You are on page 1of 198

Study of Natural Fiber Breakage

during Composite Processing

by

Carlos Jafet Quijano-Solis

A thesis submitted in conformity with the requirements


for the degree of Doctor of Philosophy
Graduate Department of Faculty of Forestry
University of Toronto

Copyright by Carlos Jafet Quijano-Solis (2015)


Study of Natural Fiber Breakage
during Composite Processing
Doctor of Philosophy, 2015

Carlos Quijano-Solis

Faculty of Forestry

University of Toronto

ABSTRACT

Biofiber-thermoplastic composites have gained considerable importance in the last century.

To provide mechanical reinforcement to the polymer, fibers must be larger than a critical

aspect ratio (length-to-width ratio). However, biofibers undergo breakage in length or width

during processing, affecting their final aspect ratio in the composites.

In this study, influence on biofiber breakage by factors related to processing conditions,

fiber morphology and the flow type was investigated through: a) experiments using an

internal mixer, a twin-screw extruder (TSE) or a capillary rheometer; and b) a Monte

Carlo computer simulation. Composites of thermomechanical fibers of aspen or wheat straw

mixed with polypropylene were studied. Internal mixer experiments analyzed wheat straw

and two batches of aspen fibers, named AL and AS. AL fibers had longer average length.

Processing variables included the temperature, rotors speed and fiber concentration. TSE

experiments studied AL and AS fiber composites under various screws speeds, temperatures

and feeding rates of the polymer and fibers. Capillary rheometers experiments determined

AL fiber breakage in shear and elongational flows for composites processed at different

ii
concentrations, temperatures, and strain rates. Finally, the internal mixer experimental

results where compared to Monte Carlo simulation predictions. The simulation focused on

fiber length breakage due to fiberpolymer interactions.

Internal mixer results showed that final fiber average length depended almost solely on

processing conditions while final fiber average width depended on both processing conditions

and initial fiber morphology. In the TSE, processing conditions as well as initial fiber

length influenced final average length. TSE results showed that the fiber concentration

regime seems to influence the effect of processing variables on fiber breakage. Capillary

rheometer experiments demonstrated that biofiber breakage happens in both elongational

and shear flows. In some cases, percentage of biofiber breakage in elongational flow is

higher. In general, simulation predictions of final average lengths were in good agreement

with experiments, indicating the importance of fiberpolymer interactions on fiber breakage.

The largest discrepancies were obtained at higher fiber concentration composites; these

differences might be resolved, in future simulations, by including the effect of fiberfiber

interactions.

iii
Acknowledgements

A long journey comes to a end for me. This journey has been not without moments of doubt
and sadness, but the path I leave now behind had far more moments of joy and satisfaction
that helped me grow as a scientist, as a person. I thank all the people and institutions who
helped me, this journey was possible because of you.

I express my deepest gratitude and admiration to Professor Ning Yan for her unparalleled
guidance. Her advice, at all times knowledgeable and focused, was pivotal to steer me in
the right direction and bring me back to concentrating on the important objectives when
I needed it the most. I am grateful for her constant support and words of encouragement
that showed me her genuine interest in my academic and personal development and success.
She is a true mentor, one from whom I was privileged to learn.

I would like to thank the members of my supervisory committee Professors Paul Cooper,
Mark Kortschot and Mohini Sain. Their interest in my development as a researcher was
perceivable through their suggestions, comments and discussions. I also thank the external
members of my defence committees, Professors Sally Krigstin and Amar Mohanty for their
valuable feedback and comments. I extend my gratitude to Dr. S. Y. (Tony) Zhang from
FPInnovations-Forintek Division. His valuable help, comments and support were essential
in helping this project start and consolidate during the first years.

I gratefully acknowledge the financial support from various institutions. NSERC and
FPInnovationsForintek Division financial support was provided through the NSERC-IPS
program. The Government of Ontario and Alpa Lumber Inc., support was provided via
the OGSST scholarship. I also acknowledge the scholarship from Forest Products Society
Eastern Canadian Section and give special thanks to the Faculty of Forestry.

The natural fibers were generously donated by Dr. Siquo Chen from Alberta Innovates-
Technology Futures (wheat straw fibers); and Dr. S. Y. (Tony) Zhang from FPInnovations-
Forintek Division (aspen fibers). The refining of the aspen fibers had the valuable assistance
of Dr. James Deng. I thank Dr. Alain Cloutier from Universite Laval and Dr. Zhirun Yuan

iv
from FPInnovations-Paprican Division for facilitating the used of their HiRes FQAs. And
I especially thank Paul Scudamore for his experienced assistance in the use of the HiRes
FQA at FPInnovations-Paprican Division.

In different stages of my experimental work I received the valuable technical assistance of Dr.
P. Syed Abthagir, Babak Doroudiani, William Garcia and Adrew Barquin. I also would like
to sincerely thank Deborah Paes, Mary-Rose Naudi, Ian Kennedy, John McCarron, Tony
Ung and Susana Diaz from the Faculty of Forestry who always patiently helped me with
my administrative and technical requirements.

Throughout all these years I was privileged to receive the constant support and encouragement
from colleagues and friends: Sabina, Sedric, Smith, Yong, Daniela, Syed, Dragica, Thierry,
Michelle, Hector, Alan, Brian, David, Claude, Darrin, Larry, Terry, Joyce and Bill I thank
you deeply.

Con mi m
as sincero amor dedico esta tesis a mi Madre, mi Padre, mis hermanas F
atima
y Elaine, y a mis sobrinas y sobrinos Cristina, Ana, Jes
us y Ricardo. Mi coraz
on siempre
ha estado con ustedes.

v
Table of Contents

Abstract ii

Acknowledgements iv

Table of Contents vi

List of Tables xi

List of Figures xii

List of Appendices xv

1 Introduction 1
1.1 Motivation and Statement of the Problem . . . . . . . . . . . . . . . . . . . 1
1.2 General Background and Fundamental Concepts . . . . . . . . . . . . . . . 3
1.2.1 Natural Fiber Composite Materials . . . . . . . . . . . . . . . . . . . 3
1.2.2 Polymer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.3 Natural Fibers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.4 Biofiber-Thermoplastic Polymer Compounding . . . . . . . . . . . . 9
1.2.4.1 Internal Mixers . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.4.2 Twin-Screw Extruders . . . . . . . . . . . . . . . . . . . . . 13
1.2.5 Rheology and capillary rheometers . . . . . . . . . . . . . . . . . . . 16
1.3 Fiber Breakage during Processing . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.1 Effect of Processing Conditions . . . . . . . . . . . . . . . . . . . . . 19
1.3.2 Effect of Initial Fiber Dimensions . . . . . . . . . . . . . . . . . . . . 23
1.3.3 Effect of the Type of Flow . . . . . . . . . . . . . . . . . . . . . . . . 25

vi
1.3.4 Simulation of Fiber Breakage . . . . . . . . . . . . . . . . . . . . . . 26
1.4 Summary and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.5 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2 Hypothesis and objectives 30


2.1 Research Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Effect of mixing conditions and initial fiber morphology on fiber dimen-


sions after processing 32
3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.2 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.3 Composites processing . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.4 Morphological measurements . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4.1 Processing variables and fiber properties . . . . . . . . . . . . . . . . 39
3.4.2 Multivariate data analysis . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.5.1 Loadings plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5.2 Scores plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.5.3 Predictors influence on fiber properties . . . . . . . . . . . . . . . . . 49
3.5.3.1 Lw and Delta-Lw . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.3.2 Wn and Delta-Wn . . . . . . . . . . . . . . . . . . . . . . . 52
3.5.3.3 L/Dn and Delta-L/Dn . . . . . . . . . . . . . . . . . . . . . 54
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

vii
4 Characterization of biofiber breakage in composite processing using a
capillary rheometer 57
4.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3.2 Processing equipment . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3.3 Composite processing . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3.4 Materials characterization . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3.4.1 Morphological measurements . . . . . . . . . . . . . . . . . 63
4.3.4.2 X-ray microtomography imaging . . . . . . . . . . . . . . . 63
4.4 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.5.1 Effect of shear flow and elongational flow . . . . . . . . . . . . . . . 70
4.5.2 Effect of fiber concentration . . . . . . . . . . . . . . . . . . . . . . . 74
4.5.3 Effect of processing temperature . . . . . . . . . . . . . . . . . . . . 78
4.5.4 Effect of shear/stretch rate . . . . . . . . . . . . . . . . . . . . . . . 79
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

5 Effect of processing conditions on the extent of biofiber breakage in a


twin-screw extruder 82
5.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3.2 Composites Processing . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.3 Materials Characterization . . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.4.1 Processing Variables and Fiber Properties . . . . . . . . . . . . . . . 88
5.4.2 Projection to Latent Structures (PLS) . . . . . . . . . . . . . . . . . 91
5.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

viii
5.5.1 Loadings Plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.5.2 Scores Plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.5.3 Factor Influence on Lw and Delta-Lw . . . . . . . . . . . . . . . . . 96
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

6 Monte Carlo simulation of biofiber breakage during processing in an


internal mixer 107
6.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3 Outline of the simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.4 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.5 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.5.1 Fibers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.5.2 Polymer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.6 Simulation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.6.1 fiber concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.6.2 RPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.6.3 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.6.4 Simulation processing . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.7 Technical description of the simulation . . . . . . . . . . . . . . . . . . . . . 122
6.8 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

7 Conclusions and Recommendations 142


7.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2 Recommendations for Further Work . . . . . . . . . . . . . . . . . . . . . . 146

References 148

ix
Appendices 167
Appendix A Chapter 3 processed samples data . . . . . . . . . . . . . . . . . . 167
Appendix B Chapter 4 processed samples data . . . . . . . . . . . . . . . . . . 170
Appendix C Chapter 5 processed samples data . . . . . . . . . . . . . . . . . . 171
Appendix D Listing of a section of the source code . . . . . . . . . . . . . . . . 175

x
List of Tables

1.1 Natural fiber breakage during processing in internal mixers. . . . . . . . . . 21


1.2 Natural fiber breakage during processing in twin-screw extruders. The tem-
peratures indicate the temperatures in the mixing section. . . . . . . . . . . 22

3.1 Factors in the X- and Y -spaces . . . . . . . . . . . . . . . . . . . . . . . . . 41


3.2 Fiber dimensions before processing. Length is reported in mm and width in
m. Aspect ratio and polydispersity index values are dimensionless. . . . . 44

4.1 Polypropylene shear and elongational viscosities at all processing tempera-


tures and shear rates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.1 Twin-screw extruders processing temperature profiles. Temperatures in C 87


5.2 Experimental design of aspen long, AL, samples extruded in the twin-screw
extruder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.3 Experimental design of aspen short, AS, samples extruded in the twin-screw
extruder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6.1 Constants for the power-law and shift factor equations. . . . . . . . . . . . . 116
6.2 Simulation design for each fiber type. . . . . . . . . . . . . . . . . . . . . . . 118

A.1 Data for all AL samples processed in the internal mixer . . . . . . . . . . . 167
A.2 Data for all AS samples processed in the internal mixer . . . . . . . . . . . 168
A.3 Data for all WS samples processed in the internal mixer . . . . . . . . . . . 169
B.4 Data for all AL samples processed in the capillary rheometer . . . . . . . . 170
C.5 Data for all AL samples processed in the twin-screw extruder . . . . . . . . 171
C.6 Data for all AS samples processed in the twin-screw extruder . . . . . . . . 173

xi
List of Figures

1.1 Polymer units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5


1.2 Hierarchical structure of plant fibers [48] . . . . . . . . . . . . . . . . . . . . 8
1.3 Schematic of dispersive and distributive mixing [53] . . . . . . . . . . . . . . 10
1.4 Schematic of shear flow: a) drag flow, b) pressure driven flow . . . . . . . . 10
1.5 Schematic of elongational flow . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Mixture quality vs. degree of fill at three rotor speeds [42] . . . . . . . . . 13
1.7 Twin-screw extruder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.8 Screw elements for (a) co-rotating twin-screw extruders and (b) counter-
rotating twin-screw extruders [62] . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1 Schematics of the fiber extraction process. . . . . . . . . . . . . . . . . . . . 38


3.2 Values of R2 and Q2 for each fiber property . . . . . . . . . . . . . . . . . . 45
3.3 Loading plot of principal components 1 and 2 . . . . . . . . . . . . . . . . . 46
3.4 Scores plot of principal components 1 and 2. N AS,  AL and  WS . . . . 48
3.5 Coefficient plot for the final average fiber length, Lw, property . . . . . . . 50
3.6 Coefficient plot for the Delta-Lw property . . . . . . . . . . . . . . . . . . . 51
3.7 Coefficient plot for the final average fiber width, Wn, property . . . . . . . 52
3.8 Coefficient plot for the Delta-Wn property . . . . . . . . . . . . . . . . . . . 53
3.9 Coefficient plot for the final average fiber aspect ratio, L/Dn, property . . . 54
3.10 Coefficient plot for the Delta-L/Dn property . . . . . . . . . . . . . . . . . . 55

4.1 A schematic of the capillary rheometer with long and zero-length dies . . 61
4.2 Pressure drop readings for polypropylene processed at 190 C . . . . . . . . 66
4.3 Pressure drop readings for a sample with 30 wt.% fibers processed at 500 s1
and 190 C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

xii
4.4 Percent of Lw reduction for samples with 10 wt.% fibers processed through
the 16-mm length die . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Percent of Lw reduction for samples with 10 wt.% fibers processed through
the zero length die. Horizontal bar indicates a difference with shear rate. 71
4.6 Percent of Lw reduction for samples with 30 wt.% fibers processed through
the 16-mm length die . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.7 Percent of Lw reduction for samples with 30 wt.% fibers processed through
the zero length die. Horizontal bar indicates a difference with shear rate. 72
4.8 Microtomography of the 10 wt.% fiber composite processed in the capillary
rheometer at 1500 s1 and 190 C through the 16-mm length die . . . . . . 75
4.9 Microtomography of the 30 wt.% fiber composite processed in the capillary
rheometer at 1500 s1 and 190 C through the 16-mm length die . . . . . . 75
4.10 Microtomography of the 30 wt.% fiber composite processed in the capillary
rheometer at 500 s1 and 190 C through the 16-mm length die . . . . . . . . 76
4.11 Microtomography of the 30 wt.% fiber composite processed in the capillary
rheometer at 1500 s1 and 190 C through the zero-length die . . . . . . . 76

5.1 Values of R2 and Q2 for each fiber property. PC1: Principal component 1.
PC2: Principal component 1. PC3: Principal component 3. . . . . . . . . . 94
5.2 Loadings plot of principal components 1 and 2. . . . . . . . . . . . . . . . . 95
5.3 Scores plot of principal components 1 and 2.  AL and N AS. . . . . . . . 97
5.4 Contribution plot between the groups of AL and AS fiber samples. . . . . . 98
5.5 Contribution plot between the groups of AS fiber samples. . . . . . . . . . . 98
5.6 Coefficients plot for the final average fiber length, Lw, property. . . . . . . . 99
5.7 Variable importance on projection. . . . . . . . . . . . . . . . . . . . . . . . 101
5.8 Analysis of AL samples only. Coefficients plot for the final average fiber
length, Lw, property. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.9 Analysis of AS samples only. Coefficients plot for the final average fiber
length, Lw, property. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.10 Coefficients plot for the Delta-Lw property. . . . . . . . . . . . . . . . . . . 105

xiii
6.1 Internal batch mixer with roller type blades. . . . . . . . . . . . . . . . . . . 112
6.2 Thermogram for aspen and wheat straw fibers. . . . . . . . . . . . . . . . . 115
6.3 Master curve for the shear rate-temperature dependence of viscosity. . . . . 116
6.4 Normalized cumulative distribution of the predicted final weight-average fiber
length, Lw, for the simulation of 10 wt.% AL fiber composites at 60 RPM,
180 C and with fibers broken in half. . . . . . . . . . . . . . . . . . . . . . . 129
6.5 Comparison between prediction and experimental data of the final weight-
average fiber length, Lw, for AL fiber composites. . . . . . . . . . . . . . . . 131
6.6 Comparison between prediction and experimental data of the final weight-
average fiber length, Lw, for AS fiber composites. . . . . . . . . . . . . . . . 132
6.7 Comparison between prediction and experimental data of the final weight-
average fiber length, Lw, for WS fiber composites. . . . . . . . . . . . . . . 133
6.8 Comparison between prediction and experimental data of the final average
fiber width, Wn, for AL fiber composites. . . . . . . . . . . . . . . . . . . . 135
6.9 Comparison between prediction and experimental data of the final average
fiber width, Wn, for AS fiber composites. . . . . . . . . . . . . . . . . . . . 136
6.10 Comparison between prediction and experimental data of the final average
fiber width, Wn, for WS fiber composites. . . . . . . . . . . . . . . . . . . . 137
6.11 Comparison between prediction and experimental data of the final average
fiber aspect ratio, L/Dn, for AL fiber composites. . . . . . . . . . . . . . . . 138
6.12 Comparison between prediction and experimental data of the final average
fiber aspect ratio, L/Dn, for AS fiber composites. . . . . . . . . . . . . . . . 139
6.13 Comparison between prediction and experimental data of the final average
fiber aspect ratio, L/Dn, for WS fiber composites. . . . . . . . . . . . . . . 140

xiv
List of Appendices

Appendix A. Chapter 3 processed samples data . . . . . . . . . . . . . . . . . . . . . 167


Appendix B. Chapter 4 processed samples data . . . . . . . . . . . . . . . . . . . . . 170
Appendix C. Chapter 5 processed samples data . . . . . . . . . . . . . . . . . . . . . 171
Appendix D. Listing of a section of the source code . . . . . . . . . . . . . . . . . . . 175

xv
Chapter 1
Introduction

1.1 Motivation and Statement of the Problem

Natural fibers have been used in composite materials since ancient times, for example, more
than 3000 years ago mixtures of clay and straw were used in buildings in ancient Egypt.
The advent of new materials, like metals, prompted a decrease in their use [1]. The past
decades, however, have seen a renaissance in the desire to use composites with natural
fibers for commercial products. This trend is happening, to some extent, again because
of petroleum, as its reserves and price fluctuation affect the cost of materials. Another
important factor that is driving this renewed interest in the use of biofibers is the increasing
public awareness on protecting the ecology of the planet. Thus the need to use renewable
and sustainable materials. Governments are also doing their part; for example, in Europe
the end of life vehicle directive (European Union directive 2000/53/EC) was enacted, which
seeks to decrease the environmental impact of end-of-life vehicles by stating that by 2015 a
minimum of 95 % in weight of the vehicle should be able to be recovered.

One of the areas where biofibers are been studied intently is the polymer composite ma-
terials, where biofibers can be used as reinforcement. Some of the plant fibers studied as
reinforcement include: wood, kenaf, flax, hemp, sisal, jute, ramie and henequen. Other

1
2

fibers that are gaining importance are the ones obtained from pineapple leaf, bamboo, rice
husk and wheat straw. Ecological-friendliness is only one of the advantages. Biofibers have
good specific mechanical properties of strength and toughness [2] and lower density than
synthetic fibers (e.g. glass, carbon or aramid) which reduces the weight of final products.
The automotive industry is the one that has been more involved in the development of
natural fiber reinforced polymeric composites. Other industries include building and con-
struction, furniture and consumer goods. Update review literature has been carried out to
study the incorporation of natural fibers into plastics derived from petroleum [3; 4] as well
as biobased polymers [510]. Some plastics used for this purpose include phenol formalde-
hyde, epoxy resins as well as various commodity plastics (especially polypropylene, PP,
polyethylene, PE, and polystyrene, PS).

Natural fiber-polymer composites have various advantages. However, their success in com-
mercial markets will depend on the reliability that the developed product can successfully
perform as it is intended. One of the most important performing criteria is that the com-
posites possess adequate mechanical properties. A composites mechanical properties are
the result of a interdependence between the composite components and their interactions
on the one hand, and the final composite microstructure (e.g. orientation and dispersion of
fibers in the polymer matrix) on the other.

Fibers reinforce along the direction of their largest dimension (length). The parameter that
characterizes the efficiency of the stress transfer from the polymer to the fiber is the aspect
ratio, which is defined as the ratio of the fiber length to the fiber diameter. Mechanical
properties, such as strength and toughness, increase with larger aspect ratios [1114]. How-
ever, a minimum (critical) aspect ratio is required in order to take full advantage of the
reinforcement capabilities of the fibers. Critical aspect ratios depend on the fiber-matrix
interfacial bond and the fibers ultimate strength [15].

Due to the inherent incompatibility between natural fibers, which are hydrophilic, and most
commercial important polymers, which are hydrophobic, a vast research has been aimed
at improving the interfacial biofiber-polymer interaction. Natural fibers and a polymeric
3

matrix could result in composites with poor mechanical properties [16; 17] and sometimes
with even lower mechanical properties than the polymer alone [18]. Extensive research is
carried out on interface-active additives, which have proven effective to make fibers more
compatible with the polymer [19; 20]. The preferred methods to increase this biofiber-
polymer compatibility have used maleic anhydride grafted polymers [19; 2124] and silane-
based coupling agents [25; 26].

On the other side, less research has been dedicated to study other factors that also affect the
critical fiber aspect ratio. One of them is the dimension of natural fibers after processing.
Biofiber-thermoplastic composites are generally manufactured using traditional thermoplas-
tic composites processing techniques, such as mixing, twin-screw extrusion, injection and
compression molding. A major issue, however, is that these pieces of equipment impart such
strong dynamics on the composites that natural fibers undergo breakage. Fibers length re-
duction is known as a major problem during processing. Also, since biofibers are, to a
great extent, bundles of elementary fibers, separation of fiber bundles with the consequence
reduction in the fibers width have been reported. The evolution of aspect ratio during
processing is, therefore, apparent. If final aspect ratios are smaller than the ones for un-
processed fibers, as is usually the case, insufficient fiber reinforcement would occur. Many
factors influence fiber breakage and in this work, the effect of various processing conditions
as well as the initial fiber morphology on fiber breakage was studied.

1.2 General Background and Fundamental Concepts

1.2.1 Natural Fiber Composite Materials

The combination of natural fibers, also known as biofibers, with a polymeric matrix is called
a composite. Natural fibers were the original reinforcement in polymers, which started
in the early twentieth century with the mixture of cellulose in phenolic resins. Use and
research on biofibers in composites have continued since [1]. However, the discovery of
nylon fibers by DuPond, a petrochemical polymer, and the start of the industrial mass
4

production of glass fibers, both in the 1930s, gradually displaced natural fibers and they
saw a sharp decrease to a very low market share since the 1960s. The last century has
seen a resurgence in a worldwide interest in the use of biofibers even bringing the United
Nations General Assembly to declare 2009 the International Year of Natural Fibers, with
the principal focus to increase their use by raising global awareness. The field of natural
fiber polymer composite materials is extensive. One of the areas encompasses the mixtures
of natural fibers with thermoplastic polymers.

The automotive industry has been the most active in their efforts to incorporate natural fiber
composites as replacement to current materials. The natural fiber composites developed
for cars show additional advantages to those mentioned in the previous section, such as
sound insulation, lower production costs, thermal comfort and reduced weight (compared
to the part that would be replaced) that turns into better fuel consumption efficiency a
25 % weight reduction saves on average 250 million barrels of crude oil [2730]. All major
car manufacturers currently use natural fiber composites in various applications. Some
non-structural applications include door panels, boot liners, seat backs, headliners and
dashboards. Research of structural parts, like the frontal bonnet of an off-road vehicle,
has been carried out [31]. DaimlerChrysler claimed to be the first in using natural fiber
composites for engine covers [32].

Research has also led, for example, to the mixture of wood fibers in thermoplastics for
commercial applications of decking and fencing, such as Trex[33] and Strandex. The
building and construction industries are also very involved in the development of these
type of materials that could replace some existing ones [34; 35]. Consumer products with
good durability could also be developed, such as furniture and sporting equipment [5]. In
particular, promising uses for biocomposites could be found in the short life-span consumer
products, such as disposable packaging and utensils. Although in the current modern society
not all short life-span articles are disposable, especially in the electronics industry many
products are discarded not because of failure but just due to mere style obsolescence, e.g.
computers, digital audio and video players [36]. Nowadays, the life-span of cell phones is
on average just 20 months in some countries [37].
5

In general, various products made with synthetic fibers could be replaced by natural fibers
and give products with good durability that have less environmental impact and that can
be more easily recycled [38].

1.2.2 Polymer

Polymer is a word that derives from the (transliterated) Greek terms polus, which means
many, and meros, which means parts. Literally, it means many parts. Polymers are
very large molecules that consist of one or several chemical units that are repeated many
times (103 + times) in the molecular chain (Figure 1.1). Polymers that have only one chemi-
cal unit are called homopolymers as opposed to copolymers (also known as heteropolymers),
which have two or more units in the polymer chain.

Figure 1.1: Polymer units

Polymers can be natural or synthetic. Cellulose, hemicellulose and lignin, which are the
primary chemical components of a plant (see section 1.2.3) are all naturally occurring poly-
mers, where cellulose is the most abundant biomass on earth. Synthetic polymers are
derived from petroleum. They can be classified as thermoplastic, thermosets and elas-
tomers. Thermoplastics and thermosets are commonly known as plastics. Thermoplastics
are polymers that can be melted when heated, solidify upon cooling and be re-melted again
(e.g. polypropylene, polyethylene, polystyrene). In thermosets, the polymer molecules are
chemically crosslinked and once formed they cannot be melted again (e.g. polyurethanes,
phenol-formaldehyde, vulcanized rubber). Elastomers are polymers with some degree of
crosslinking such that the molecules still have high degree of flexibility that allows them to
deform and recover their initial shape (e.g. silicone rubber, polybutadiene).
6

Depending on the complexity, conformation and arrangement of the molecular chains, a


thermoplastic polymer, when in its solid state, can have regions of sufficient structural reg-
ularity that allow the formation of crystallites [39]. The rest of the molecules instead, will
remain in an rigid disordered arrangement that give amorphous regions. Thermoplastics,
with both crystalline and amorphous zones, are called semicrystalline polymers. Ther-
moplastics that do not have regions of crystallinity are called amorphous polymers. Both,
crystallites and amorphous zones, have characteristic temperatures that turn the solid poly-
mer into a melt. The crystalline regions have a melting temperature. Since crystals in a
polymer are not uniform or perfect there is a range of melting temperatures, Tm . Polymer
technical data give only one melting temperature that corresponds to the peak temperature
of a DSC curve, which for many materials represents the temperature where most crys-
tallites are melted [40]. Amorphous regions have instead a glass transition temperature,
Tg , below which they are hard and brittle. Glass transition temperatures are lower than
melting temperatures. Above the Tg semicrystalline polymers are rubbery and amorphous
polymers melt [40; 41].

A great number of different polymers exist and they are ubiquitous in our modern society.
They can be found in a great variety of consumer, structural and advanced products. Nev-
ertheless, only a few of them are consumed in very high volume; these are called commodity
plastics. Some examples are polypropylene, polyethylene and polystyrene. Polymers are
seldom sold without additives (plasticizers, pigments, colorants) and they can be mixed
with different solids to create composites (see section 1.2.3). The mixture of a polymer
with a solid particle is called a compound and the process of mixing them is referred to as
compounding. This is opposed to blends and blending, which refer to the mixture of
different polymers [42].

1.2.3 Natural Fibers

Reinforcement of a material occurs when fibers effectively strengthen and/or stiffen the
composite. There are various factors related to the fibers that influence their level of
7

reinforcement, e.g., their orientation and concentration in the composite. Fiber length,
L, is another important factor, such that a critical length, Lc is required for an adequate
transfer of the applied load to the fibers by the matrix according to [15]:

f D
Lc = (1.1)
2

where f is the fibers ultimate (tensile) strength and is the fiber-matrix bond strength.
Equation 1.1 refers to fibers with a round cross-section, with a diameter of value D. There-
fore, there are two characteristic physical dimensions in a fiber, its length and diameter.
The ratio L/D is called the fibers aspect ratio and a critical aspect ratio could be referred
to when D is on the left side of Equation 1.1. Natural fibers morphological cross-sections
are variable and irregular; however, it is not uncommon to assume it is circular, especially
in mechanical test studies [4345].

Natural fibers in this study were extracted from plant fibers. An elementary fiber is a single
plant cell. The field of natural fiber polymer composites use mostly fiber bundles, which
are known as technical fibers [46]. These bundles are elementary fibers bonded together.
Throughout this study the term fiber refers to technical fibers. Also, with regard to the
scale, in this work macroscopic fibers were used, as opposed to the field of nanofibers [47].

Plant fibers, unlike synthetic fibers, are not single units of one component but rather they
have complex compositions. Fibers are themselves composites of natural polymers, in which
each fiber consists primarily of rigid microfibrils of a semicrystalline polymer, cellulose,
embedded in a matrix composed of amorphous polymers: lignin, hemicellulose, pectin [49].
These fibers are also known as lignocellulosic fibers.

The hierarchical structure of an elementary plant fiber starts with the cellulose polymeric
chain. The cellulose molecule consists of thousands of D-glucose monomeric units chemically
linked into long linear chains. Groups of cellulose molecules are connected together, through
hydrogen bonds, into the microfibril bundles [50]. Microfibrils have crystalline regions
(micelles) and amorphous regions. Groups of microfibrils cemented together by amorphous
8

Figure 1.2: Hierarchical structure of plant fibers [48]

polymers form macrofibrils. Macrofibrils are, in turn, assembled and glued together with
additional non-cellulosic amorphous polymers to form the walls of the plant cells. Groups
of plant cells are joined together to form plant tissue. A schematics of the hierarchical
structure of a plant cell is shown in Figure 1.2.

Depending on the source, natural fibers are classified as follows: bast or core fibers (flax,
jute, hemp, kenaf), leaf fibers (sisal, pineapple, agave, banana, abaca), seed fibers (cotton,
kapok, coir), grass and reed fibers (wheat, corn, rice, bamboo) and all other fibers (wood,
roots) [3; 51]. Fibers can also be classified according to their length as short or long. Short
fibers are the individual (technical or elementary) plant fibers whereas long fibers are short
fibers that are secure together, either industrially or by hand, into rovings [52]. In this
work, the breakage of short fibers during processing was studied.
9

1.2.4 Biofiber-Thermoplastic Polymer Compounding

The process of obtaining a thermoplastic compound that is ready to be shaped into final
form involves five elementary processing operations [53]: 1) handling of particulate solids,
2) melting, 3) mixing, 4) devolatilization and 5) pressurization and pumping. Mixing is
the process that improves the uniformity of a mixture. Mixing can happen by diffusion,
turbulent motion or convective motion. Polymer melts and their suspensions (the mixture
polymer-solids is referred to as a suspension when the polymer is in the molten state) have
very high viscosities (see section 1.2.5) . Typically, commercial polymers viscosities are in
the range of 102 Pas or higher compared to waters viscosity of 103 Pas [54]. Due to their
high viscosities, molecular diffusion coefficients in polymer melts are very small and can be
considered negligible. Also, during polymer melts flow, viscous forces continually prevail
over inertial forces, ensuring that the flow is always laminar [55]. Therefore, convection is
the only important mixing mechanism in thermoplastic melts. Convective motion refers to
the movement of groups of both fluid particles and solids. There are two types of convective
mixing: dispersion and distribution.

Dispersive mixing brings about the disaggregation of a cohesive component into smaller
parts, for example, the break up of fiber agglomerates into smaller flocs and/or in-
dividual fibers (Figure 1.3). It is also referred to as intensive mixing. The cohesive
forces that hold the smaller parts together must be surpassed by the hydrodynamic
forces of the polymer melt if dispersive mixing is to happen. Mixing forces during
flow are influenced by: a) the polymers viscosity, and b) the type of flow.

In laminar mixing, shear and elongational flows will generate the forces that are
conducive to dispersive mixing. A shear flow field is one in which the direction of
the fluids velocity is perpendicular to the velocity gradient (Figure 1.4). This is
in contrast to an elongational flow field where the direction of the fluid velocity is
parallel to the velocity gradient (Figure 1.5). Shear flows can be generated by a
moving boundary or a pressure gradient in constant width geometries. Elongational
flows can be generated by pressure gradients in contraction geometries [56]. Studies
10

have shown that elongational flows generate higher hydrodynamic forces and thus
are much more effective to cause dispersive mixing than shear flows [42]. The higher
hydrodynamic forces in elongational flow are mainly due to elongational viscosities
are much higher than shear viscosities (see section 1.2.5).

Figure 1.3: Schematic of dispersive and distributive mixing [53]

(u=V at y=h) (u=0 at y=h)


V
u(y)
u(y) velocity u P0 h
h gradient = y V P1
y y
x x
(u=0 at y=0) (u=0 at y=0)
a) Moving boundary (drag) flow b) Pressure driven flow (P0 > P1)
(max velocity = V) (max velocity = V)

Figure 1.4: Schematic of shear flow: a) drag flow, b) pressure driven flow

Distributive mixing involves the continuous spatial rearrangement, over a period of time,
of the components of the mixture (Figure 1.3). This is carried out by imposing on
the suspension a continuos and permanent deformation through shearing (shear flow),
stretching and squeezing (both elongational flows). Thus the important variable in
distributive mixing is the strain undergone by the material and not the stresses (forces)
imparted during the mixing. Dispersive mixing is also called extensive mixing [53].
11

(u=0 at wall)

velocity u u(x)
gradient = x P0 V P1

y
x
(u=0 at wall)
Pressure driven flow in contraction (P0 > P1)
(max velocity = V)

Figure 1.5: Schematic of elongational flow

These two mixing mechanisms do not necessarily happen separately in processing equip-
ment. Reduction in size of a solid component may also happen in the mixing zones where
distributive mixing is the predominant mechanism, while dispersive mixing will always in-
volve the occurrence of some distributive mixing as well. Numerous pieces of equipment
for mixing are available from the many machine manufacturers worldwide. Internal mixers
are used to compound biofiber-thermoplastic composites in both, laboratory analyses as
well and in industry. Nevertheless, the dominant mixing devices in industry are twin-screw
extruders.

1.2.4.1 Internal Mixers

The first single-rotor internal mixer was developed in the first half of the nineteenth century
and the first twin-rotor mixer was patented and manufactured around 1880 [57]. Since then,
rotor designs have varied, all of them aimed at improving mixture quality [42]. A typical
internal mixer is shown in Figure 6.1. The internal mixer is a ubiquitous device in the
processing laboratory of thermoplastic composites due to its many advantages:

low amount of materials is required for every recipe


easy to change recipes for composites
feed stock in various forms and large clumps can be used
12

precise control of amount of material fed without the need to use expensive gravimetric
or volumetric feeders
it has a well-defined residence time
easy to vary processing procedures
wide range of processing procedures can be evaluated
easy cleaning

The internal mixer has two characteristic geometric zones. A narrow space between the
rotor wings tip and the internal chambers wall and large areas between rotors. During
mixing, materials continuously move between these two zones. The narrow gaps provide
a high-shear region while the larger area provides low-shear. Therefore, the larger zone is
considered to generate basically distributive mixing while dispersive mixing is considered to
happen chiefly in the vicinity of the narrow gaps [58]. However, there is evidence of some
dispersion of chopped glass fibers outside the regions of high stresses [59]. In addition to the
high-shear regions developed in the gaps, visualization studies on internal mixers have also
shown strong elongational flows in the convergent regions before the gaps, as the material
is stretched to enter the rotor tips [60].

The degree of fill and the rotors speed in the internal mixer strongly influence the quality
of the mixing. Studies have shown that the optimum degree of fill should be in the range of
70 and 80 v.% [42; 60]. In this range the most favourable pressure gradients are developed
inside the internal mixer. Larger pressure is generated in the leading edge of the rotors and
lower pressure at their receding edge; this has been confirmed visually and with simulations
[58; 60]. These pressure gradients allow an adequate interchange of material between the
rotors, between rotor wings and also help the material to pass through the gap regions.
Below this fill range not enough pressure is developed and above the range, the chamber is
too full to generate a proper pressure gradient. An example of the various mixture qualities
for suspensions of carbon black processed at several degrees of fill and rotor speeds is shown
in Figure 1.6.
13

poor 6
25 rpm
110 C
4

Mixture 140 C
quality
2
170 C

rpm
63 rpm

38
200 C
best 0
0.3 0.5 0.7 0.9
Degree of fill

Figure 1.6: Mixture quality vs. degree of fill at three rotor speeds [42]

1.2.4.2 Twin-Screw Extruders

Twin-screw extruders (TSE) are by far the most common pieces of compounding equipment
that are used in the natural fiber-thermoplastic composite industry. They carry out the
task of continuous production of the mixture. The twin-screw extruder used in this work
is shown in Figure 1.7. Material transport from the feeding entrance to the end of twin-
screw extruders is, for the most part, accomplished by positive displacement [61]. This is
opposite to the material transport occurring in a single-screw extruder, which depends on
the friction and viscous drag of the solids and melt, respectively [62]. The degrees of positive
conveying flow generated in a twin-screw extruder vary according to the purpose and type
of the extruder. Positive conveying characteristics are given by the degree of intermeshing
between the screws. In fully intermeshing equipment, the flights of one screw fit completely
into the channel of the adjacent screw. The better the seal provided by the screws system
the less back leakage flow and more positive conveying. Non-intermeshing TSE do not have
positive conveying and their throughput depends on the friction between the material and
the barrel, in a similar fashion to single-screw extruders [63; 64].
14

There are two major types of TSE compounding extruders and the TSE used for profile
extrusion. Compounding TSE do not generate large positive pressures and their rotating
speeds can be very high. In comparison, TSE used in profile extrusion can generate high
and uniform pressures and their rotating speeds are low. TSE have enormous flexibility due
to their modular design in both the screws (see examples in Figure 1.8) and the barrel.

Figure 1.7: Twin-screw extruder

Compounding twin-screw extruders can be co-rotating or counterrotating, with a larger per-


centage of co-rotating screws used in industry. The screws can be fully intermeshing, par-
tially intermeshing, non-intermeshing and also non-intermeshing with distance [42]. Screw
speeds can be very high, usually around 200-500 but speeds can reach as high as 1400 rpm
is some equipment [42]. Such high speeds provide good dispersive and distributive mixing
due to the good deformation rates the material experience inside the extruder during the
mixing time. Good dispersive mixing and/or good distributive mixing also depend on the
15

effective selection of mixing elements. A good mixing element will generate forward, reverse
and circumferencial flows. White and Kim [65] correlated the different flows to the mixing
that the fluid elements experience. Reverse flow promotes distributive mixing because the
flow is recirculated in the mixing element while circumferential flow promotes dispersive
mixing because the material is forced to pass through the small gap regions created by
the mixing elements geometry. In the vicinity of the gaps, the converging zones lead to
strong elongational flows [66]. Two types of narrow gaps are formed: a) between the mixing
elements tips and the barrel, and b) in-between adjacent mixing elements. The gaps found
in-between mixing elements generate the highest deformation rates and elongational flows
in co-rotating screws since, at the gap, the two surfaces move in opposite direction.

(a) (b)

Figure 1.8: Screw elements for (a) co-rotating twin-screw extruders and (b) counter-
rotating twin-screw extruders [62]

Elongational flows from converging zones and recirculation flows are often taken into account
during the design of screws mixing elements [67]. Nevertheless, the flow patterns generated
by mixing modules are transient and very complex and to fully assess mixing quality in
screw design not only the flows generated (shear and elongational) should be analyzed but
also the trajectory of the fluid particles in the flow field [68].

Twin-screw extruders have another distinctive characteristic. The material is fed into the
extruders in a starve fed fashion. In starve-feeding the material does not fill the screws
(flood feeding), as is carried out in single-screw extrusion. Starve-fed conditions provide
various benefits. First, the melting of the thermoplastic is faster than in flood feeding as the
16

pellets get to be immersed in a polymer melt and experience a more efficient heat transfer.
Also, there is no development of very high pressures along the extruder, which would be
detrimental to the mixing. High pressures compress the material and induce the formation
of agglomerates or prevent their break-up. In starve feeding the extruder throughput is
determined by the feeders. This allows the analysis of different conditions, for example,
the variation of screw speed at different throughputs or vice versa, by studying different
throughputs using the same screw speeds [62]. Due to these advantages, starve feeding has
also been studied as an alternative for single-screw extrusion [69; 70]. In summary, there are
great possibilities of control of the extrusion process and mixing in twin-screw extruders.

1.2.5 Rheology and capillary rheometers

The compounding of natural fiber-thermoplastic composites requires that at some point,


after the polymer is fed into the processing equipment, the fibers be in contact with the
polymer in the fluid state. The flow of polymer melts and suspensions is very complex
[61]. The discipline that studies the flow and deformation of materials is rheology. The
most important rheological flow property is the viscosity. Viscosity is the resistance to
flow, in either shear or extension. In a rigorous sense, the term viscosity alone refers to
the resistance to shear flow. The resistance to extensional flow is named extensional (or
elongational) viscosity.

Viscosity. There are two basic types of simple shear flow. Drag flow occurs when one of
the boundaries moves while the other is stationary. Pressure driven flow comes about
when there is a pressure differential between the two ends of a channel, e.g fluid in a
tube. Schematics of the two flows are shown in Figure 1.4. Viscosity, , is defined as
the ratio of the imposed shear stress, , to the shear rate, :

F du
= , = , = (1.2)
A dy
Shear stress is defined as the force, F , that is applied divided by the area, A, and has
units of Pascals, Pa, (in SI units). In shear flow, the force is applied tangentially to
17

the area and is perpendicular to the velocity gradient. The shear rate is the velocity
gradient and has units of reciprocal seconds. The units of viscosity are, accordingly,
Pas. The viscosity of fluids with simple molecules and low molecular weight (e.g
water, air, benzene, xylene) does not change with the shear rate. These liquids are
called Newtonian fluids. Polymers, on the other hand, have very long and complex
molecular structures with very high molecular weights. This characteristic causes
their viscosities to change with the shear rate. Fluids whose viscosity is shear rate
dependent are called non-Newtonian. In the case of thermoplastics, at very low shear
rates (aprox. < 4s1 ) their behavior is Newtonian; as the shear rates increase
their viscosity decreases. The decrease in viscosity with shear rate is called shear-
thinning and is attributed to polymer molecules disentanglement and realignment in
the direction of flow. Therefore, the higher the shear rates the less resistance to flow.
The shear-thinning zone is commonly represented in a log-log plot by the power-
law equation = m0 n1 . m0 and n are the consistency and power-law indices,
respectively. Lower values of n indicate a faster decrease of viscosity with shear
rate. Another important factor that influences viscosity is temperature. Viscosity
decreases exponentially with temperature. This behavior is commonly represented by
an Arrhenius type equation [71]. The effects of shear rate and temperature can be
combined into one equation, which it is known as a master curve, e.g. see Equation
6.5.

Elongational (extensional) viscosity. In a similar manner to viscosity, the elongational


viscosity, e , is defined as the ratio of the applied elongational stress, e , to the
extensional rate, :

e Fe du
e = , e = ,  = (1.3)
 A dx

The extensional force, Fe , is applied perpendicular to the area and is tangential to


the velocity gradient. All fluids, Newtonian and non-Newtonian exhibit resistance to
extensional flow. The ratio of extensional to shear viscosity is called the Troutons
ratio, and is equal to 3 for Newtonian fluids, independently of the shear or extensional
18

rates [54]. The Troutons ratio for polymers and suspensions, however, is much higher,
values of 103 or 104 can be found, and also is shear and extensional rates dependent.

During processing, shear rates vary from a range of (approximately) 10 to 103 s1 in ex-
trusion and internal mixers to as high as 103 to 105 s1 in injection molding [54]. At these
shear rates, the viscosity shows shear-thinning (power-law) behavior. The most common
piece of equipment that is used to characterize the viscosity in the power-law region at
different temperatures is the capillary rheometer. In a capillary rheometer the polymer or
suspension melt is forced to flow from a large reservoir into a capillary (typical diameters are
1, 2 mm). The shear rate is calculated from the flow rate. A pressure transducer is located
just before the capillary to record the total pressure drop experienced by the material as it
is processed. The total pressure drop is caused by two factors. The first factor is the flow
resistance in the capillary die. The second factor is due to the materials resistance to pass
from the wide barrel into the capillary die. The latter is often referred to as the excess or
entrance pressure drop. The shear stress is measured from the pressure drop experienced
in the capillary die.

A popular method to separate the pressure drop in the capillary die from the entrance
pressure drop is to concurrently run tests in two capillaries. One capillary with an almost-
zero length from which the entrance pressure drop is calculated. A schematic of a two-bore
capillary rheometer is shown in Figure 4.1. Cogswell was the first to established that the
excess pressure drop is due to the elongational viscosity of the flow [72; 73]. Cogswell theory
has been extensively verified by experiments and simulations and currently the Cogswells
method is widely used to obtain extensional viscosities, stresses and strain rates from cap-
illary rheometers [7477]. Elongational strain rates are around one order of magnitude
lower than the shear rates in capillary rheometers. Detailed theoretical background on the
capillary rheometer can be found elsewere [71].
19

1.3 Fiber Breakage during Processing

1.3.1 Effect of Processing Conditions

Fiber breakage during processing have been acknowledged to be an important problem


and many studies on synthetic fiber breakage have been published. Research works have
analyzed the effect that processing conditions and different sections of processing equipment
have on the damage of synthetic fibers when processed in injection molding and extrusion
equipment. It has been found that high shearing during the preplasticisation stage in an
injection molding equipment caused considerable fiber attrition [78]. Gate dimensions and
dies were also shown to control the breakage [78; 79], while the screw compression ratio
had little effect on it [80]. Synthetic fiber concentration has also been shown to play an
important role as higher percentage of fibers rendered shorter post-processing lengths [81];
this result was corroborated during analysis of breakage in a twin-screw extruder [82].

Nevertheless, studies on twin-screw extruders have revealed that the most severe breakage
occurred in the mixing sections and different mixing section designs were analyzed by Shon
et. al. [83] and Ramani et. al. [82].

Studies on natural fiber-polymer composites have also reported fiber breakage during pro-
cessing in internal mixers, extrusion or injection molding equipment. Natural fibers used
have included sisal [8488], jute fibers [89], hemp strands [90], pineapple leaf fiber [91],
flax fiber [92; 93]. Fiber attrition was not the main research objective of most of these
studies. However, relationship between processing conditions and fiber breakage is evident
as fiber dimensions were reported before and after processing. For example, breakage of
short pineapple leaf fiber increased with rotor speeds in an internal mixer [91]. A study
of composites of sisal fiber in a polypropylene matrix processed in an internal mixer found
that fiber breakage increased at lower temperature (that is when the viscosity and shear
stresses in the mixture are higher), at higher rotor speeds and at longer mixing times [87].
20

Fiber breakage has also been reported during twin-screw extrusion. A study on flax fibers
mixed in a polypropylene matrix showed that fiber length was greatly affected by the pro-
cessing speed, fiber content and extruder layout [94]. Also, when processed with polycapro-
lactone, hemp fiber lengths were shown to decrease at high screw speeds, low feeding rates,
lower moisture content and higher temperatures [95; 96]. Decreased resident processing
times have been attributed to lower breakage of miscanthus and bamboo fibers [97].

A distinct characteristic of natural fibers is that, after processing, they commonly show a
reduction of their width when compared to their initial value. This reduction has been re-
lated to the separation of fiber bundles. Fibers cross-section is commonly assumed circular
and the fiber width is considered the fiber diameter, which is used to calculate the fibers
aspect ratio [4345]. Reduction of biofiber widths after processing have been conducive
to higher average fiber aspect ratios in the final composites [88; 98]. However, reduction
of fiber width has not always occurred and it has been independent of the occurrence of
biofiber length breakage [97; 99]. See Tables 1.1 and 1.2 for works that have reported fiber
breakage in internal mixers and twin-screw extruders.

Previous studies have provided good insight on the effect of processing conditions on fiber
breakage in different fiber-polymer systems. However, synthetic fibers and biofibers are not
alike. Synthetic fibers like glass are rigid and highly homogeneous. On the other hand, nat-
ural fibers from lignocellulosic plants are flexible and microscopically heterogeneous because
they themselves are composite materials of cellulose, hemicellulose and lignin. The arrange-
ment, orientation and percentage of the different constituents within a fiber are the result
of the natural process of plant growth, location of the fiber within the plant and fiber matu-
rity [100102]. Also, fiber heterogeneity might be further expanded during fiber extraction
[103]. As a result, biofibers have a high degree of property variability amongst fibers from
different species, and even among fibers from different sections of a plant [49; 104]. Studies
of fiber breakage of different biofibers is important as results for one type of fiber might not
be extrapolated to other natural fibers. For example, dependence of the material flow on
fiber species has been found by Li et al. [105].
Table 1.1: Natural fiber breakage during processing in internal mixers.

System wt.% fiber Processing Conditions Summary Ref.


Sisal-PP 30 165, 170 and 175 C at 50 RPM Shorter lengths at 165 C than at 175 C. [87]

and 30, 40, 50 and 60 RPM at 170 C Intensive breakage at 60 RPM

Pineapple 10,20,30 60 and 80 RPM at 130 C Higher breakage at 80 RPM [91]


leaf-LDPE
shorter lengths at higher temperatures [98]
Sisal-Mater Bi 20 80 and 120 C at 20 RPM
and higher speeds. Fiber width reduction
50 and 150 RPM at 120 C
is observed.
21
Table 1.2: Natural fiber breakage during processing in twin-screw extruders. The temperatures indicate the temperatures in
the mixing section.

System Fiber feed Processing Conditions Summary Ref.


Flax-PP-MAPP 10,20,30 100, 200 and 300 RPM Increased speed and fiber content gave [94]
Theoretical 4 extruder layouts lower fiber length but depended on ex-
fiber content 210/220/230 C truder configuration. Fiber width was re-
duced ten percent in average.

Hemp-Polycaprolactone 1,2,3 kg/h 1,2,3 kg/h at 200 RPM Collected fibers along the screw profile. [95]
100, 150, 200 and 300 RPM at 2 kg/h Length and width breakage increased at
100 C high screw speeds and low feed rates.

Hemp-Polycaprolactone 0.85,1.5 kg/h TSE Profile 1: 9.8,14,22.5 % moisture, 100 and Fiber damage differs with TSE pro- [96]
file. Profile 1 gave more width reduc-
140 C at 250 RPM, 0.85 kg/h, 140 C
TSE Profile 2 (less severe kneading): 100, 200, tion. Fiber breakage increased with screw
300 and 400 RPM, 0.85,1.5 kg/h at 100 C, 9.8 speed. Aspect ratio varies with feed rate
% moisture and screw speed.

Miscanthus- various total 100, 150, 225 and 300 RPM at 20 kg/h (TFR) Miscanthus: Increasing RPM and feeding [97]
bamboo- feed rates 100, 150, 225 and 300 RPM at constant rate preserved fiber length. Higher aspect
Poly(lactic acid) (TFR): TFR/RPM ratio at higher RPM. Bamboo: higher
PLA+fibers 165 C fiber breakage given by longer fibers

Cellulose-PP 20,30 wt.% 100 and 300 RPM at 8 kg/h. Higher screw speeds and fiber content in- [99]
185/190/195/200 C creased fiber breakage and thermal degra-
Some materials processed twice dation. Fiber diameter was not affected.
22
23

1.3.2 Effect of Initial Fiber Dimensions

The importance of unprocessed synthetic fiber length and diameter on the extent of breakage
during processing was shown by Maschmeyer and Hill [106] and Yilmazer and Cansever
[107]. It was found that shorter initial aspect ratios resulted in less synthetic fiber breakage
during processing and that fiber breakage stopped beyond a certain small fiber length. The
latter happened after materials were reprocessed several times. Opposite to this result,
Le Duc et al. [108], reported that final biofiber length did not show dependence on the
initial length of natural fibers. To the knowledge of this author, no work has correlated the
influence of initial biofiber width on fiber breakage.

Synthetic fibers have narrow distributions of length and diameter such that they could be
considered uniform. Natural fibers, on the other hand, do not show unique initial lengths
and diameters but usually have broad distributions. And due to breakage, these initial
distributions are continuously modified during processing with the consequent constant
variation of the material flow properties [109].

Also, contact surface between fibers vary with both fibers length and width. Fiberfiber
contact is one of the interactions that is considered to be influential on fiber breakage [110].
Contact could be maintained either by static friction [111] or attractive forces between fibers
[112].

A distribution would represent the entire spectrum of fiber lengths or widths. However, it is
usually desired to quantify and characterize the effect of a distribution by a single value or
point estimate [113]. Fiber length has been represented by different averages. The common
averages used are the number average (Ln), weight average (Lw) and Z-average (Lz), which
are defined as [114]:

X X X
Ni Li Ni L2i Ni L3i
Ln = X ; Lw = X ; Lz = X ; (1.4)
Ni Ni Li Ni L2i
24

While the number average gives the same weight to all lengths, the weight and z-average
give more weight to longer lengths and decrease the effect of fines on the length values. The
z-average is influenced the most by longer lengths.

The adequacy of using any of these averages on mechanical properties have been analyzed in
several papers [115118]. Also, one study has evaluated their suitability on simulation mod-
els to predict the dependence of viscosity on shear rate of glass fiber-polymer suspensions
[109]. The study found that the weight average fiber length, Lw, predicted viscosities closer
to experimental values. In addition, Lw is the average most commonly used to characterize
natural fiber length [119].

Fiber widths are commonly represented by the number average (Wn), and this study follows
this usage [119]. Also, as mentioned in section 1.2.3 natural fibers morphological cross-
section is commonly assumed circular, especially in mechanical test studies [4345]. This
assumption is followed in this work, and the measured fiber widths are considered as the
fibers diameters. The ratio of a fibers length (L) to its diameter (D) is the fibers aspect
ratio (L/D) and the estimator of the distribution used in this study is the number average
(L/Dn).

The standard error (


) for the average of a distribution (fiber width or fibers aspect ratio)
is given according to [113]:

S

X = (1.5)
nsample

where X is the mean of the distribution, S is the sample standard deviation and nsample is the
sample size. And the standard error for the weight-average length (sLw ) is given by [120]:

r
s2 ( Li )2 N0 ni Li (Li Lw)2
P P
2
sLw = ; e= P ; s = (1.6)
(Li )2 (N 0 1) Li
P
e

where i = 1, 2, M categories (bins in the distribution); n is the fiber count in the ith
category, L is the histogram class center length in the ith category, N 0 is the number of
non-zero weights and s is the weighted standard deviation of the sample.
25

1.3.3 Effect of the Type of Flow

In general, fiber breakage mechanisms have been summarized to be caused by one of the
following interactions: a) fiberfiber, b) fiberpolymer and c) fibersurface (fiber contact
with processing equipment surfaces) [12; 110; 121]. All these interactions are forced on the
fibers at different times during the mixing process. If the stresses applied on a fiber, due
to any of these interactions, exceed a fibers strength or the cohesive forces on the fiber
bundles, length breakage or width breakage, respectively, would occur. As mentioned in
section 1.2.4, the stresses developed in mixers or twin-screw extruders generate from shear
and elongational flows. It has been a common assumption that biofiber breakage occurs
only due to the relatively high shear stresses experienced while passing through the shear
flow of narrow gaps in processing equipment [98]. The latter assumption might derive from
earlier studies on fiber suspensions, which focused mostly on synthetic fibers, e.g. glass
fibers.

A capillary rheometer, as seen in section 1.2.5, separates the shear flow from the elongational
flow. Biofiber breakage during capillary rheometer studies has been reported by Kumar
et. al. [122]. Their study found that fiber breakage increased at higher extrusion rates.
However, no distinction was made on the level of breakage that happened in each section
of the capillary rheometer.

However, rheological flow properties of lignocelulosics-thermoplastic suspensions have been


carried out in capillary rheometer. Studies have shown that viscosity (sisal-polyethylene
suspensions) increased with the fiber content and fiber length [122; 123] and decreased
with temperature [123]. The effect of fiber length was marginal at low shear rates but was
very important at higher shear rates [123]. Regarding the temperature, the viscosity of the
suspensions was found to be more temperature sensitive than that of the unfilled polymer.
Other studies have found that the suspensions become more shear thinning (power law index
decreases) when the concentration of lignocellulosics increases [105; 122125]. However,
Ayora et. al. [126] found that as fiber concentration increases, the power law index (n)
increases as well. This tendency reached a maximum at an approximate fiber concentration
26

of 20 v.%. At higher fiber concentrations, the power law decreased with fiber content.
The maximum obtained was proportional to fiber flexibility. The behavior of n with fiber
concentration was explained in terms of the possible bending of the natural fibers during
flow, which would disrupt it, causing higher stresses than expected. Increase in the power-
law index has also been reported when an ester based internal lubricant was added to a
suspension [125].

1.3.4 Simulation of Fiber Breakage

Understanding the factors that influence biofiber breakage is important as it would help
composite producers and equipment manufacturers diminish damage to fibers during com-
pounding. Designing processing equipment involves, to a great extent, theoretical work
and simulation as they decrease costs, waste of materials and save time. In addition, when
simulation is paired with experimental data, it could also give assistance to determine the
relevant factors that influence a process.

Simulation works on fiber suspensions have been carried out. Nevertheless, the majority
of them have been aimed at modeling the time evolution of fiber orientation during flow.
Simulations have been developed for rigid fibers [127] and for flexible fibers as well [128
130]. Modeling fiber flexibility has been carried out considering the fibers as a chain of
beads joined with connectors [129] or as chains of rods connected by hinges [130]. And fiber
flexibility is expressed in terms of a phenomenological parameter related to the fibers Young
modulus. Two other factors that have been studied for their effects on fiber orientation are
fiber agglomeration [130] and wall effects [131].

There are simulation studies related to fiber breakage. In those simulations, breakage of
fibers can be caused by bending, buckling, torsion or stretching. Forgacs and Mason [132]
first introduced the idea of a critical buckling condition and described the types of motion
that a fiber would have during flow depending on its level of flexibility. Salinas and Pittman
[133] used an empirical approach to predict breakage of rigid fibers of known aspect ratio,
Youngs modulus and ultimate tensile strength under known fluid shear stress. They ap-
27

plied the equation to diluted suspensions of glass fibers in a Newtonian matrix. Yamamoto
and Matsuoka [134], developed a theoretical model that included fiber breakage by either
stretching, torsion or bending and obtained breakage threshold curves for fibers of different
bending strength. The curves gave a fiber fracture condition using the ratio of the polymeric
fluid shear stress to the fibers Youngs modulus. However, the number of published simu-
lation works on biofiber breakage during processing is limited. This author only found the
interesting work by Iannace et al. [98], where the minimum fiber length during simulation
was a fixed value. All these analyses have dealt only with the fiberpolymer interaction,
which is considered one of the interactions that causes fiber breakage [110; 121] as well as
an important interaction that influences the rheology of the composite during processing
[135].

1.4 Summary and Scope

Biofiber breakage has mainly been reported as a secondary result in other studies and a lim-
ited number of works have focused on improving the understanding on this area. Although
previous research works have provided good insight, there are some discernible areas that
have not been addressed adequately. The area of biofiber breakage requires further study
on the effect of processing conditions for more biofiber-thermoplastic systems. Also, the
different factors that would influence fiber breakage have been studied separately; a simul-
taneous analysis of variables would be important to identify the interaction between factors
on their influence on natural fiber breakage. Additionally, the influence of initial biofiber
length and width are important as well but, with the exception of the work of Le Duc et al.
[108], they have not been addressed sufficiently in literature. Furthermore, all studies have
presumed that biofiber breakage happens only due to high stresses in shear flows. Finally,
studies that simulate breakage of natural fibers during processing are scarce, more work on
this topic is required. In this study, the above gaps were addressed by studying biofiber
breakage during processing in two pieces of processing equipment: a) an internal mixer and
b) a twin-screw extruder. Variables included processing conditions as well as initial fiber
28

dimensions and their influence was analyzed simultaneously through multivariate statis-
tics. In addition, a more fundamental understanding of fiber breakage through the use of
a capillary rheometer was carried out. This analysis was used to determine the extent of
fiber breakage that happens in shear and in elongational flows. Lastly, a non-isothermal
computer simulation to predict biofiber length breakage during processing in an internal
mixer with a non-Newtonian fluid was carried out. The final fiber length depended on
processing conditions and fiber morphology. All simulation inputs were experimental data.
The simulation was based on the model by Yamamoto and Matsuoka [134] and considered
only the fiberpolymer interaction.

Two types of thermomechanically extracted biofibers were used, one wood source (aspen)
and one non-wood source (wheat straw). Both types of fibers have shown great potential
in their use as reinforcement for thermoplastics [3; 27; 136; 137]. Polypropylene was used
as the matrix for the composites. Polypropylene is the second most important commodity
plastic with sales projected to reach US$145 billion in four years as its demand keeps growing
[138]. Polypropylene has excellent mechanical properties of strength, stiffness and impact
resistance and it is used in many industries, including the automotive industry, as a pristine
resin or in composites [137; 139]. Of all the thermoplastic matrices available, polypropylene
(PP) shows the most potential benefits when combined with biofibers for making composites
of industrial value [140].

1.5 Thesis Structure

This thesis is presented in a paper based format where each research chapter has been
written for publication. This format resulted in some repetition in the background and
other details within the chapters. Each chapter is self-contained, independent and addresses
a specific objective. The chapters combined form a coherent work that contributes to the
body of knowledge on biofiber breakage.
29

Each research chapter is introduced in the order of publication or submission to peer-


reviewed scientific journals. Chapter 2 presents the hypothesis and objectives of the re-
search. The study of the influence of fiber morphology and processing conditions on fiber
breakage during processing in an internal mixer is presented in Chapter 3. This chapter
has been published: Quijano-Solis C., Yan N., Zhang S.Y. (2009). Effect of mixing con-
ditions and initial fiber morphology on fiber dimensions after processing. Composites Part
A: applied science and manufacturing, 40(4), 351-358.

Chapter 4 corresponds to the study of fiber breakage in a capillary rheometer. The results
of this chapter have also been published: Quijano-Solis C. and Yan N. (2014). Characteri-
zation of biofiber breakage in composite processing using a capillary rheometer. Journal of
reinforced plastics and composites, 33(16),1463-1473.

Chapter 5 and Chapter 6 show the experimental study on fiber breakage in a twin-screw
extruder and the Monte Carlo simulation, respectively. Both works have been submit-
ted for publication. Finally, chapter 7 presents the overall conclusions of the thesis and
recommendations for future work.

The author of this thesis has been the leading contributor to the formulation, execution of
the research and the writing of all papers submitted for publication.
Chapter 2
Hypothesis and objectives

2.1 Research Hypothesis

Natural fiber breakage during mixing with thermoplastic polymers is influenced by


the initial fiber dimensions.

Biofiber breakage occurs during the elongational flow developed in the entrance of
small gaps in processing equipment.

Natural fiber breakage depends on the relationship between processing conditions,


initial fiber dimensions and the compounding equipment.

Simulation of fiber breakage on a dilute system can help identify important factor
influence on this process.

30
31

2.2 Objectives

This study encompasses two main sections, an experimental study and a simulation. The
experimental sections have two general objectives: a) To determine how processing condi-
tions and initial fiber dimensions influence the breakage of natural fibers during processing;
b) to determine the influence of elongational flow on fiber breakage. The simulation has
the objective to determine the influence of processing conditions on biofiber breakage for a
dilute fiber concentration system.

Specific objectives:

to determine how fiber initial dimensions and processing conditions influence the
breakage of the fibers during processing in an internal batch mixer;

to determine the extent of breakage that happens in shear and elongational flow during
extrusion through a capillary rheometer;

to determine how fiber initial dimensions and processing conditions influence the
breakage of the fibers during processing in a twin screw extruder;

to develop a model to simulate fiber breakage during processing taking into account
the fiberpolymer interaction.
Chapter 3
Effect of mixing conditions and initial fiber
morphology on fiber dimensions after
processing

3.1 Abstract

In this study, the influence of initial fiber morphology and processing conditions on the
breakage of fibers during processing in an internal mixer was investigated. Composites of
aspen and wheat straw in a polypropylene matrix were processed at different temperatures,
intensities of mixing and fiber concentrations. Amongst the fiber parameters studied were
the fiber length, width and coarseness. The simultaneous analysis of variable influence on
fiber breakage was carried out using multivariate statistics. Results showed that the final
average fiber length depends mostly on processing conditions while the final average fiber
width depends on both processing conditions and initial fiber morphology.


The contents of this chapter have been published in: Quijano-Solis C., Yan N., Zhang S.Y. (2009).
Compos, Part A Appl Sci Manuf., 40(4), 351-358.

32
33

3.2 Introduction

The use of lignocellulosic fibers as reinforcements for thermoplastics has gained considerable
importance due to the many advantages that natural fibers offer over traditional synthetic
materials (e.g., glass, carbon or aramid). Lignocellulosics cause less abrasive wear to pro-
cessing equipment, they offer excellent specific properties of strength and stiffness and are
more environmentally friendly [2].

Natural fiber polymer composites from lignocellulosics can be used for many sectors that
include building and construction, furniture, consumer goods and the automotive indus-
try. Commercial success, however, will depend on the final mechanical properties of the
composite. If reinforcement of the thermoplastic is required, fiber aspect ratio needs to be
controlled.

Aspect ratio is defined as the ratio between the fiber length (L) and the fiber diameter
(D). Larger aspect ratios will more effectively transfer load from the matrix to the fiber,
giving composites with higher mechanical properties, such as strength and toughness [11].
In addition, a minimum (critical) aspect ratio is required in order to take full advantage of
the reinforcement capabilities of the fiber. The critical aspect ratio decreases with higher
fibermatrix bond strength and with lower fiber ultimate strength [15].

Attention to aspect ratio is especially important in short fiber thermoplastic composites.


During processing, fibers, thermoplastics and additives are usually mixed using common
thermoplastic processing equipment; e.g., mixers and twin-screw extruders. These process-
ing equipment impart such severe stresses to the materials that fibers may undergo breakage.
In general, fiber breakage is considered to be caused by one of the following interactions: (a)
fiberfiber, (b) fiberpolymer and (c) fibersurface (fiber contact with processing equipment
surfaces) [110; 121].

The breakage of the fibers length is well known to be a major concern for synthetic fibers
[83]. Studies on twin-screw extruders revealed that most breakage of glass fibers occurred
34

in the mixing sections [82; 83]. Many mixing element geometries exist and their different
combinations give a vast array of screw designs within a twin-screw extruder. A study
has found that screw designs that provided higher intensity of mixing and residence time
induced more fiber attrition as concentration of glass fibers within the composite increased
[82].

The length of lignocellulosic fibers has also shown to be reduced during the manufacturing
of composites. Higher rotor speeds resulted in an increased breakage of short pineapple leaf
fiber when processed with low density polyethylene in an internal mixer [91]. Also, studies of
sisal fiber polypropylene composites in an internal mixer found that fiber breakage increased
at lower temperatures, at higher rotor speeds and at longer mixing times [87]. Breakage
during processing has also been reported for other lignocellulosics such as jute fibers [89],
hemp strands [90], flax fiber [92] and wood fiber [22]. Although fiber attrition was not the
main research objective of these studies, a close relationship between processing conditions
and fiber breakage was reported.

Equivalent results in the breakage of fibers for synthetic and lignocellulosic fibers have
been observed. It has been found that glass fiber attrition stops beyond a certain small
fiber length [141]. This happened after materials were reprocessed several times in an
extrusion rheometer without a die. A limiting fiber length was also found for sisal fibers
when processed in an internal mixer and given enough processing time [98]. This limiting
fiber length did not significantly change when the fiber strength was lowered by a treatment
with an alkaline solution. Only the kinetics of length reduction was altered as weaker fibers
showed a faster reduction compared to untreated fibers [98].

Even though both synthetic and lignocellulosic fibers undergo reduction of their initial
length during processing, there is significant dissimilarities between them. Synthetic fibers
are rigid, highly homogeneous and with narrow distributions of length and diameter such
that they could be considered uniform. These characteristics are opposite to lignocellu-
losic fibers, which are flexible and microscopically inhomogeneous as they themselves are
composites materials of cellulose, hemicellulose and lignin. The arrangement, orientation
35

and percentage of different constituents within a fiber are the result of the natural process
of plant growth, which is influenced by the environment, conditions of growth, etc. Such
variations in composition and structure render a high degree of property variability even
among the lignocellulosic fibers [49].

Also, unlike synthetic fibers, where breakage consists of only reduction in fiber length, lig-
nocellulosic fibers also show reduction of the initial fiber diameter. This reduction has
been related to the separation of fiber bundles leading to a lower average diameter and
higher average aspect ratio of fibers in the final composite [88; 98]. In addition, lignocel-
lulosic fibers are temperature sensitive. Lignin has a glass transition temperature around
100 C [142] and major degradation of all components start around 200 C [143]. These
temperatures are within the normal range of processing of commodity thermoplastics such
as polypropylene and polyethylene. It is therefore expected that even though both synthetic
and lignocellulosic fibers undergo damage during processing, fiber property-breakage and
processing-breakage relationships obtained on synthetic fibers need to be analyzed for their
extent of applicability to the lignocellulosic fibers.

The aforementioned studies on lignocellulosic fibers have provided good insight into their
breakage during processing. However, further studies are required. Comparison of different
species of lignocellulosics as well as the simultaneous study of the influence of different
processing conditions have not been carried out. Also, it appears that no study has taken
into account the influence of initial fiber morphological parameters, such as coarseness and
fiber dimensions (length and diameter), on fiber breakage. In theory, the level of stress built
up on the fibers that results in breakage is related to the initial morphology of the fibers
themselves. Coarseness is defined as the weight per unit length of fiber, and is a measure
of fiber flexibility.

This study, therefore, is aimed at achieving a better understanding of the effect of initial
fiber morphology and processing conditions on the breakage of two types of lignocellulosic
fibers. Fibers of aspen and wheat straw were processed in an internal batch mixer with a
polypropylene matrix. The influence of fiber length and diameter is assessed by the use of
36

both an average of the distribution and its polydispersity index. The processing conditions
included different rotor speeds, temperatures and fiber concentrations.

3.3 Experimental

3.3.1 Materials

Fibers extracted thermomechanically from one wood and one non-wood source were used
in the composites. The wood fibers were hardwood aspen provided by FPInnovations-
Forintek Division Eastern Region. Fiber extraction was carried out in the MDF pilot
plant at FPInnovations-Forintek Division under two refining conditions aimed at obtaining
batches of two different lengths, which will be referred to as long and short fibers. Both the
aspen long fibers (AL) and aspen short fibers (AS) were refined using the same preheating
time and pressure of 2.5 min and 6 bar, respectively. The refining energy and speed were
as follows: 150 kW h/t and 2000 RPM for AL fibers; and 250 kW h/t and 2800 RPM for AS
fibers. The non-wood fibers were wheat straw (WS) refined at the following conditions:
3 min, 9 bar and 3000 RPM. These WS fibers were provided by Alberta Research Council.
All batches of fibers were provided dried.

The polypropylene (PP) homopolymer Profax 6523 by Basell was used as the matrix for
the composites. This polymer has a nominal melt flow rate of 4 g/10 min (230 C/2.16).
The peak melting point as determined using a DSC1000 by TA Instruments was 160 C.

3.3.2 Equipment

Composites were processed in a CWBrabender 3-piece mixer/ measuring head with roller
type blades. These blades are of irregular shape, and subject the samples to intensive mix-
ing under controlled conditions of temperature, shear and time. The mixer has a capacity
of 260 cm3 when the roller blades are used. However, for optimum mixing performance, it
is usually filled to 7080 v.% capacity [42]. This type of mixer is standard laboratory equip-
37

ment that is often used to characterize materials processability and performance prior to
final manufacture. Correlation between this mixer and large-scale extruders is established
empirically by comparison of the optimum operating conditions in the internal mixer be-
tween new materials and materials of known performance in the industrial process analyzed
[56].

3.3.3 Composites processing

The processing conditions included the mixer temperature set point and the rotor speed or
intensity of mixing (RPM). The temperature set points used were 180 and 195 C and the
RPMs were 40 and 60. In addition, two levels of fiber concentration (10 and 30 wt.%) were
used for each fiber. The mixing procedure followed the one suggested by Rauwendaal [42].
After material loading, the mixing time for all composites was kept constant at 6 min. This
mixing time has been reported to be sufficient to obtain an homogeneous mixture and for the
process to reach steady state conditions [144]. The effect of mixing time on fiber breakage
have shown that the strongest reduction in fiber dimensions occurred during the initial steps
of mixing with no significant change in dimensions observed after reaching steady state [98].
Therefore, no significant breakage is expected to occur in our mixtures after steady state
conditions are attained. Five replicates were carried out for every combination of processing
conditions. From every replicate a representative portion was selected for fiber extraction.

3.3.4 Morphological measurements

Morphological properties of fibers, before and after processing, were measured using a HiRes
FQA by Optest Equipment Inc. The HiRes FQA combines hydraulic, optical, and image-
processing systems to measure fiber length and width, amongst other fiber data. Sheath
flow of a dilute suspension of fibers transports them into a cell that renders an almost
two-dimensional plane. Flow hydrodynamics orients the fibers in the cell where the optical
and image systems measure fiber dimensions [145]. It can also be used to determine fiber
coarseness. As described earlier, coarseness is defined as the weight per unit length of fiber.
38

The range limits of the HiRes FQA are: for fiber length between 0.04 and 10 mm and for
fiber width between 7 and 300 m.

Fibers were extracted from the processed composites in order to measure fiber dimensions
after processing. For every combination of processing conditions, all portions that were
selected for fiber extraction were combined and underwent the extraction process simulta-
neously. The extraction involved the dissolution of the polypropylene in xylene at boiling
temperature . The composite was immersed for 1 h in boiling xylene in a flask. The sus-
pension was then filtered at high temperature to prevent precipitation of the polymer. This
procedure was carried out three times for each sample to assure that all polypropylene was
dissolved. Finally, the extracted fibers were dried at room temperature for at least 20 h
to ensure the evaporation of the xylene. Figure 3.1 shows a schematic of the extraction
process. Three replicates from every batch of unprocessed as well as extracted fibers were
measured in the HiRes FQA.

Figure 3.1: Schematics of the fiber extraction process.



Cervantes M. Private communication; 2006.
39

3.4 Analysis

3.4.1 Processing variables and fiber properties

The processing variables studied include the temperature set point, the intensity of mixing
(RPM) and fiber concentration. The temperature set point is the temperature at which the
mixer is heated up prior to mixing the fibers and polymer. Thus, it is the initial temperature
of the mixing process. However, the temperature of the mixture after processing was higher
than the set point. Final temperatures ranged from 10 to 20 C above the starting set point.
It was decided, therefore, to record the final processing temperatures and also include them
in the analysis of the data.

Fiber morphological parameters measured by the HiRes FQA included length and width.
Due to the fact that natural fiber dimensions are not uniform, a distribution of lengths and
widths was obtained. The effect of a distribution is usually quantified and characterized
by a single value, the distribution average. The two most common averages used are the
number average and weight average [114]. The number average is highly influenced by
the small values of the distribution, while the weight average gives more weight to larger
values. Fiber widths are commonly represented by the number average (Wn) [119], and
this study follows this usage. On the other hand, fiber lengths are commonly represented
by the weight-average length (Lw). Lw gives more weight to longer fibers and decreases the
effect of fines on the length values [119]. The equations for Wn and Lw are

X X
ni Wi ni L2i
Wn = X ; Lw = X (3.1)
ni ni Li

where i = 1, 2, , N categories (bins in the distribution); n is fiber count in the ith


category and L (or W ) is the histogram class center length (or width) in the ith cat-
egory. Regarding fibers width. Natural fibers morphological cross-sections are variable
and irregular. The HiRes FQA measures the width of each fiber multiple times along its
40

length and based on these measurements provides only one value of width (W) for each
fiber [119]. The average width, Wn, is the mean taking into account all fibers width, W. In
biofiber composites studies, biofibers cross-section is commonly assumed circular and the
fiber width is considered as the fiber diameter [43; 46; 98]. In this work, this assumption is
followed by considering the fiber widths as the fiber diameters.

In addition to the length and width distributions, distributions of the aspect ratio were
calculated. The fiber aspect ratio or L/D is the ratio of the fiber length to its diameter. To
P P
represent the distribution of aspect ratios, its number average (L/Dn= ni (L/D)i / ni )
was used.

The effect of a distribution can also be represented by the polydispersity index, which is
the ratio of the weight average to the number average of the distribution, e.g., the fiber
length polydispersity is Lpol = Lw /Ln . Width and aspect ratio polydispersity values are
represented as Wpol and L/Dpol, respectively. Polydispersity is a measure of the ratio of
long to short values in a distribution and with the exception of uniform distributions, poly-
dispersity values are always greater than one. The last fiber morphology factor considered
was the fiber coarseness, which was also measured using the HiRes FQA. Polydispersity in-
dices and coarseness values were only calculated for the unprocessed fibers. To distinguish
between fiber properties before and after mixing, all properties for the unprocessed fibers
were identified with the prefix UP-, e.g., UP-Lw.

Finally, new factors that account for the change in fiber dimensions were calculated. These
factors are the difference between the properties before and after processing. To identify
these factors the prefix Delta- is used before the property name. These factors are the
following:

Delta-Lw = UP-Lw Lw
Delta-Wn = UP-Wn Wn (3.2)
Delta-L/Dn = UP-L/Dn L/Dn
41

All processing conditions and fiber morphological values that correspond to the properties
before processing are the variables or predictors considered influencing fiber breakage, and
they are called the X-space. On the other hand, fiber morphological values after processing
and the Delta-properties were considered the response properties or Y -space. Table 3.1
summarizes all factors in the X- and Y -spaces.

Table 3.1: Factors in the X- and Y -spaces


Variables or predictors (X-space) Responses (Y-space)
Temperature (Temp.) Lw
Final temperature (Final temp.) Wn
RPM L/Dn
Fiber concentration (Conc.) Delta-Lw
UP-Lw Delta-Wn
UP-Lpol Delta-L/Dn
UP-Wn
UP-Wpol
UP-L/Dn
UP-L/Dpol
UP-Coarseness (UP-Coar)

Processing conditions, such as the temperature set point and RPM, as well as the fiber
concentration are factors that can be set to the desired values according to the experimental
design. However, there are also factors whose values could not be controlled, such as the
final processing temperatures and all initial fiber morphological properties. The inclusion
of variables that do not vary according to a design of experiments will promote correlation
among the predictors. In addition, there were two runs for which the final temperature
was not recorded. It was decided, therefore, to carry out the analysis of variance using
Projection to Latent Structures (PLS). PLS successfully handles missing data and extracts
the meaningful variance correlation in the X- and Y -spaces while relating the X- and Y -
spaces to each other [146; 147].
42

3.4.2 Multivariate data analysis

Projection to Latent Structures (PLS) is a multivariate regression tool that helps to reveal
correlation amongst input-variables or predictors (X-space) and also their impact on several
responses (Y -space). This is done by separating regularities from noise in the data. PLS
handles data with strong collinearity, noise and missing values in both the X- and Y -spaces.
This tool reduces the dimension of the system to a smaller number of latent variables
(referred to as principal components or scores) that can simultaneously explain the signif-
icant variance in X, and also predict Y . The higher the correlation in the data the fewer
principal components that are computed. The scores are independent of each other and are
a linear combination of the original predictors. The weight of each predictor that is used to
calculate the scores is directly related to their level of influence on the measured Y -space
properties. An important aspect of PLS is the ability to show the interrelationship among
all predictors, the relationship among all responses, and simultaneously the predictors in-
fluence on the measured responses, all of them in a single plot, the w c plot. All these
calculations are usually carried out by first centering the data to have a mean of zero and
then scaling to unit variance. This process of mean centering and scaling is done in order
to give each variable the same weight and importance prior to the analysis. This is done to
counteract the effect of scaling in different measurements units, and to allow each variable
to contribute equally to the model. As a regression tool, PLS provides a measure of the
goodness of fit, R2 . R2 is an indication of how much variance in the data is explained by the
model. R2 , for any regression tool, including PLS, can always be increased by adding more
terms (complexity) to the hypothesized model. A far better metric to gauge model perfor-
mance is by using the so-called Q2 metric. Q2 is an indicator that measures how well the
regression model can predict new data. One technique to estimate Q2 is by cross-validation.
This method consists of dividing the data into a number of groups. Models are built with a
group of data left out one group at a time. With each model, the corresponding omitted
data are predicted and the total prediction error sum of squares calculated. Q2 , like R2 ,
varies between 0 and 1, where values closer to 1 indicate better prediction ability. The Q2
value will always be smaller than R2 . Finally, Q2 is used to select the number of principal
43

components (model complexity) to avoid overfitting.

PLS models can be converted to a standard linear regression form as given by the following
equation:

X
y = k + bn xn (3.3)
n

where k is a constant, bn is the coefficient corresponding to the predictor xn and y the


predicted y-property. Details of the PLS calculations can be found elsewhere [146]. Several
software packages are available to create PLS models. The SIMCA-P software by Umetrics
was used in this work.

3.5 Results and discussion

The statistical analysis to determine the influence of processing conditions and initial fiber
morphological properties on the breakage of lignocellulosic fibers during processing was
carried out using PLS. All predictors and responses included in the analysis are shown in
Table 3.1. The initial fiber morphological data is part of the predictors (X-space) and
therefore one value of each unprocessed (UP) property for each type of fiber (AL, AS, WS)
was used, e.g one UP-Lw, UP-Wn, etc. These values were calculated by combining into one
distribution the data from all replicates of measurement from the batch of each unprocessed
fiber. Tables 3.2 reports the average dimensions for unprocessed fibers. It can be noted
that the two different refining conditions for aspen (AS and AL) indeed rendered fibers
with two different morphological properties. The aspen short fibers are both shorter and
thinner than the aspen long fibers by 11 % and 26 %, respectively. The highest values of
Lw and Wn, however, do not correspond to the aspen long fibers, which are 26 % shorter
and 24 % thinner than the wheat straw fibers. This clear pattern in initial Lw and Wn
dimensions where AS < AL < WS is not reflected on the aspect ratio. The smallest average
L/Dn is shown by AL fibers, whose aspect ratio is 5 % and 10 % smaller than the ones
for AS and WS, respectively. The polydyspersity index of the distributions also shows
44

variation between all fibers. AL fibers had the broadest distribution of length and width.
Aspen long UP-Lpol exceeded those of AS and WS by more than 50%. The polydispersity
indices of aspect ratio, on the other hand, followed a pattern similar to that of Lw and Wn:
AS < AL < WS.

Table 3.2: Fiber dimensions before processing. Length is reported in mm and width in
m. Aspect ratio and polydispersity index values are dimensionless.
UP-Lw UP-Lpol UP-Wn UP-Wpol UP-L/Dn UP-L/Dpol
AS 1.5(0.011) 2.4 35.6(0.37) 1.8 36.8(0.2) 1.3
AL 1.7(0.014) 4.2 48.1(0.75) 2.1 35.2(0.38) 1.5
WS 2.3(0.028) 2.8 63.8(1.77) 1.9 38.9(0.60) 1.7

Values in parenthesis are the standard error of the averages (see Equations 1.5 and 1.6)

Regarding fiber dimensions after processing, a decrease in fiber average length (Lw), diame-
ter (Wn) and aspect ratio (L/Dn) with respect to initial values was obtained for all natural
fibers. Processed data measurements form part of the responses (Y -space) in PLS, this
statistical technique analyzes the data mean centered and scaled to unit variance (averages
of the processed data as used in PLS are shown in Appendix A). In general, the percentage
reduction in average fiber length (38.981.5 %) was greater than the percentage reduction
in average fiber width (0.450.8 %). This implied that fibers after processing had smaller
average aspect ratio than initial fibers. Fiber aspect ratio had a reduction between 25.6 %
and 56.7 %. The xylene treatment might remove certain amounts of lignin, wax and oils
from fibers surface that would have an effect on fiber dimensions. However, we consider
these effects were negligible when compared to the reduction in fiber length and width that
occurs due to the processing of the fibers.

The PLS analysis gives four principal components that together explain more than 80 % of
the average variance (R2 ) in both X and Y -spaces. The first two components, however,
explain most of the Y -space variance (70 %). The ability of the PLS model to predict the
data (Q2 ) was good as well, predicting 79 % of the average Y -space variance.
45

Figure 3.2 shows the values of R2 and Q2 for every fiber property. It can be corroborated
that the two first principal components explain most of the variance in fiber properties after
processing. Final average width (Wn), Delta-Lw and Delta-L/ Dn are mostly explained by
the first principal component (PC1) while final average length (Lw) is almost exclusively
explained by the second principal component (PC2) with practically no effect from PC1.
The explained variance of the remaining responses (L/Dn and Delta-Wn) depended on a
combination of all components.

Figure 3.2: Values of R2 and Q2 for each fiber property

3.5.1 Loadings plot

Principal components are latent variables that depend on the original predictors. The fact
that the number of initial variables is 11 but that there are only four principal components,
confirms the high degree of correlation among the predictors. To assess this correlation and
the predictors influence on the scores and ultimately on the final fiber properties one looks
at the loadings plots.
46

The loadings plot between the first (w*c[1]) and second (w*c[2]) principal components,
which are the most important, is shown in Figure 3.3. The loadings plot shows the weight
or importance that every predictor has on each principal component. This weight is directly
related to the distance from the origin, either horizontally for w*c[1] or vertically for w*c[2].
The Y -space fiber properties are also shown in the loadings plot and the weight of a principal
component on each property depends on the distance of the specific fiber property from the
origin. The influence of each predictor on a fiber property is therefore given by the combined
weighted effect of predictors and principal components.

Figure 3.3: Loading plot of principal components 1 and 2


47

One key feature of the loadings plot is that it shows factor correlations in a simple graph-
ical form. Variables that are located near each other are highly and positively correlated.
Negative correlation is observed when variables are located on opposite sides of the origin.
For example, the processing conditions RPM and concentration are both located very close
to the final temperature, which emphasizes their positive correlation with this predictor.
Final temperature represents the temperature reached after the process is finished. A higher
RPM means higher shearing of the polymer, which in turn promotes more viscous heating
dissipation [56; 148]. Higher fiber concentration, on the other hand, increases the frequency
of fiber contacts resulting in an increase in heat dissipation by friction.

The first direction, w*c[1], which explains the greatest degree of variability, is described
by raw material variation. The second direction, w*c[2], which describes the next greatest
portion of the data, is described by changes in the processing conditions. Processing con-
ditions (final temperature, RPM and concentration) are grouped together and are separate
from raw material data (UP- Lw, UP-Wn, UP-Coar, UP-L/Dpol).

3.5.2 Scores plot

The scores plot, Figure 3.4, is a PLS output that shows the exact relationship amongst
all experimental samples. The scores plot shows the samples, or experimental runs, as a
function of the underlying principal components. Each sample can be assigned a set of
score values; a value per component, which are called t[1] for principal component one and
t[2] for principal component two. Score values of each sample are determined by a linear
combination of the predictors.

By plotting t[1] versus t[2], experiments with similar predictors and properties would clus-
ter together. In addition to showing clustering, a scores plot helps discover data inho-
mogeneities, grouping and outliers. Identifying and removing outliers are important steps
towards developing a reliable predictive PLS model. It is important to mention that the
analysis to identify outliers was carried out previous to reaching any conclusion about the
influence of predictors on fiber properties.
48

Figure 3.4: Scores plot of principal components 1 and 2. N AS,  AL and  WS

In the scores plot we can see a clear separation between all natural fibers in the horizontal
axis and in the vertical axis all the runs for each sample are distributed. The position of
any experiment in the scores plot is readily explained by the loadings plot as the two can be
visualized as superimposed on one another. Thus, the higher the positive t[1] value for an
experimental point the greater the initial average width and length the unprocessed fibers
had. Accordingly, we can observe that under the same combination of processing conditions
(same vertical distance) wheat straw fibers had the highest final average width, followed by
the aspen long fibers. Aspen short final average fiber widths were the smallest. Lw values,
on the other side, vary along the second component. As Lw is opposite to the predictors
final temperature, RPM and concentration, a negative correlation is deduced. A high value
of t[2] indicates that the sample ended with large Lw and that it was processed at low values
of RPM and fiber concentration.

Due to the fact that the highest variability is given by raw material data, a material clus-
tering is observed in the scores plot (Figure 3.4).
49

It is important to point out that the initial temperature set point appears, in general, to
have little effect on any of the experimental outcomes, as it has low weight in the model.

3.5.3 Predictors influence on fiber properties

3.5.3.1 Lw and Delta-Lw

The responses Lw and Delta-Lw have different meanings. Lw represents the average length
of the distribution of final fiber length, while Delta-Lw is a measure of the level of breakage
that occurs during processing.

It was noted previously that PC1 explains a mere 1 % of the final average length, Lw, and
predicts none; but Lw is mostly explained (92 %) by the second principal component. This
result is visually corroborated in Figure 3.3. The loadings plot w*c[1] versus w*c[2] shows
that Lw is located on PC2 with practically zero influence from PC1. Therefore, final values
of Lw are primarily a function of processing conditions: RPM, concentration and final
temperature. All these processing conditions are negatively correlated with final average
fiber length, which indicates that increasing any of them will induce shorter Lw.

Negative correlation with RPM is in accordance with the results reported in the literature
[91]. The flow inside the internal mixer is very complex and is a combination of shear and
elongational fluid motion. RPM is a measure of the rates of shearing and elongation that
the material undergoes within the internal mixer [56]. The stresses developed in the mixture
are proportional to these shear and elongational rates, so that increase in RPM promotes
higher stresses. If the hydrodynamic stresses on the fibers are high enough they can exceed
their strength and cause breakage [98].

On the other hand, higher fiber fractions increase the viscosity of the system [149]. The
latter also increases the stresses in the system, as a direct proportionality exists between
stresses and viscosity. As mentioned previously, fibers are more prone to damage as stresses
increase.
50

Figure 3.5: Coefficient plot for the final average fiber length, Lw, property

The third major predictor correlated with Lw is the final temperature. The negative cor-
relation with Lw that is found appears to be opposite to results reported in the literature
[87]. This seeming contradiction is because viscosity decreases when temperature rises and
thus lower stresses and breakage are expected. In this study, final temperature values are in
the range 190234 C. Thermogravimetrical analysis of the aspen and wheat straw showed
that major degradation of both types of fibers start around 195 C. At these temperatures
major degradation is experience by the lignin and hemicelluloses [143]. It is believed then,
that the decrease in fiber length with the increase final temperature predictor found herein
is the result of fiber weakening due to thermal degradation.

All observations mentioned previously are the result of the first two principal components.
Although these components explain most of the variance, the final correlation relation-
ships are the result of the combination of all principal components. The coefficients plot,
when mean centered and scaled, shows the final correlation pattern between predictors and
fiber properties. Figure 3.5 presents the coefficients plot for the final average length. The
dominance of RPM, concentration and final temperature is apparent, while other X-space
variables show minor or no effect on their influence to determine final values of Lw. There-
51

Figure 3.6: Coefficient plot for the Delta-Lw property

fore, values of Lw after processing could be determined mainly by processing conditions


with practically no regard of their initial length, width and coarseness.

Unprocessed fiber morphology, on the other hand, has more influence over the level of
breakage that occurs or Delta-Lw (see Figure 3.6). For example, wheat straw fibers are
initially longer and wider than aspen short fibers. After processing, WS fibers underwent a
harsher reduction in length to reach final average lengths similar to those of AS fibers.

Coarseness is also positively correlated to Delta-Lw as coarser fibers had a larger decrease
in length. This could be because greater fiber coarseness increases flocculation [150] and
floc strength [151] via stiffening of the fibers.

Kerekes and Schell [150] reported that coarser fibers gave suspensions with lower uniformity
and suspension uniformity is directly related to the number and size of flocs. The effect
of fiber coarseness on flocculation can also be correlated through fiber flexibility. Lower
fiber flexibility, which is obtained by increasing coarseness, has been reported to increase
flocculation [152]. This result is consistent with the works that reported that more rigid
fibers have higher cohesive forces via interfiber friction [153; 154]. This favors floc formation
52

as too flexible fibers were found not to flocculate [154].

The ideal floc dispersion during mixing aims at fiber disentanglement. However, instead of
disentanglement, fiber breakage may occur due to higher fiberfiber intimate contact with
higher flocculation or due to the floc strength being higher than the strength of the fibers
that form it.

3.5.3.2 Wn and Delta-Wn

The analysis of the final average width, Wn, and the level of width reduction, Delta-Wn,
follows a similar reasoning as the one for fiber length. Figures 3.7 and 3.8 show the coefficient
plots for these properties. Processing conditions as well as initial fiber morphology are
correlated with Wn and Delta-Wn.

Figure 3.7: Coefficient plot for the final average fiber width, Wn, property

Figure 3.7 shows a clear separation between processing conditions and initial fiber morphol-
ogy in the way they influence Wn. Processing conditions are all inversely correlated to Wn.
On the other hand, a positive correlation is found between the initial fiber morphology and
the final average width.
53

Figure 3.8: Coefficient plot for the Delta-Wn property

The harsher processing conditions promoted by higher rotor speeds and fiber concentration
will effectively fibrillate the fibers into entities of lower width. In addition, the increase in
temperature may facilitate the path to bundle separation by softening the lignin as its glass
transition spans over a range of temperature between 100 and 170 C [142].

However, the final width dimensions are not solely determined by processing conditions.
In fact, as mentioned above (see also Figure 3.2), initial fiber morphology plays an almost
dominant role in the determination of Wn. Fibers that are thicker and with broader dis-
tribution of widths and aspect ratios will promote a higher degree of thickness reduction.
Thicker fibers have a greater probability of interacting with other fibers or with equipment
surfaces. On the other hand, a broad distribution of width may facilitate the packing of the
fibers, which would increase fiberfiber interaction [155]. Larger aspect ratios, on the other
hand, increase the viscosity of the system [149]. The fact that longer lengths also serve to
increase width reduction may be due to the positive correlation with fiber width. Finally,
the increased flocculation [150] and floc strength [151] promoted by higher fiber coarseness
adds to the probability of increase breakage due to higher fiberfiber interaction or due to
fiber strength being lower than floc strength.
54

An interesting result is that thicker fibers undergo greater thickness decrease, even though
this reduction is not enough to convey these initially thick fibers to the low levels of thick-
ness acquired by fibers that are initially thinner. Notice the position of Wn very close to
unprocessed morphological predictors in the loading plot, which indicates that fibers that
initially have the largest thickness will be the thickest ones after processing.

3.5.3.3 L/Dn and Delta-L/Dn

Results for L/Dn and Delta-L/Dn are shown in Figures 3.9 and 3.10. These graphs couple
the observations obtained for length and width. It is clear that both processing conditions
and fiber morphology influence their values. More severe processing conditions, as expected,
promote more breakage to give final lower values of aspect ratio. Regarding the initial fiber
morphology, the high negative influence of the polydispersity index for the length and width
predictors on L/Dn could be due to the increase in level of packing that fibers experience
when the distributions are broader.

Figure 3.9: Coefficient plot for the final average fiber aspect ratio, L/Dn, property
55

Figure 3.10: Coefficient plot for the Delta-L/Dn property

3.6 Conclusions

In this study, projection to latent structures (PLS) statistical analysis was used to examine
the influence of processing conditions and initial fiber properties on the breakage of thermo-
mechanical fibers of aspen and wheat straw during processing in an internal batch mixer.
A decrease in average fiber length (Lw), diameter (Wn) and aspect ratio (L/Dn) after pro-
cessing was obtained. It was found that the longest and thickest fibers underwent greater
reduction in length than shorter fibers to reach approximately the same final Lw. However,
these thicker fibers had the highest final values of Wn. As a result, values of aspect ratio
after processing were larger for fibers with initially smaller length and thickness.

Processing conditions, such as RPM, fiber concentration and the level of processing terminal
temperature are proportional to the reduction in both fiber length and width. Values of Lw
after processing could be determined mainly by processing conditions. Each combination of
operating parameters is conducive to a final Lw, with practically no regard of their initial
length, width and coarseness. The more demanding the processing conditions the lower the
final values of average length obtained.
56

The final average fiber width is also strongly dependent on the initial fiber morphology.
Thicker fibers, higher aspect ratios, broader L/Dn distributions, and higher coarseness are
correlated to a larger decrease in fiber width. An interesting result is that thicker fibers
undergo a larger thickness decrease. This reduction, however, is not enough to bring their
thickness to the low levels of thickness acquired after processing by fibers that initially were
thinner. Final aspect ratios, therefore, are a function of both processing conditions and
initial fiber properties.

These results point out the importance of using fibers with low thickness even if the length
is also short. Length will be shortened during processing regardless of the initial length,
but fibers that are initially thinner will result in fibers that have higher L/D ratios after
processing. The analysis in this work is empirical and thus valid only within the system
range and conditions used. Results herein can be used as guidelines but total extrapolation
to other type of fibers (synthetic or natural) or to other processes should be made with
caution.
Chapter 4
Characterization of biofiber breakage in
composite processing using a capillary
rheometer

4.1 Abstract

In this study, a capillary rheometer was used to determine the impact of shear and elon-
gational flow on the breakage of aspen fibers in a polypropylene matrix. Composites were
processed at different concentrations, temperatures, and rates of strain. The influence of
the processing parameters was also analyzed. The difference in breakage of the fibers was
assessed through their weightaverage fiber length. Results showed that both types of flow
exert influence on the total breakage of the fibers. In some cases, the percentage of break-
age in the elongational flow is higher. The level of fiber breakage was similar for all fiber
concentrations and shear rates. Temperature influence depended on the type of flow.


The contents of this chapter have been published in: Quijano-Solis C. and Yan N. (2014). J. Reinf
Plast Comp, 33(16),1463-1473.

57
58

4.2 Introduction

Interest in biofibers as reinforcements for thermoplastics is growing at a rapid pace world-


wide. Compared to synthetic fibers, biofibers have the advantage of lower density and good
specific mechanical properties [2]. They are also cheaper to produce than their synthetic
counterparts. Biofibers are more environmentally friendly as they are obtained from renew-
able resources, can be recycled, and offer CO2 neutrality [156]. In addition, biofibers cause
less abrasive wear to processing equipment than synthetic fibers.

Unfortunately, biofibers can get damaged by breaking in length or width during composite
processing [157159]. Fiber breakage is an undesired outcome when fibers are expected to
act as reinforcement in a composite because their aspect ratio is reduced. The shorter the
fiber aspect ratio, the less efficient is the transfer of the applied loads from the matrix to
the fiber [15]. Past work has shown that biofiber breakage depends on fiber characteristics
and processing conditions such as temperature, time, and processing (mixer or extruder)
speed [87]. All these factors influence the fibers mechanical strength and the dynamics of
the mixing.

Fiber breakage happens when the stresses exerted on the fiber are high enough to exceed
the fibers ultimate strength. It is a common assumption that breakage of biofibers occurs
due to only the relatively high shear stresses experienced while passing through the shear
flow of narrow gaps in processing equipment [98]. The latter assumption might derive from
earlier studies on fiber suspensions, which focused mostly on synthetic fibers, e.g. glass
fibers.

In addition to shear flow, elongational flow is also developed in commercial mixers [62].
Strong elongational flows could occur upstream of the narrow gaps where shear flow occurs
[66] or in the intermeshing region in multiple screw extruders [160]. Even though the stress
levels developed in elongational flow are larger than in shear flow[62], breakage of synthetic
fibers in elongational fields has been considered unlikely as synthetic fibers get oriented
in the direction of flow, are highly homogeneous, and have high Youngs modulus [133].
59

However, Lee and George [161] reported some breakage of glass fibers during visualization
of flow in an elongational field.

This chapter explores the hypothesis that biofiber breakage in elongational flows is also a
possibility as biofibers have lower stiffness than synthetic fibers, are inhomogeneous, and
show variability of mechanical properties within each fiber. This variation is attributed
to defects both along the fiber length and on their surface, which occur naturally or as a
consequence of the fiber extraction process[49].

The knowledge of where breakage occurs, shear versus elongational flow, and the extent of
biofiber breakage in each flow type could help both composite producers and equipment
manufacturers to minimize reduction in fiber aspect ratio while maintaining good mixing
performance. In this work, a capillary rheometer is used to determine the extent of fiber
breakage that happens in shear and in a convergent flow. The breakage that occurs in the
convergent section is used as an indication of the breakage that happens in elongational
flow. Fiber breakage through capillary rheometer has been reported for both glass- and
biofiberthermoplastic composites [85; 162]. However, no analysis of the type of flow field
on breakage has been carried out.

Composites of aspen fiber in a polypropylene matrix were processed at two levels of fiber
concentration (10 and 30 wt.%), two shear rates (500 and 1500 s1 ), and three temperatures
(190, 200 and 210 C). The degree of fiber breakage as a function of these processing
conditions was also analyzed and discussed.

4.3 Experimental

4.3.1 Materials

Wood fibers from aspen provided by FPInnovations-Forintek Division Eastern Region were
used for the composites. The batch of fibers was provided dried and was thermomechanically
extracted in a medium density fiberboard pilot plant at FPInnovations-Forintek Division.
60

The refining conditions were set as preheating time of 2.5 min, pressure of 6 bar, refining
energy of 150 kW h/t, and speed of 2000 RPM. These conditions correspond to the aspen
long (AL) fibers.

The homopolymer Polypropylene Profax 6523 (Lyondellbasell, the Netherlands) was used
as the matrix for the composites. This polymer has a nominal melt flow rate of 4 g/10 min
(230 C/2.16). The peak melting point as determined using a DSC1000 (TA Instruments,
USA) was 160 C.

4.3.2 Processing equipment

Two processing equipments were used. One was a three-piece mixer/measuring head with
roller-type blades from CWBrabender Instruments with a capacity of 260 cm3 . For optimum
mixing performance, it was filled to 7080 v.% of its capacity [42]. Processing parameters
that can be set with this mixer include temperature, time, and rotors RPM. The final
processing temperature, which increases during processing, can also be monitored.

The other equipment was a twin-bore Rosand bench-top capillary rheometer series RH2000
(Malvern Instruments, UK). This rheometer consists of two bores, each with a capillary die.
The material is melted in the bores and forced through the capillary dies by pistons (Figure
4.1). Viscosities, strain rates, and stresses of the material are calculated through the flow
rates and the pressure drop. The total pressure drop experienced by the material as it is
processed, which is recorded by pressure transducers, is caused by two factors. One factor
is the flow resistance through the capillary die. The second factor is the materials resistance
to pass from the wide barrel into the capillary die. The latter is often referred to as the
entrance pressure drop, which is due to the elongational viscosity of the elongational flow
field developed [72; 73]. Cogswell theory is widely used to calculate extensional viscosities
from capillary rheometers [7477].

In order to decouple the pressure drop that occurs in the capillary die from that of the
excess pressure drop, both bores have dies of equal diameter but of different lengths. One
61

Piston

Capillary die Capillary die


16 mm legth 0.25 mm legth
1mm diameter 1mm diameter

Figure 4.1: A schematic of the capillary rheometer with long and zero-length dies

bore used a 16-mm length die. The other used the zero-length die, which has a length of
0.25 mm. Both dies are 1 mm in diameter and have an entry angle of 180. The capillary
bore has a diameter of 15 mm. It is a common practice for this type of capillary rheometer
to specify only the temperature and the apparent (nominal) shear rates in the long capillary
die as processing parameters. Detailed theoretical background on a capillary rheometer can
be found elsewhere [71].

4.3.3 Composite processing

The Rosand capillary rheometer is a ram extruder. If a composite is to be processed in this


capillary rheometer, the compounding of the composite should be carried out in different
equipment. In our study, the mixing of the materials was carried out in an internal mixer.
Our intention is to study the breakage of fibers inside the capillary rheometer, and it is
desirable that fibers undergo the lowest breakage possible during compounding with the
polymer. In Chapter 3 it was shown that the more demanding the processing conditions,
62

the higher the breakage induced. Therefore, fibers were mixed at a low speed of 20 RPM,
and 180 C was fixed as the temperature set point. Fiber concentrations studied were 10
and 30 wt.%. The mixing procedure followed the one suggested by Rauwendaal [42]. After
material loading, the mixing time for all composite mixtures was kept constant at 6 min.

After processing in the internal mixer, a Delta bench-top band saw was used to cut the
composite specimens into smaller pieces of roughly 1 cm in thickness, so that they could be
fed into the capillary bores. In the capillary rheometer, all samples were given the same
pre-processing time (sample loading and pre-heating time) of 4 min.

Each composite sample was extruded once in the capillary rheometer following an experi-
mental factorial design that consisted of three different temperatures (190, 200 and 210 C)
and two constant shear rates (500 and 1500 s1 ). Although one load of material in the
capillary barrel allows various shear rates to be run, in this study, only one shear rate and
one temperature were evaluated with each load.

As a composite is mixed during processing, shear will cause an increase in temperature


due to viscous dissipation effects. Some extent of wood/biofiber thermal degradation is
unavoidable in real-life composites during processing due to the increased temperatures.
Thermogravimetrical analysis of aspen fibers carried out in a TGA500 (TA Instruments,
USA) showed thermal degradation of the fibers that starts around 195 C.

In this work, real-life conditions where fiber has undergone different degrees of thermal
degradation have been emulated by processing below and above the onset temperature of
fiber thermal degradation. Processing of polypropylene at 190 C for natural fiberpolymer
composites is a common practice. At 200 C and 210 C, aspen thermal degradation is
expected to occur. The shear rates were chosen from values commonly used during capillary
processing of biofiber composites [22]. Also, during processing in compounding equipment,
e.g. twin-screw extruders, screw speeds can be very high, usually around 200-500 rpm but
speeds can reach as high as 1400 rpm is some equipment [42]. At these ranges of RPM, shear
rates vary from (approximately) 10 to 103 s1 in extrusion and internal mixers [54; 163].
63

4.3.4 Materials characterization

4.3.4.1 Morphological measurements

Fibers from the composites processed in the internal mixer as well as fibers from the ma-
terials extruded in the capillary rheometer were extracted to measure their morphological
properties in order to determine the extent of fiber breakage before and after processing in
the capillary. Unprocessed fiber dimensions were also measured. Morphological properties
of fibers were measured using a HiRes FQA by Optest Equipment Inc. This equipment mea-
sures fiber length and width, amongst other fiber properties. The fiber width is commonly
considered as the fiber diameter [43; 46; 98].

For every combination of shear rate and temperature two samples were obtained, one cor-
responding to the 16-mm length die, the other to the zero-length die. Fiber extraction
was carried out on a representative portion from every sample. Fiber extraction was done
by dissolution of the polypropylene in xylene at boiling temperature . A composite sample
was immersed for 1 h in a container filled with boiling xylene. The suspension was filtered
at high temperature to prevent precipitation of the polymer. This procedure was carried
out three times for each sample to ensure that all polypropylene was dissolved. Finally, the
extracted fibers were dried at room temperature for at least 20 h to allow evaporation of
xylene prior to measurements. Three replicates from every extracted sample were measured
with a minimum of 10,000 fibers per replicate.

4.3.4.2 X-ray microtomography imaging

Four samples from the capillary rheometer were examined using a high-resolution SkyScan-
1172 desktop X-ray microtomograph (SkyScan, Belgium). The procedure for the X-ray
imaging was to set the X-ray source to 40 keV and 167 mA. No filter was applied. Each sam-
ple was rotated, in steps of 0.4, over a range of 180, using camera resolution of 2000 1024

Cervantes M. Private communication; 2006.
64

pixels. The system was controlled by a PC workstation. The cross-sectional pixel size was
4.01 mm. The raw dataset was reconstructed with the software NRecon (SkyScan, Bel-
gium). The reconstructed 2D images were saved as a stack of uncompressed 16-bit TIFF
files. These TIFF files were then used to create a three-dimensional image of microstructure
with the software Image-Pro Plus 6.1 (MediaCybernetics, USA). Two snapshots from the
3D images were taken at two different points around the samples circumference for visual
analysis.

4.4 Data analysis

Biofibers are nonhomogeneous with a distribution in fiber length. To represent the distri-
bution, it was decided to use the weightaverage length (Lw). Lw gives more weight to
longer fibers and decreases the effect of fines on the average value [119]:

X
ni L2i
Lw = X (4.1)
ni Li

where i = 1, 2, , N are the categories (bins in the distribution), n is the fiber count in
the ith category, and L is the histogram class center length in the ith category. The
processing variables analyzed included fiber concentration, temperature, and shear rate. In
addition, the two extruding bores, one with a long die and one with a zero-length die,
allowed us to decouple the influence of the elongational and shear flow.

During compounding in the internal mixer, fibers can still undergo breakage even though
the processing conditions are mild. The extent of breakage could differ when the fiber
concentration is changed (see Chapter 3). Therefore, the initial average fiber length prior
to feed into the capillary rheometer could be different when fiber concentration changes. To
better compare the fiber length reduction inside the capillary rheometer, the change of Lw
was analyzed:
65

I Lw
Delta-Lw = Lw (4.2)
ILw

where ILw and Lw are the average fiber length before and after processing in the
capillary rheometer, respectively. Delta-Lw represents the percentage of Lw reduction or
breakage experienced by the fibers.

4.5 Results and discussion

The weightaverage length, Lw, and average width, Wn, of aspen fibers before compounding
were 1.7 0.014 mm and 48.1 0.075 m, respectively. The mixing in the internal mixer,
as expected, caused some fiber breakage, and the length reduction was different for the two
fiber concentrations. The higher the fiber concentration, the greater the degree of fiber
breakage experienced in the internal mixer. The initial weightaverage fiber length (Lw)
for capillary samples with 10 wt.% of fibers was 1.06 0.01 mm and for the 30 wt.% fiber
samples was 0.68 0.007 mm. Average fiber width after processing was also reduced, but
no statistical different between both samples was found. Wn was 29 0.18 m on average.
The low RPM used also helped maintain final temperatures to a maximum value of 186 C.
Since the onset of aspen fiber thermal degradation starts around 195 C, it is expected that
no major thermal degradation of the fibers occurred during the mixing step. The later was
also visually verified as all composites were of a light brown color instead of a very dark
shade of brown (almost black) typically found in composites that have undergone severe
thermal degradation.

After mixing, the composites were processed in the capillary rheometer. Figure 4.2 shows
pressure drop readings for neat polypropylene at 190 C (similar behavior was obtained for
all polypropylene extrusions). Pressure drops under steady state are used for rehological
calculations. Polypropylene viscosities are reported in Table 4.1. They were calculated
following a standard test in the capillary rheometer for shear viscosities [71] and elongational
viscosities [72].
66

Figure 4.2: Pressure drop readings for polypropylene processed at 190 C

Pressure drops for all composites processed through the 16-mm length die showed large
fluctuations with stable readings only reached for few seconds. Pressure drops for the zero-
length die showed more steady values with sporadic surges. An example is shown in Figure
4.3 for 30 wt.% fiber composites extruded at 190 C. Pressure fluctuations in the 16-mm
length die are indicative of flow instabilities that occur at high shear rates during capillary
measurements. The flow is usually called spurt flow and was shown by the composites
in this work. Spurt flow is most frequently attributed to the stick-slip of the polymer
close the die wall [164]. Like neat polymers, woodpolymer composites also experience
stick-slip phenomenon. However, they show larger pressure fluctuations than for polymer
resins during capillary extrusion [165]. Pressure drop readings in the capillary die in this
work are, however, less steady and larger than those reported in a work on wood sawdust
composites[165]. The reason could be that the current work uses fibers with a larger aspect
ratio (L/D averaging 20 and 15 for the 10 and 30 wt.% fiber composites, respectively), and
the capillary diameter is smaller than those in the work by Hristov and Vlachopoulos [165]
67

viscosities (Pa s) at all


Table 4.1: Polypropylene shear [( a )] and elongational [e ()]
processing temperatures and shear rates.
190 C 200 C 210 C
500 s1 121 117 112
( a )
1500 s1 61 58 56

1
27,438 (=23.5)
14,560 (=31.6)
11,238 (=35.2)

e ()
[=]s

12,120 (=75.7)
7911 (=90.8)
6451 (=98.8)

The power law index was 0.44.

(L/D averaging between 3 and 5 and a die diameter of 2 mm). The larger aspect ratios of
the fibers in the current work are expected to render higher suspension viscosity [166].

This combined with the narrower capillary die would be the reason for higher resistance
during capillary flow coupled with stick-slip phenomena. All pressure drops in shear flow
were, on average, much higher than those in elongational flow. Rheological data of the
composites were not calculated due to their pressure drop behavior. Also, fiber dimensions
impact composites viscosities; to analyze this influence on the composites studied herein is
a larger task and outside the scope of this work.

There have been issues concerning the experimentation of fiber suspensions in capillary
rheometers. These issues, however, are mostly concerned with the validity of the rheo-
logical data (viscosities, shear stresses) calculated from the experiments and not to the
processing per se, since polymer suspensions with a significant load of fibers have be pro-
cessed through the capillary rheometer. A critical ratio of the particle size to die diameter
should not be surpassed so that the assumptions under which rheological calculation equa-
tions were developed remain valid. Several values of this ratio have been proposed [167].
Unfortunately, there is no other fully characterized equipment adequate to measuring the
shear and elongational viscosity of polymers and their suspensions at high shear rates than
the capillary rheometer. In capillary dies the shear rates are inversely proportional to the
third power of the capillary die radius [71]. Therefore, small capillaries are required to
68

Figure 4.3: Pressure drop readings for a sample with 30 wt.% fibers processed at 500 s1
and 190 C

develop high strain rates. In this work, however, rheological data of the suspensions was
not calculated. The elongational and shear flow developed in the capillary rheometer were
only used to assess fiber breakage.

The most relevant characteristic of the capillary rheometer for this study, is the equivalence
of its dimensions and flow developed inside to those of processing equipment. Gap openings
of about 1 mm or less are common in processing equipment. The gap height in the internal
mixer used in Chapter 3 is approximately 0.7 mm. It has been pointed out the importance
of carrying out experiments with dies of similar size to the processing equipment studied,
as it would be more relevant and avoid the extrapolation of the flow behavior [167].

With regard to the flow, a significant convergent flow is developed when the suspension
passes from the large barrel into the capillary die. Similar convergent flows are found when
the flow passes from a large section into a small opening in processing equipment, like the
69

flow before small gap regions between the screws and barrel of mixing equipment. The con-
vergent nature of entry flows contains a large component of extensional deformation that
is captured by the entry pressure drop [7277]. Even for fiber reinforced polymeric sus-
pensions, flow in convergent channels, have shown to induce large extensional deformation
[168171].

The major processing difficulty when extruding from a large reservoir into a small gap is
that instabilities could also occur through blockage of the capillary die entrance. Murty
and Modlen [170] reported that to minimize jamming of the die entrance, the viscosity of
the liquid matrix must be greater than a critical value that depends on the fiber length
and volume fraction. They carried out experiments with glass fibers (L = 3, 6 and 12 mm
and D = 10m) in golden syrup through a 0.8 mm height slit gap at various fiber volume
concentrations that reached concentrations were fiberfiber interactions are of importance
(from 0.005 to 0.04). All critical values of viscosity were around 10 Pa s. Polymer viscosities
are much higher than 10 Pa s and this would be the reason why fiber suspensions at high
load of fibers can be extruded through capillary rheometers [170]. The ability to extrude
fiber suspensions, where the length of the fibers is around the same dimensions than the
narrow passages, was explained in terms of higher viscosity fluids would exert larger drag
forces on the fibers that orient them in the direction of flow and overcome the frictional
forces that would otherwise obstruct their flow [170]. For example, glass fibers (L = 500m
and D = 10m before compouding) in polypropylene suspensions have been extruded in a
capillary rheometer with dies of 1 and 1/2 mm [171]. Also, biofiber-polymer suspensions
have been processed through capillaries at high mass concentrations, e.g. 50 wt.% jute
fiber-PP composites through 1 mm capillary die [172]. The maximum weight percent of the
suspensions in this work is 30 wt.% and their entry pressure drops were mostly steady. The
behavior of the entrance pressure drop obtained in the zero-length die is indicative of the
extensional flow experienced during processing.

The extrusion through the capillary rheometer also caused fiber length breakage. All results
of the percentage of fiber length breakage (Delta-Lw) after processing are shown in Figures
4.4 to 4.7 for all 10 and 30 wt.% fiber composites (Lw data are shown in Appendix B).
70

Opposite to the reduction in fiber length, no statistical difference between the width of
fibers before and after capillary extrusion was found. This does not indicate that some
reduction in fiber width did not occur but that very few fibers might have been broken
widthwise. The capillary processing would be the equivalent of passing the composites only
once through one narrow gap between the rotors and the mixing equipment wall. Many
passes through the capillary equipment might be needed to observe a statistical difference
in fiber width breakage.

4.5.1 Effect of shear flow and elongational flow

Samples that passed through the 16-mm length capillary die experienced both the elon-
gational flow in the convergent section from the bore to the die as well as the shear flow
developed in the capillary die. On the other hand, it is reasonable to assume that only
elongational flow has been experienced by the samples extruded through the zero length
die. Similar reasoning is used for the calculation of elongational rheological data in this
type of rheometer. The comparisons of Figure 4.4 vs Figure 4.5 and Figure 4.6 vs Figure
4.7 provide the difference in fiber breakage that occurred in each region. Statistical analysis
was performed to determine the statistically significant differences. All comments are based
on the results of the analysis.

Results show that fiber attrition not only occurs in the shear field region but also in the
extensional one. The breakage that occurs in the extensional field is of importance and,
in some cases, even has a major impact on the total fiber breakage. At 190 C, a greater
percentage of the fiber breakage actually occurs in the extensional flow. As the processing
temperature is increased to 200 C, the same behavior is observed for the samples extruded
at high shear rates. Conversely, more breakage was experienced through shear flow for
samples processed at low shear rates. At 210 C and at low fiber concentrations, almost all
breakage occurred in shear flow but at high concentrations the opposite is observed. The
intricate behavior described above points out the close interaction that the type of flow has
with the processing factors in determining the final breakage of the fibers.
71

Figure 4.4: Percent of Lw reduction for samples with 10 wt.% fibers processed through
the 16-mm length die

Figure 4.5: Percent of Lw reduction for samples with 10 wt.% fibers processed through
the zero length die. Horizontal bar indicates a difference with shear rate.
72

Figure 4.6: Percent of Lw reduction for samples with 30 wt.% fibers processed through
the 16-mm length die

Figure 4.7: Percent of Lw reduction for samples with 30 wt.% fibers processed through
the zero length die. Horizontal bar indicates a difference with shear rate.
73

X-ray microtomography of four samples extruded at 190 C was carried out to visualize the
fibers within the composites. The X-ray microtomography analysis rendered 3D images
from which snapshots were taken at two different areas around the samples circumference.
Figures 4.8 and 4.9 show the snapshots for the 10 wt.% and 30 wt.% fiber composites,
respectively, after extrusion through the 16-mm length die at high shear rate. Figure 4.10
shows the 30 wt.% fiber composites extruded through the 16-mm length die at low shear
rate. All previous samples experienced both elongational and shear flows, in that order.
Finally, Figure 4.11 corresponds to the 30 wt.% fiber composites, extruded at high shear rate
through the zero-length die. It can be noticed that all samples, even the one extruded
at low shear rates, show some degree of fiber orientation in the direction of flow. Fiber
orientation is more pronounced in some samples (Figures 4.8 and 4.11) than in others. This
seems to indicate that elongational flow in the convergent section plays a significant role in
the orientation of the biofibers in the direction of flow. This orientation is preserved even
after the fibers pass the shear flow section of the 16-mm length die. Shear flow effect of
fibers is to rotate them [173]. The reason fiber orientation is maintained after shearing flow
is probably that most of the shearing happens close to the walls, which is characteristic
of velocity profiles more blunt and less parabolic (partial plug flow) [173]. This causes the
center of the composite to flow at the same velocity while all deformation occurs close to
the capillary wall. Plug flow type in small channels has been reported for suspensions with
high concentration of solids [161].

It can also be noticed in all tomography scans of the samples that were extruded through
the 16-mm length die that the edge of the suspensions shows some depletion in fiber con-
centration. This indicates that radial migration away from the wall toward the capillary
axis occurred. This phenomenon has been documented previously [166] and could promote
slip at the wall. Slip at the wall is highly related with plug-like flow [174].

The breakage of fibers in the elongational field indicates that the ultimate strength of the
biofibers was reached upstream of the capillary die. The mechanistic process on how fibers
break in elongational flow would be complex and this author believes depends on the inter-
action between fibers. As it will be shown in the following section, the suspensions studied
74

in this work would have large interactions between fibers due to the fiber concentration.
These interactions would tend to cluster the fibers into agglomerate networks [153]. It is
through the flow induced stretching of the fiber agglomerates in the extensional flow that
the fiberfiber interaction would come into play to cause fiber breakage. As the fibers travel
through the capillary die, no more relevant breakage occurs, since the bulk of the fibers
could be flowing in a partial plug flow and the areas where shear occurs are depleted in
fibers. At this moment, the influence of other factors, e.g. concentration, temperature and
shear rate, start to gain importance.

4.5.2 Effect of fiber concentration

The impact of fiber concentration on fiber breakage can be observed when comparing Figure
4.4 versus Figure 4.6 and Figure 4.5 versus Figure 4.7. These figures show that the per-
centage of Lw reduction that occurs at 30 wt.% of fiber concentration is similar or slightly
less than the length reduction that happened to the 10 wt.% fiber composites processed at
the same temperatures and shear rates. The exception are the samples processed at 210 C
in the zero-length die, where the percentage of breakage is much higher for the 30 wt.%
fiber composites.

These results could be explained based on the concentration regime of the fibers in the
suspensions. The study of flow of suspensions distinguishes three types of fiber volume
concentration regimes, namely dilute, semi-concentrated, and concentrated. The densities
of the polymer and fiber were used to calculate the concentration regime of the composites.
The melt polypropylene density was estimated using the dilatometry of liquid polypropylene
data and the Tait equation reported in literature [175]. The average value of density (p =
0.86 g/cm3 ) obtained between the temperatures of 190 and 210 C, and an average pressure
drop of P = 15 MPa (based on the actual data for pressure drop in the capillary) was
used. For the fiber density, the value of f = 1.5 g/cm3 was used. The fiber volume fraction
for 10 wt.% fiber suspensions was 0.06, and the fiber volume fraction for the 30 wt.% fiber
suspensions was 0.19 according to:
75

1 mm

Figure 4.8: Microtomography of the 10 wt.% fiber composite processed in the capillary
rheometer at 1500 s1 and 190 C through the 16-mm length die

1 mm

Figure 4.9: Microtomography of the 30 wt.% fiber composite processed in the capillary
rheometer at 1500 s1 and 190 C through the 16-mm length die
76

1 mm

Figure 4.10: Microtomography of the 30 wt.% fiber composite processed in the capillary
rheometer at 500 s1 and 190 C through the 16-mm length die

1 mm

Figure 4.11: Microtomography of the 30 wt.% fiber composite processed in the capillary
rheometer at 1500 s1 and 190 C through the zero-length die
77

Vf Mf p
= = (4.3)
Vp + Vf Mp f + Mf p

Mf Mp
Vf = ; Vp = (4.4)
f p

where is the fiber volume fraction, Mf is the fiber mass concentration, and Mp is the
polymer mass concentration. The criteria for suspensions in the concentrated regime is
[176]:

 
D
> (4.5)
L

where D is the fiber diameter and L is the fiber length. The right side of Equation 4.5 is
the inverse of the fiber aspect ratio. This equation was proposed for suspensions where all
fibers have the same dimensions. In this work, the inverse of the average fiber aspect ratio
was used. The values for the inverse average fiber aspect ratio for the 10 wt.% and 30 wt.%
fiber suspensions were 0.05 and 0.065, respectively. Therefore, both suspensions studied in
this work fell into the concentrated regime.

In this fiber concentration regime, fiberfiber contacts are considered one of the major
interactions that could cause fiber breakage, as rubbing of fibers against one another dur-
ing their movement could induce bending stresses, surface abrasion, or potentially induce
stress concentrations. This seems to be more relevant for the 10 wt.% fiber composites.
At 30 wt.% of fiber concentration, fibers might have got so close they could have formed
a network, which was strengthened by the friction forces at the contact point between the
fibers in a type of elastic fiber interlocking proposed by Soszynksi and Kerekes [153] for
floc formation, thus decreasing fiber mobility and damage. In addition, partial plug flow
has been reported to increase with concentration of suspended particles [173]. Thus, the
region where deformation happens would be larger for 10 wt.% fiber composites.
78

On the other hand, samples processed at 210 C in the zero-length die show much greater
breakage for composites with the highest fraction of fibers. The authors believe that this in-
crease in breakage with fiber concentration is more a consequence of the interaction between
concentration and temperature, as explained in the next section.

4.5.3 Effect of processing temperature

The effect of processing temperature is different in elongational flow than in shear flow.
In elongational flow, all samples but one with 10 wt.% fiber concentration that were ex-
truded through the zero-length capillary die (Figure 4.5) show basically no dependence
of fiber breakage on temperature. The exception was the composite processed at 200 C
and 1500 s1 . Independence on temperature also seems to be the case for 30 wt.% fiber
composites processed through the zero-length die at 190 C (both shear rates) and 200 C
at 1500 s1 (Figure 4.7). As processing temperature increases, the viscosity and thus the
stresses in the system decrease and less breakage is expected, as was also reported in the
work by Joseph et al [87]. Maximum processing temperatures in the work of Joseph et al.
[87] were below 180 C. Thermogravimetrical analysis of aspen fibers showed that major
thermal degradation of fibers started around 195 C. Thermal degradation is experienced
by the lignin and hemicelluloses, weakening the fiber [143]. At 190 C, major fiber thermal
degradation has not started, while at 210 C, it has started in earnest, and 200 C is the
middle between these two opposites. The interaction between the factors strain rate and
temperature will play a major role through the balance between stresses developed and
the degree of fiber thermal degradation. In samples that do not show change in the level
of breakage through temperature increase, the lower stresses were sufficient to break the
biofibers, whose mechanical integrity had been reduced. In the samples with 10 wt.% fiber
extruded at 200 C and 1500 s1 , the stresses were enough to cause more breakage than in
the samples at the other temperatures. On the other hand, in samples of 30 % fiber ex-
truded at 200 C and 500 s1 , the stresses were not high enough. In the case of the 30 wt.%
fiber composites extruded at 210 C (Figure 4.7), it could be that the higher fiber breakage
is the result of the interaction between the temperature, the high concentration, and the
79

type of flow. In the elongational field, at this temperature, fibers are weaker and are forced
to increase their interaction as the flow deforms to enter the capillary die.

The effect of temperature on the breakage of fibers for samples that have passed through
16-mm length die (Figures 4.4 and 4.6) shows a different trend. As the processing tem-
perature increases, the level of fiber breakage grows. This trend could be due to increased
viscous dissipation effects associated with the capillary die. Dissipative heating induces a
temperature rise in the melt, proportional to the pressure drop. Temperature rise in the
capillary is more noticeable for long dies (e.g. L/D = 32). Although the capillary die in this
work is moderate in length, large pressure drops were registered during the compounding.
The average adiabatic temperature rise can be calculated by

P
T = (4.6)
Cp

Assuming a melt polymer density of = 0.86 g/cm3 , a suspensions heat capacity of Cp =


1800 J/kg C, and an average pressure drop in the capillary of P = 15 MPa, we get
a temperature rise of about 10 C. The actual conditions during extrusion are between
adiabatic and that of constant temperature at the wall, and thus, the temperature rise
might be lower. However, the increased temperature, above the set point, will promote
more thermal degradation of the fibers, rendering them more prone to damage.

4.5.4 Effect of shear/stretch rate

In general, the level of breakage seems not to be influenced by the level of shear and
stretch rates. This is opposite to what is expected, as higher shear/stretch rates promote
larger stresses during processing. This result seems to indicate that the increased breakage
observed at higher RPMs in reality could be due more to the number of passes through
the narrow gaps rather than to the increased stresses due to higher shear/ stretch rates.
Indeed, the extrusion through the capillary dies was the equivalent of one pass through the
narrow gap. During mixing, a sample passes many times through the narrow regions, and
80

the number of passes through these regions increases as the RPMs increase, for the same
total mixing time. The increase in stresses could in fact have some influence, but it might
be of lower importance than the number of times a sample passes through the narrow gaps.
The only cases where there is difference are for samples extruded in the zero-length die at
200 C. The composites processed at 1500 s1 had more fiber breakage than those processed
at 500 s1 . This could be due to the interaction between the processing shear rates and
temperature, where at 200 C, the fibers have started to thermally degrade.

4.6 Conclusions

Aspen fiberpolypropylene composites with 10 and 30 wt.% of fiber were extruded in a two
bore capillary rheometer at two shear rates (500 and 1500 s1 ) and three temperatures (190,
200 and 210 C) in order to determine the influence of the shear and elongational flows on
fiber breakage. The effect of processing conditions was also assessed. The experiments
intend to simulate one pass of the material through one of the narrow gaps between mixing
elements and a processing equipment barrel. During compounding, composite mixtures
pass many times through these gaps.

Fiber length breakage was detected by the reduction of the weightaverage fiber length. The
analysis showed that it was possible to decouple the length breakage that occurred in the
shear flow region from the breakage that occurred in the elongational field. The breakage
that occurred in the extensional field was shown to be of importance and, in some cases,
even had a major impact on the total fiber breakage. The influence of the type of flow,
elongational versus shear, works in conjunction with the processing conditions to determine
the final breakage level of the fibers.

In general, the percentage of length reduction for the 30 wt.% fiber composites was similar or
slightly less than for the 10 wt.% fiber composites. These results are explained in terms of the
fiber concentration regime of the suspensions. The two composites fall into the concentrated
fiber regime. The amount of fibers is high enough that their compaction might reduce fiber
81

mobility and damage. However, fibers in the 10 wt.% fiber composites seemed to be less
compacted, which could have permitted them to have more mobility causing additional
breakage through fiberfiber contacts. The processing temperature factor showed that in
the elongational field, the balance between stresses developed in the system and degree
of fiber thermal degradation will play a major role. On the other hand, for samples that
also passed through the shear region, it was found that higher temperatures are conducive
to more breakage. The latter was possibly due to increased thermal degradation of fibers
caused by additional temperature rise in the capillary die during extrusion.

Finally, the level of shear/stretch rates at which composites were processed was shown to be
almost non-influential on fiber breakage. Shear rates are proportional to flow stresses during
processing. The larger breakage observed at higher RPMs during processing in reality might
be due more to the number of passes through the narrow gaps than to the increased level
of stresses. The number of passes through these regions increases as RPM increases for the
same total mixing time.

The work presented here is experimental in nature. Other biofibers could have different fail-
ure mechanism than the biofibers used herein. However, the authors believe that the overall
understanding from these experiments will be highly valuable for composite researchers and
manufacturers for developing biofiber composites with enhanced performance.
Chapter 5
Effect of processing conditions on the
extent of biofiber breakage in a twin-screw
extruder

5.1 Abstract

In this study, the influence of processing variables and initial fiber length on fiber breakage
during processing in a twin-screw extruder was investigated. Composites of biofibers were
processed with a polypropylene matrix. Processing variables included the screws RPM,
the feeding of the polymer and fibers, fiber concentration, and both set-point and final
temperatures. The influence of the variables was analyzed using multivariate statistics.
The feeders, the screw speed and initial fiber length showed the greatest effect on final fiber
length. Fiber concentration fell into the semi-concentrated and concentrated regimes, and
this seemed to have influence on the final effect of processing variables.


The contents of this chapter have been submitted for publication.

82
83

5.2 Introduction

The interest in using biofibers as reinforcements for thermoplastics has grown in importance
over the last decades. Biofibers offer several advantages over synthetic fibers. From the
environmental point of view, natural fibers are a renewable resource and they are considered
as carbon neutral materials [177]. The economic advantages could be substantial as the new
opportunities for usage would increase the added-value of the biofibers. Biofibers have good
specific mechanical properties [2] and cause less abrasive damage to processing equipment
than synthetic fibers. Due to these advantages, composite products made of biofibers and
thermoplastics are continually introduced into the market. Automotive and construction are
some of the industries that have been most involved in the biofiber composite development.
And continuing research on the incorporation of biofibers into thermoplastics is important,
as there are still many issues that remain to be resolved.

For reinforcement of the polymer matrix (e.g. increase strength and toughness) preservation
of an adequate fiber aspect ratio after processing is most important. Aspect ratio is defined
as the ratio of fiber length (L) to diameter (D). Minimum (critical) fiber aspect ratios are
needed to take full advantage of the reinforcement capabilities of the fibers when loads are
transferred from the matrix to the fibers. The critical aspect ratio is dependent on the
fiber-matrix bond strength and the fiber ultimate strength [15].

Attention to aspect ratio is especially important for short biofiber-thermoplastic composites


as they are processed using standard thermoplastic equipment, such as mixers and twin-
screw extruders (TSE). The dynamic nature of the processing in these pieces of equipment,
results in the mixture being constantly subjected to high stresses, causing a significant
degree of fiber breakage.

Various studies of biofiber-thermoplastic composites in internal mixers have reported fiber


breakage. In Chapter 3 it was found that final fiber length was almost solely dependent
on processing conditions regardless of the initial fiber morphology, while the reduction in
fiber width showed dependence on other factors including both fiber initial morphology
84

and processing conditions. Similar results on cellulose fibers were obtained by Le Duc et
al. [108], where final length did not show dependence on the initial length of the fibers.
Increased fiber breakage of short pineapple leaf fiber was observed when mixing rotor speeds
increased during processing with a low density polyethylene matrix [91]. Close relationship
between processing conditions and fiber breakage has also been reported for other biofibers
such as sisal [87], jute [89], hemp [90] and flax [93].

Fiber breakage has also been reported during twin-screw extrusion. However, there are not
as many studies dedicated to analyze the influence of the different processing parameters
on biofiber breakage. A study on flax fibers mixed in a polypropylene matrix showed
that fiber length was greatly affected by the processing speed, fiber content and extruder
layout [94]. Also, when processed with a quasi-Newtonian fluid (polycaprolactone), hemp
fiber lengths decreased at high screw speeds, low feeding rates, lower moisture content and
higher temperatures [95; 96]. The impact of the screw speed was explained in terms of the
increase in shear stresses with higher RPMs, while they attributed the increased breakage
with low feeding rates to the increased resident time. Accordingly, decreased residence time
was attributed as the factor behind the preservation of lengths of miscanthus and bamboo
fibers when mixed with poly(lactic acid). No statistical difference between initial and final
fiber length was found at some higher levels of feeding rates and screw speeds [97].

Biofibers have also shown reduction of their initial diameter [96]. This reduction has been
related to the separation of fiber bundles leading to a lower average diameter. However,
reduction of fiber diameter has not always occurred and it has been independent of the
occurrence of biofiber length breakage [97; 99].

All previous analyses have provided good insight on the breakage of biofibers during pro-
cessing. However, biofibers are not homogeneous but rather they show strong property
variability even among the same type of fibers. Biofibers are microscopically inhomoge-
neous as they are themselves composites of cellulose, a semicrystalline polysaccharide, and
various amorphous polymers, e.g. pectins or hemicellulose and lignin. The arrangement,
orientation and percentage of the different constituents within a fiber are the result of
85

the natural process of plant growth, which is influenced by the environment, conditions of
growth, etc. And fiber properties depend on their composition, structure and also on the
method of fiber extraction [49]. It is for these reasons that results on one type of biofiber
may not be extrapolated to other types of biofibers. In order to gain a better understand-
ing of how different variables influence biofiber breakage further experimental studies with
different biofibers are required.

A contribution to the study of biofiber breakage during processing is shown in this work by
evaluating the effect of different processing conditions and initial fiber length on the break-
age of aspen biofibers when mixed with a polypropylene matrix in a twin-screw extruder.
The nature of the extrusion process leads many times to statistical correlation among pro-
cessing variables. Different methods to address this issue during analysis exist [178]. In
this work, Projection to Latent Structures or PLS was used [146]. The processing variables
studied included the feeding rates of the polymer and fiber, the mixing speed rate, fiber
concentration and temperature.

5.3 Experimental

5.3.1 Materials

Wood fibers from aspen provided by FPInnovations-Forintek Division Eastern Region were
used for the composites. The fibers were thermomechanically extracted in a medium density
fiberboard, MDF, pilot plant at FPInnovations. Fiber extraction was carried out at two
refining conditions that gave batches of two different lengths, which will be referred to as
aspen long (AL) and aspen short (AS) fibers. Both AL and AS fibers were refined using the
same preheating time and pressure of 2.5 min and 6 bar, respectively. The refining energy
and speed were as follows: 150 kW h/t and 2000 RPM for AL fibers; and 250 kW h/t and
2800 RPM for AS fibers. All batches of fibers were provided dried.
86

The homopolymer polypropylene (PP) Profax 6523 (Lyondellbasell, the Netherlands) was
used as the matrix for the composites. This polymer has a nominal melt flow rate of
4 g/10 min (230 C/2.16). The peak melting point as determined using a DSC1000 (TA
Instruments, USA) was 160 C.

5.3.2 Composites Processing

Composites were processed in a 25 mm co-rotating twin-screw extruder (ONYX extrusion


technology, Canada). The twin-screw extruder (TSE) has ten heating zones, two feeders,
two vent ports and an L/D ratio of 40:1. Material is fed through basic open-loop rotating-
screw volumetric feeders. The first feeder deposits its contents from above, while the second
feeder conveys material through the side of the extruder, between the heating zones 3 and 4.
The polypropylene was fed into the first feeding port and the wood fibers were fed into the
second port. The split feed was used in order to focus on the study of fiber breakage that
happens in the mixing section of the extruder. At the second port section the polypropylene
is expected to be melted, thus minimizing fiber breakage due to solid polymer melting [179].
The processing parameters included five screw speeds or RPMs (80, 120, 200, 280 and 300)
and three temperatures (190, 200 and 210 C). The full processing temperature profiles
are shown in Table 5.1. Feeding rates of both the polypropylene and fibers were varied by
adjusting the speeds of the feeders, which is the standard procedure when using volumetric
feeders. To have an approximation of the mass rate fed into the extruder, both feeders were
detached from the extruder and a first calibration of their feeding was carried out. Three
speeds were used for the polymer feeder (3, 6 and 9 RPMs, aprox. 1.6, 1.9 and 2.2 kg/h,
respectively) while four speeds were used for the fiber feeder (10, 15, 40 and 60 RPMs aprox.
AL: 0.07, 0.11, 0.29 and 0.44 kg/h and AS: 0.08, 0.12, 0.32 and 0.48 kg/h). Composites
rendered starve fed conditions in the TSE [62]. Tables 5.2 and 5.3 show the experimental
design carried out. All experiments were processed once using the same screw design. Thus
the effect of screw design on fiber breakage was not studied in this work. No shaping die
was used at the end of the extruder to prevent fiber breakage due to die entrance flow. The
throughput from every extrusion run was also calculated through the average weight of 15
87

samples, each one collected every minute during the extrusion process. The throughput of
only polypropylene resin obtained for every set of the processing variables PP-feed, Temp,
and RPM was also calculated.

Table 5.1: Twin-screw extruders processing temperature profiles. Temperatures in C


zone 1 zone 2 zone 3 zone 4 zone 5 zones 6 to 10
Temperature profile 190: 170 175 180 190 190 190
Temperature profile 200: 170 175 180 195 195 200
Temperature profile 210: 170 175 180 195 200 210

5.3.3 Materials Characterization

Fiber length before and after processing was measured using a Fiber Quality Analyzer,
FQA, (Optest Equipment Inc, Canada). The FQA combines hydraulic, optical, and image-
processing systems to measure fiber length, among other fiber data. Sheath flow of a dilute
suspension of fibers transports them into a cell that renders an almost two-dimensional
plane. Flow hydrodynamics orients the fibers in the cell where the optical and image
systems measure fiber length [145]. Processed fibers were extracted from the composites in
order to measure their morphological properties. From each one of the 15 samples that were
collected to calculate the throughput, a representative portion was taken for fiber extraction.
All portions from every combination of processing conditions, that were selected for fiber
extraction, were combined and underwent the extraction process simultaneously.

Fiber extraction was done by dissolution of the polypropylene in xylene at boiling temper-
ature. A composite sample was immersed for one hour in a container filled with boiling
xylene. The suspension was filtered at high temperature to prevent precipitation of the
polymer. This procedure was carried out three times for each sample to ensure that all
polypropylene was dissolved. Finally, the extracted fibers were dried at room temperature
for at least 20 h to allow evaporation of xylene prior to measurements. Three replicates
from every extracted sample were measured in the FQA with a minimum of 10,000 fibers
per replicate.
88

Table 5.2: Experimental design of aspen long, AL, samples extruded in the twin-screw
extruder.
RPM PP feeders RPM Fiber feeders RPM Temperature
(PP-feed) (F-feed) (C)
120 6 60 190
200 6 60 190
280 6 60 190
120 9 60 190
200 9 60 190
280 9 60 190
120 9 15 190
200 9 15 190
280 9 15 190
120 6 40 190
200 6 40 190
280 6 40 190
120 9 60 200
200 9 60 200
300 9 60 200

5.4 Analysis

5.4.1 Processing Variables and Fiber Properties

The processing variables studied included the screws intensity of mixing (RPM), the
polypropylene feed (PP-feed), the fibers feed (F-feed). Also, two temperature variables
were included. One corresponds to the set-point temperature (Temp). These were the tem-
peratures at which the last extruders zones were heated up prior to the extrusion process
(190, 200 and 210 C). However, the temperature of the mixture increases during extrusion
due to viscous heat dissipation [62]. Final temperatures (F-temp) of the composites were
registered for aspen long, AL, samples in the last zone of the extruder, before the material
exit, and included in the analysis of the data.
89

Table 5.3: Experimental design of aspen short, AS, samples extruded in the twin-screw
extruder.
RPM PP feeders RPM Fiber feeders RPM Temperature
(PP-feed) (F-feed) (C)
120 6 40 190
120 6 40 200
120 6 40 210
120 3 40 190
120 3 40 200
120 3 40 210
120 6 10 190
120 6 10 200
120 6 10 210
80 3 40 190
80 3 40 200
80 3 40 210
80 6 40 190
80 6 40 200
80 6 40 210
80 6 10 190
80 6 10 200
80 6 10 210

Regarding the polymer and fiber feeding, an initial calibration of the volumetric feeders was
carried out to have an approximation of the relationship between the polymer and fibers
mass rate input with the feeders RPM. However, mass rate fed through this type of feeders
is not constant and some difficulties might arise when feeding low density materials, like the
fibers in this work. This occurs because volumetric feeders cannot self-adjust to compensate
for material bulk density variations and are sensitive to material build-up, which affects
their calibration. Adjustment to maintain a constant throughput is carried out manually.
Some studies have ingeniously fed the fibers manually, directly into the extruder housing
to avoid variation in the fiber mass rate variable [94; 95]. Manual feeding of this sort was
not possible in the twin-screw extruder used in this work as the second feeders opening
90

is on the side and not at the top of the housing. Also, constant calibration of the feeders
was not possible due to fibers supply constraint. All in all, this author believes that the
extrusion carried out here reflects more the reality of this type of process. However, due to
the difficulties mentioned above, it was decided to use as feeding variables unit the RPM
of both feeders. RPM is constant and is the process set-point parameter. This way the
effect of the feeders will be more precise as it will avoid both, using a value of mass rate
that might be a misrepresentation of the actual value as well as doing a regression over a
parameter that has already undergone a regression via RPM-mass rate calibration.

Finally, to account for the actual ratio of fibers-to-polymer in the final samples, fiber con-
centration (Conc) for every combination of fiber-to-polymer feeding was calculated using
the extrusion throughputs (mass flow rate) and was also included in the analysis. Fiber
concentration of the composites was calculated following the equation: 100(Mc Mp )/Mc ,
where Mc and Mp are the mass flow rates of the composite and polymer, respectively.

Length of the fibers was measured using an FQA. Since biofiber dimensions are not uniform,
a distribution of lengths was obtained. It is customary in the study of biofibers to quantify
and characterize the effect of the fiber length distribution by a single value, a distribution
average [145]. The fiber length was represented by the length-weighted average (Lw), which
it will be referred to as the weight-average fiber length. Lw is the average most commonly
used to characterize fiber length because it gives more weight to longer fibers and decreases
the effect of fines on the length values [119]. The equations for the length-weighted average
(Lw) and its standard error (sLw ) are:

X
ni L2i
Lw = X (5.1)
ni Li

r
s2 ( Li )2 N0 ni Li (Li Lw)2
P P
2
sLw = ; e= P ; s = (5.2)
(Li )2 (N 0 1) Li
P
e

where i = 1, 2, M categories (bins in the distribution); n is the fiber count in the ith
category, L is the histogram class center length in the ith category, N 0 is the number of
91

non-zero weights and s is the weighted standard deviation of the sample [120].

Fiber lengths for the unprocessed fibers are identified as UP-Lw and processed fibers
lengths are identified as Lw. Also, a factor named Delta-Lw that accounts for the
change in fiber dimensions was calculated according to:

Delta-Lw = UP-Lw Lw (5.3)

All processing conditions and average fiber lengths that correspond to the properties before
processing are the variables considered to influence fiber breakage, and they are called the
X-space. On the other hand, Lw and Delta-Lw are considered the response properties or
Y -space.

Processing conditions, such as the set-point temperature and screws and feeders RPM are
factors that can be set to the desired values according to the experimental design. However,
there are also variables for which this is not possible, such as the initial weight-average fiber
length, the final processing temperature and fiber concentration. This type of variables will
promote correlation among the variables. In addition, final temperatures were not recorded
for AS fiber composites. It was decided, therefore, to carry out the analysis of variance
using Projection to Latent Structures (PLS). PLS successfully handles missing data and
extracts the meaningful variance correlation in the X- and Y -spaces while relating the X-
and Y -spaces to each other [146; 147].

5.4.2 Projection to Latent Structures (PLS)

Projection to Latent Structures (PLS) is a robust multivariate regression technique that has
various advantages over multiple linear regression (MLR) [146; 180]. Among the advantages
of PLS are its capability to analyze data with mild and strong collinearity (correlation),
noise as well as being suitable to handle missing data for both input-variables (X-space)
and responses (Y -space). Also, multiple response variables in the Y -space can be analyzed
simultaneously. PLS reduces the X-space to a smaller number of non-collinear latent
92

variables, which are known in PLS as principal components or scores. In this sense, scores
are the link between processing variables and the experimental results.

The two main graphs given by PLS are the w*c plot and the scores plot. The relationship
among all input-variables, the relationship between all responses and the variables influence
on the responses are all shown in the w*c plot. The scores plot shows the relationship among
all experimental data. Two parameters are provided to gauge the goodness of the regression
model, R2 and Q2 . R2 is the standard coefficient of determination that indicates how much
data variance is explained by the model. Q2 , on the other hand, indicates how well the
regression model will predict new data and it also ranges between 0 and 1, where values
closer to 1 indicate a better data prediction. Q2 is used to prevent overfitting of the data,
which happens when unnecessary regression coefficients are added to the model to increase
the value of R2 . In order to give all data the same weight and importance in the analysis, it
is standard in PLS that the raw data is centered to have a mean of zero and scaled to unit
variance. PLS models can be converted to standard linear regression equations. Details
of the PLS analysis can be found elsewhere [146]. The SIMCA-P software by Umetrics
was used to develop the PLS model in this work. SIMCA-P carried out the centering and
scaling of the data and presents the graphs in this transformed state for better visualization
of variable influence and importance.

5.5 Results and discussion

The two different refining conditions rendered aspen fibers with two different length dis-
tributions, named aspen long (AL) and aspen short (AS) fibers. Aspen long fibers had a
weight-average length, UP-Lw, of 1.74 0.014 mm and a polydispersity index of 4 before
composite processing. Unprocessed aspen short fiber values were: UP-Lw=1.55 0.011 mm
and polydispersity=2.8.

Aspen fibers were shortened during composite processing, with every combination of pro-
cessing conditions breaking the fibers to a different degree. The statistical analysis to
93

determine the influence of processing variables and initial average fiber length, UP-Lw, on
the breakage of AL and AS fibers during processing was carried out using the multivariate
PLS technique. PLS analyzes data mean centered and scaled to unit variance, (averages of
the processed data as used in PLS are shown in Appendix C).

The experimental design considered AL and AS fiber composites processed under the same
combinations of processing conditions. However, due to issues outside the authors control,
not all samples were processed. In the end, AL and AS composites had different combi-
nation of processing conditions except for one sample in each fiber type. The processing
conditions for these samples were RPM=120, PP-feed=6, F-feed=40 and Temp=190 C.
Due to different feeding rates between AL and AS, the fiber concentration reached by each
sample was different. The sample with AL fibers had a fiber concentration of 4.3 wt.%,
while the sample with AS fibers showed a higher concentration of 15.6 wt.%.

The overall fitting of the data with PLS gives three principal components that on average
explain 80 % of the variance (R2 ) in the X-space and 62 % of the variance in the Y -space.
The ability of the PLS model to predict the Y -space variance (Q2 ) was 53 %. Figure 5.1
shows the breakdown per component of the values of R2 and Q2 . It can be noticed that the
first two principal components explain most of the variance.

5.5.1 Loadings Plot

To graphically assess the influence of variables on Lw and Delta-Lw the loadings plot is used.
The loadings plot between the first (w*c[1]) and second (w*c[2]) principal components is
shown in Figure 5.2. This plot shows multiple factor correlations in a simple graphical form.
Variables that are located near each other are positively correlated. Negative correlation is
observed when variables are located on opposite sides of the origin.

It can be observed, for example, the proximity of the variables UP-Lw, F-feed and RPM
with F-temp. This indicates that higher values of the first three variables are related to
higher final temperatures. Indeed, longer fibers or an increase in the amount of fibers
94

Figure 5.1: Values of R2 and Q2 for each fiber property. PC1: Principal component 1.
PC2: Principal component 1. PC3: Principal component 3.

would increase the shear viscosity of the system [181] and thus giving higher viscous heat
dissipation. The increase in the screws RPM will also increase viscous heat dissipation
through shearing of the polymer [62].

On the other side, the fibers initial average length (UP-Lw) and the average fiber concen-
tration (Conc) are opposite to each other. This indicates a negative correlation between
these variables in this work where the shorter the value of UP-Lw the higher the fiber
concentration reached in the sample. Fiber concentration of AL fiber composites ranged
from 2.3 to 15.6 % with a median of 8 % while fiber concentration of AS fiber composites
ranged from 6.3 to 17.1 % with a median of 14 %. Unprocessed aspen long fibers had both
a higher UP-Lw and a wider length distribution than aspen short fibers. Packing density of
particles is expected to increase as the particle size distribution widens [166], and thus fiber
concentration for AL fibers would have been expected to be higher than for AS fibers. The
opposite behavior observed in this work could be due to the better packing of AL fibers
95

Figure 5.2: Loadings plot of principal components 1 and 2.

worked against their feeding in the fibers feeder channel. Here, there is no melted polymer
that could behave as lubricant between the agglomerates and the channel walls and thus
their passing through the channel was slower, more difficult. The slower feeding of the AL
fibers was confirmed during the calibration of the side feeder where AL fibers gave lower
mass rate per RPM.
96

5.5.2 Scores Plot

All data points are shown in the scores plot of principal component 1 versus principal com-
ponent 2 (Figure 5.3). Score values of each sample are determined by a linear combination
of the variables (X-space). Each sample can be assigned a set of score values; a value per
component, which are called t[1] for principal component 1 and t[2] for principal component
2. It can be observed that there is a clear separation between samples along the horizontal
axis. Samples on the negative side of t[1] are the composites of AL fibers, while AS fiber
composites are on the positive side. Also it is noticed that AS fiber composites cluster into
two groups along t[2]. A contribution plot between the groups is used to identify what
separates the clusters of data. Contribution plots in a PLS analysis show differences for all
factors between two data points or data sets. The contribution plot between AL and AS
fiber composites is shown in Figure 5.4. It shows that the main factors influencing the sepa-
ration are UP-Lw and the screws RPM. AL fibers had longer UP-Lw and their composites
were processed, in general, at higher screw speeds. Figure 5.5 shows the contribution plot
between the two groups of AS fiber composites. The main factor influencing their separa-
tion was PP-feed and to a lesser extent F-feed. This points out the great influence that the
amount of polymer had on the final length of the fibers.

The position of any experiment in the scores plot is readily explained by the loadings plot
as the two can be visualized as superimposed on one another. For example, AS samples on
the first quadrant of the scores plot (Figure 5.3) are closer to Lw, which is located in the
first quadrant in Figure 5.2, indicating that these composites ended up with longer fiber
lengths than the cluster of AS samples on the fourth quadrant of Figure 5.3.

5.5.3 Factor Influence on Lw and Delta-Lw

The influence of the processing variables and UP-Lw on Lw including all principal compo-
nents is shown in the coefficients plot of Figure 5.6. Final fiber length is positively corre-
lated to PP-feed, Conc and Temp, while negative correlation with RPM, F-feed, UP-Lw and
97

Figure 5.3: Scores plot of principal components 1 and 2.  AL and N AS.

F-temp is observed. The variable importance on projection or VIP (Figure 5.7) indicates
the importance of a variable for the entire system. The VIP plot shows that the importance
of the factors from higher to lower is: PP-feed, RPM, UP-Lw, F-feed, Conc, Temp, F-Temp.
According to Wold [182] the more relevant variables are those with values higher than 0.8.

Longer unprocessed fiber lengths (UP-Lw) would increase fiber-to-fiber interaction and the
likelihood of fiber damage resulting in lower final fiber lengths.

The positive correlation of the polymer feeding with Lw could be due to the level of fiber
concentration in the system. The study of flow of suspensions distinguishes three types of
fiber volume concentration regimes, namely dilute, semi-concentrated, and concentrated.
In the dilute regime fibers flow and rotate freely in the suspension without encountering
other fibers. As the concentration increases, the semi-concentrated regime is reached where
the distance between fibers shortens and they start to interact with one another. Further
increase in fiber concentration would bring the distance between fibers so close that fiber
98

Figure 5.4: Contribution plot between the groups of AL and AS fiber samples.

Figure 5.5: Contribution plot between the groups of AS fiber samples.


99

Figure 5.6: Coefficients plot for the final average fiber length, Lw, property.

motion will be restricted and the suspension might behave as a solid [176]. The criteria for
suspensions in the semi-concentrated regime is [176]:

 2  
D D
<< (5.4)
L L

where is the fiber volume concentration, D is the fiber diameter and L is the fiber length.
2
Fiber volume concentrations lower than D L indicate dilute suspensions, while fiber volume
D
concentrations greater than L indicate concentrated suspensions. D/L is the inverse of the
fiber aspect ratio. This equation was proposed for suspensions where all fibers have the
same dimensions. However, natural fibers do not have a single value of length. Also, the
diameter of the fibers after processing is not measured by the FQA equipment. The mean
values for each fiber type between the minimum and maximum fiber average aspect ratios
obtained in Chapter 3 were used as estimates in Equation 5.4. These values were AL:
L/D = 18.3 and AS: L/D = 23.3.
100

To calculate the fiber volume concentration the following equations were used:

Vf Mf p
= = (5.5)
Vp + Vf Mp f + Mf p

Mf Mp
Vf = ; Vp = (5.6)
f p

where Mf and Mp are the fiber and polymer mass concentration, respectively. The minimum
and maximum fiber mass concentrations reported above for AL and AS were used and it
was assumed a melt polypropylene density, p , of 0.8 g/cm3 and a fiber density, f , of 1.5
g/cm3 . The suspensions in this work fall between the semi-concentrated and concentrated
regime according to:

AL: min = 0.013 max = 0.047 (5.7)

AL (Equation 5.4): 0.003 < < 0.055 (5.8)

AS: min = 0.037 max = 0.085 (5.9)

AS (Equation 5.4): 0.002 < < 0.043 (5.10)

Fiberpolymer interaction is an important factor that is found conducive to fiber breakage.


This would indicate a negative correlation between PP-Feed and Lw. However, the high level
of fiber in the system would hinder the hydrodynamic interactions on the fibers in favour
of direct fiberfiber mechanical interaction [181], which would become the more important
factor for fiber breakage. The polymer then would act more as a lubricant between the
fibers and reduce their interaction, helping preserve fiber length.

With respect to the effect of the fibers, two factors were considered, the fibers feed (F-feed)
and fiber concentration (Conc). These variables show opposite influence on the final fiber
101

Figure 5.7: Variable importance on projection.

length. F-feed shows a negative correlation while Conc displays a positive one. Their
opposite influence on Lw is interesting. To better visualize the reason behind this behavior,
the samples AL and AS were analyzed separately. Figures 5.8 and 5.9 show the coefficients
for Lw for each type of fiber, AL and AS, when their samples were analyzed separately. It
can be noticed that both F-feed and Conc have actually the same negative influence on Lw
for AL fibers and both have the same positive influence on Lw for AS fibers.

The different effect, positive or negative correlation, on Lw could also be due to the fiber
concentration regime in the composites. It can be seen in Equations 5.7 to 5.10 that AL
fiber samples fall within the semi-concentrated regime, while AS fiber samples span from
the upper lever of the semi-concentrated regime to the concentrated regime. At the semi-
concentrated levels, where fiber interaction starts gaining importance, the expected result
of a negative correlation between the amount of fibers in the composite and Lw is observed
(AL fiber samples). This due to an increase in the amount of fibers in the mixture which
increases the mechanical contacts between fibers and thus the likelihood of fiber breakage
102

Figure 5.8: Analysis of AL samples only. Coefficients plot for the final average fiber
length, Lw, property.

and lower Lw. Similar results were found in Chapter 3. Nevertheless, as the level of fiber
enters the concentrated regime, the close proximity between fibers might compact them
into a close network similar to the ones proposed by Soszynksi and Kerekes [153] where
friction at the fiber contacts would strengthen the flocs, therefore decreasing fiber mobility
and damage. This would revert the effect of the amount of fibers in the mixture on Lw from
negative to positive (AS fiber samples), as higher concentrations would protect the fibers
more. A similar result was found in the study of fiber breakage in Chapter 4.

Negative correlation between the amount of fibers and Lw in the concentrated regime would
be obtained if fiber damage increases with fiber concentration. The latter would require
that flocs formed in the concentrated regime pass a minimum number of times through
the regions where dispersive mixing is experienced to fully break them apart. And such
conditions might not have been reached during processing. In this sense, the results in
the twin-screw extruder in this work are a compromise between the two studies mentioned
103

Figure 5.9: Analysis of AS samples only. Coefficients plot for the final average fiber
length, Lw, property.

above, where composites in the internal mixer (Chapter 3) where mixed for 6 min. while
in the capillary rheometer (Chapter 4) the samples were extruded for only one minute.
However, an analysis of the screw design, residence time distribution and their effect on the
mixing is outside the scope of this work.

When all samples are analyzed together, PLS seems to combine the value of the coefficients
on the global system. Comparing the analysis of all samples together against the separate
analyses of AL and AS, the global analysis helped include the influence of initial fiber length,
UP-Lw. It was also found that the global analysis gave higher R2 and Q2 . And finally, the
global analysis reflects the transition between these two regimes by showing a balance on
the influence of the amount of fiber on Lw through the effect of the variables F-feed and
Conc.
104

Higher RPMs will promote greater shear rates and higher stresses on the suspension [62].
This will likely increase fiber breakage as there will be a higher probability of reaching
the fibers ultimate strength during processing. In addition, increasing the revolutions per
minute will also increase the probability of the sample to pass more times through the
regions where breakage occurs through high shear and elongational stresses. Therefore the
negative correlation between RPM and Lw is observed.

Finally, based on the variable importance on projection plot of the global analysis (Figure
5.7) and those of the separate analyses for AL and AS (Figures not shown), both tempera-
ture variables had negligible influence on Lw [146] and would not be statistically significant.
However, a negative correlation between temperature and Lw was expected, as obtained by
the F-temp variable. This is because final temperatures (F-temp: 206 to 227 C) fell above
the onset of fiber thermal degradation. Thermogravimetrical analysis using a TGA500 (TA
Instruments, USA) showed that aspen fibers onset of thermal degradation starts at 195 C.

The responses Lw and Delta-Lw (Figure 5.10) have different meanings. Lw represents the
average length after processing, while Delta-Lw is a measure of the degree of breakage that
occurs during processing. Results in the internal mixer of Chapter 3 showed that regardless
of the initial fiber morphology, final fiber lengths (Lw) depended almost solely on processing
conditions. And Delta-Lw depended not only on processing conditions but also on fibers
initial morphology, in which the level of breakage was higher for fibers whose initial average
length was longer. In the current work, however, Lw also shows dependence on the initial
average fiber lengths and the dependence of Delta-Lw on the processing variables is almost
the exact opposite to the dependence of Lw. The internal mixer is standard laboratory
equipment that is often used to characterize materials processability and performance prior
to final manufacture in twin-screw extruders. The results in this work point out a difference
between a twin-screw extruder and an internal mixer that warrants further analysis.
105

Figure 5.10: Coefficients plot for the Delta-Lw property.

5.6 Conclusions

In this work, the influence of processing variables and unprocessed fiber length (UP-Lw)
on the final weight-average fiber length (Lw) and degree of fiber breakage (Delta-Lw) of
aspen fibers during composite processing with polypropylene in a twin-screw extruder was
studied. Two different batches of aspen fibers were used. One with longer fiber lengths
and a wider length distribution, AL. Another one with smaller fiber lengths and a narrower
length distribution, AS. The processing variables included the screws RPM, the polymer
and fiber feeders RPM and set-point temperature. The fiber concentration and the final
temperature of the system were also included.

It was found that the screws RPM and UP-Lw are negatively correlated to the final weight-
average length of the fibers. While positive correlation with Lw was found for the polymer
feeders RPM. Temperature variables showed negligible influence on Lw in this work.
106

The level of fiber concentration regime is shown to play an important role. Samples in
this work fall in both the semi-concentrated and concentrated category. These regimes
appear to determine the influence of the amount of fiber, given by fiber feeders RPM
and fiber concentration, on Lw. In the semi-concentrated regime fiberfiber contacts seem
to dominate fiber damage. On the other hand, a possible stronger floc formation in the
concentrated regime might protect fibers against breakage and more passes of the mixture
through the high stress zones would be required to break flocs apart. Increasing the amount
of polymer might ease the flow of fibers, therefore the positive correlation with Lw. Initial
fiber morphology, UP-Lw, had an effect on the final fiber length reached, pointing out a
difference between the experiments in twin-screw extruder and the analysis in an internal
mixer (Chapter 3)

In this work, the effect of processing conditions on fiber breakage were studied. It is an
empirical study and similar analysis with other fibers and polymers would be helpful to
ascertain the similarities and differences within other systems.
Chapter 6
Monte Carlo simulation of biofiber
breakage during processing in an internal
mixer

6.1 Abstract

Biofiber breakage during composite processing in an internal mixer with a non-Newtonian


fluid was carried out via a non-isothermal Monte Carlo simulation. Experimental data from
aspen fibers and wheat straw were used as inputs for the simulation along with rheological
data of polypropylene as the polymer matrix. The processing variables studied included
RPM (40 and 60), temperature (180 and 195 C) and fiber mass concentration (10 and
30 wt.%). Fiber breakage criteria considered only the effect of the fiberpolymer interaction.
Simulation results were compared to the experimental results of Chapter 3. Predicted
weight-average fiber lengths, Lw, and average fiber widths were close to experiments but
they tended to over-predict values for higher fiber concentration composites. Simulation
values of the fibers average aspect ratio were, in general, lower than experiments.


The contents of this chapter have been submitted for publication.

107
108

6.2 Introduction

Natural plant resources, such as wood or crops, are good sources of biofibers that can be
used as reinforcements to thermoplastics in composite materials [183]. Past decades have
seen a growing interest in the study and development of these types of composites for various
applications in the automotive, consumer goods, construction and building industries[184
186]. Natural fibers offer various advantages over synthetic fibers. They have good specific
strength and toughness, lower density, and the added benefit of being environmentally
friendly [187]. In addition, they are not as abrasive as their synthetic counterparts, resulting
in less wear to processing equipment [2].

Biofiber-thermoplastic composites are compounded using traditional thermoplastic process-


ing equipment, such as internal mixers and twin-screw extruders (TSE) [188]. However, a
side effect from the mixing is that biofibers are broken in length or width, thus changing the
initial aspect ratio [157159]. The aspect ratio of a fiber is its most characteristic geometric
property and is defined as the ratio of fiber length (L) to fiber diameter (D): L/D. Fibers
with higher aspect ratios strengthen a composite better because they carry the loads, trans-
ferred from the matrix, more efficiently. Also, there is a minimum (critical) fiber aspect
ratio that is required for effective fiber reinforcement. Critical fiber aspect ratios decrease
with higher fiber-matrix interfacial strength and lower fibers ultimate strength [15]. Fiber
length breakage decreases the aspect ratio and thus it is not a welcome outcome brought
about by the mixing when mechanical reinforcement is required.

The influence of processing conditions on biofiber breakage during mixing, in internal mixers
and TSE, has been apparent from the numerous works that have reported it for various
composite systems. Biofibers studied have included cellulose [108], pineapple-leaf fiber [91],
sisal [87], jute [89], hemp [90] and flax [189]. For example, studies in internal mixers have
shown that shorter final fiber lengths were observed at higher rotor speeds [87; 91] , increased
processing temperatures and longer mixing times [87].
109

Experimental studies on fiber breakage are a required step to understand the effect of the
different variables and would help composite manufacturers in their selection of processing
conditions. On the other hand, process optimization and equipment design strongly rely
upon theoretical work and simulation, as they decrease costs, waste of materials and save
time. Although fiber breakage during processing is a problem when length reduction occurs,
the field of its study via simulation is limited. Turkovich and Erwin [110] were the first to
summarize that fiber breakage could happen by any of three interactions: fiberpolymer,
fiberfiber or fiberequipment surfaces. All works that have developed fundamental rela-
tionships describing the conditions for polymer flow induced deformation or breakage of
fibers, have been carried out for dilute systems (fibers do not have contact with each other)
of synthetic fibers in simple shear flow. The first studies were carried out by Forgacs and
Mason [132; 190]. They derived an equation to calculate the shear stress () at which fiber
bending first occurs. Fiber bending is not necessarily a breakage criterion but the contin-
uance of bending could eventually reached fiber fracture. Their equation, which has been
the basis or the point to which most works compare, indicates that bending (and eventually
fracture) solely depends on the fibers aspect ratio (L/D) and Youngs modulus (E).

Using this concept, a simulation by Yamamoto and Matsuoka [134] obtained threshold
breakage curves of /E vs. L/D for fibers of various bending strengths. Their simulation
modeled fibers as bonded spheres that could stretch, bend and twist. Other works have
explored different numerical solutions [191; 192] to Yamamoto and Matuokas model. Nev-
ertheless, an interesting step to follow is the implementation of Yamamoto and Matsuokas
findings into simulations that model composite compounding to predict final fiber length.
In this work, one such simulation in an internal mixer is carried out.

Simulations of this type are lacking, especially in the field of biofiber composites. The
author only finding the interesting simulation on biofiber breakage in an internal mixer by
Iannace et al. [98]. To model fiber breakage, Iannace et al. used the equations developed by
Manas-Zloczower [193] for rupture of agglomerates of spherical particles, e.g., carbon black
clusters. Those equations considered the agglomerates as elongated bodies of revolution
that break under the tension imposed by the extensional components of simple shear flow.
110

An important conclusion from Manas-Zloczowers equations was that agglomerate rupture


is independent of agglomerate size but depends on the diameter of the spherical particles
that make up the agglomerate. However, a simulation model that involves fiber breakage
by fiber bending would be more pertinent, as this damage mechanism is considered more
relevant than breakage by tension in shear flows [194]. In addition, developed equations
for fiber breakage depend on the fibers aspect ratio, which is more in accordance to the
physical phenomenon. In the simulation by Iannace et al. a fiber always broke into four
equal-sized daughter fibers, each with length and diameter that were half the initial fibers.
Although in a real process a fiber could end up into several smaller fibers by the end of the
mixing process, fiber breakage, in length or width, at any given point in time, is more likely
to occur at a single plane rather than two. One part of Iannace et al. experiments consisted
in mixing their feed stock (sisal fibers and Mater Bi bioplastic) in an internal mixer under
extreme conditions to determine the minimum fiber length and width reached. Then, in
their simulation, they used these values as the lowest limits for fiber dimensions, so that
fibers would not continue breaking once they had reached either the length or width low
limit.

In this study, a Monte Carlo simulation was carried out on fiber breakage that followed a
different approach to the one by Iannace et al. by using the findings by Yamamoto and
Matsuoka [134]. No a priori minimum value of length was assumed. Instead, the criteria
under which a fiber would stop breaking was limited by properties of the flow and the
fibers: the fluid shear stresses, and fibers Young modulus and aspect ratio. The fibers were
assumed to break only in length. This work represents an extension to the experimental
study in the internal mixer carried out in Chapter 3. That work studied the breakage of
aspen fibers and wheat straw when mixed with a polypropylene matrix. All required input
information for the simulation, namely fibers morphology before processing and polymer
rheology, used the experimental data of the pre-processed fibers and polymer, respectively.
The simulation conditions included two RPMs (40 and 60), two fiber concentrations (10 and
30 wt.%) and two temperatures (180 and 195 C). Simulation results were compared to the
experimental results of Chapter 3. The results of the simulation showed good agreement
between predicted and the experimental final average lengths although values were closer
111

at lower fiber concentrations and lower RPMs. Change of the final average fiber width of
the composites happened as well due to the change in the distribution of widths every time
a fiber broke and their prediction was also close to experiments. Simulation values of the
fibers average aspect ratio were, in general, lower than found in the experiments.

6.3 Outline of the simulation

The internal mixer is standard processing equipment that is commonly used to compound
natural fiber-thermoplastic composites. Flow inside the mixer is very complex. However,
it has two characteristic mixing zones that have been widely acknowledged. Narrow gaps
formed between the rotors flight tips and the housing and larger areas between the rotors
[193]. Break up of agglomerates and fiber breakage happen predominantly in the narrow
gaps, due to the high stresses developed, while only distributive mixing is considered to
happen in the large zone [193].

Two types of flow are generated in the zone around the gaps. Elongational flows are
generated in the contraction leading to the gap and shear flow is generated inside the gap.
Both flows generate strong stresses, and as was shown in Chapter 4, fiber breakage happens
in both elongational and shear fields, and the share of fiber breakage that occurs in each field
is correlated to processing conditions. However, it has been a common assumption, in both
theoretical and simulation works on fiber breakage, that fiber damage happens exclusively
in the shear flow of the narrow gaps [98; 134]. The simulation in this work followed this
approach. Also, the analysis assumes a non-isothermal, Non-Newtonian flow and that the
suspension is in a dilute concentration regime. In dilute fiber suspensions, fibers do not
interact with one another and the fiberpolymer interaction is considered to be the leading
cause for fiber breakage. Accordingly, the effects of the fibers interaction with equipment
surfaces were considered negligible. The assumptions mentioned above that were followed in
the simulation in this study fulfill the assumptions under with the model for fiber breakage
was developed by Yamamoto and Matsuoka [134].
112

It must be pointed out that even though the simplification of the system is high, the
hydrodynamic forces on the fiber, in shear flow, are considered a major interaction that
causes fiber breakage [110; 121] as well as an important interaction that influences the
rheology of the composite during processing [135]. Bsased on observations on glass fiber
polymer processing, Turkovich and Erwin [110] indicated that fiber breakage, even for fiber
concentrated regimes, can be described solely by a dilute suspension mechanism.

6.4 Equipment

The internal batch mixer considered in this work is a CWBrabender 3-piece mixer/mea-
suring head with non-intermeshing, counter-rotating roller type blades as shown in Figure
6.1. These blades are of irregular shape and subject the samples to control conditions
of temperature, shear and time during mixing. The shear is controlled by the level of
revolutions-per-minute (RPM) of the rotors.

Figure 6.1: Internal batch mixer with roller type blades.


113

The mixer consists of two interconnected chambers, each one with a rotor. The rotor of
one of the chambers is connected to the shaft of the motor. The variable RPM refers to
the rotations of the shaft and the blade connected to it. The other blade rotates two thirds
of the shaft. The simulation in this work focused only on the small gaps, where intensive
mixing and high stresses are developed [58]. Each rotor has three gaps through which the
mixture passes recurrently during processing. In one rotation of the rotor, each gap runs
through 80 % of each chamber, while the other 20 % is the space where the chambers are
interconnected. The total volume of the high shear zone (HSZT ) created by all gaps is:

HSZT = HSZa + HSZb = 76 cm3 (6.1)

HSZa = 0.8 Rc2 Rr2 L 3 = 45.6 cm3



(6.2)

2
0.8 Rc2 Rr2 L 3 = 30.4 cm3

HSZb = (6.3)
3

where HSZa and HSZb are the high shear zone volumes created by the three gaps of the
rotor attached to the shaft and the secondary rotor, respectively. The chamber internal
radius, Rc , is 3.5 cm, the radius of the rotor at the wing tip, Rr , is 3.43 cm and the length
of the gaps, L is 10 cm. Due to the irregular shape of the blades, all measurements are
approximates taken from of an internal mixer with a capacity of 260 cm3 . For optimum
performance the mixer is filled between 7080 v.% of its capacity [42].

6.5 Materials

6.5.1 Fibers

The simulation used the experimental data of length and width from thermomechanically
extracted biofibers as input. Fibers experimental information corresponding to their ex-
114

traction and morphological measurement are reported in Chapter 3. The biofibers consisted
of two batches of hardwood aspen fibers and one batch of non-wood wheat straw (WS).
The batches of aspen fiber are referred to as aspen long, AL, and aspen short, AS. The
length-weighted average length, Lw, and average width, Wn, of the batches were: AL:
Lw=1.7 mm Wn=48.1 m; AS: Lw=1.5 mm Wn=35.6 m; WS: Lw=2.3 mm Wn=63.8 m.
Fibers length and width were measured using a HiRes FQA (Optest Equipment Inc,
Canada). Three replicates for each batch of fiber gave a pool of 35,458 AL fibers, 64,772
AS fibers and 12,856 WS fibers.

During experimental processing of polymers and polymeric suspensions, the temperature


increases due to viscous heat dissipation [42; 195]. If the temperature increases beyond the
onset of biofiber thermal degradation, the fibers mechanical integrity would be compromised
and higher fiber breakage would be expected. To take into account this effect, thermal
degradation characteristics of the fibers were obtained using a TGA500 (TA Instruments,
USA) under air atmosphere with a heating rate of 10 C/min covering a range from 50 to
400 C. TGA thermograms show weight percent against temperature (Figure 6.3). Major
weight loss for both types of fibers started around 195 C.

Also, experimental values of the Youngs modulus of the fibers were estimated by carrying
out tensile tests on the composite samples processed in the internal mixer (Chapter 3) and
assuming the composites follow the upper-bound of the rule-of-mixtures for the composite
modulus, according to [15]:

Ec = Ef Vf + Ep Vp (6.4)

where E and V represent the Youngs modulus and volume concentration of the composite,
c; fiber, f ; and polymer, p. The tensile tests were measured on ten specimens for each
sample according to the ASTM D-638 standard at a crosshead speed of 2.5 mm/min. The
average of the fibers Youngs modulus were: AL: E = 5.2 1.4 GPa; AS: E = 4.0 0.8 GPa
and WS: E = 3.6 1.2 GPa.
115

Figure 6.2: Thermogram for aspen and wheat straw fibers.

6.5.2 Polymer

The rheological data of the polypropylene (PP) Profax 6523 (Lyondellbasell, the Nether-
lands) was used for the polymer matrix. This polymer has a nominal melt flow rate of
4 g/10 min (230 C/2.16). The peak melting point as determined using a DSC1000 (TA
Instruments, USA) was 160 C. Polymer fluid viscosity () is affected by the change in
both shear rate ()
and temperature (T ). The polypropylene viscosity was measured with a
twin-bore bench-top Rosand capillary rheometer series RH2000 (Malvern Instruments, UK)
at four temperatures (180, 190, 200 and 210 C) and ten piston speeds ranging from 0.67
to 16.66 mm/min, which correlated to shear rates between 30 and 730 s1 . The rheological
data were used to develop a master curve for the shear rate-temperature dependence of
viscosity. The Ostwald-de Waele or Power-law equation was used to account for the effect
of shear rate [55]. The power-law is an inelastic generalized Newtonian viscosity equation.
Polymer elastic effects are considered negligible [196]. The temperature effect was included
through the shift factor of an Arrhenius type dependence [55]. The master curve was:
116

  
n1 E 1 1
= aT m0 ; Shift factor: aT = exp (6.5)
R T T0

where m0 is the consistency index at the reference temperature T0 , n is the power-law


index, aT is the shift factor, E is the activation energy, R is the universal gas constant. The
parameters used for the power-law and shift factor equations are given in Table 6.1.

Figure 6.3: Master curve for the shear rate-temperature dependence of viscosity.

Table 6.1: Constants for the power-law and shift factor equations.
m0 n E R T0
Pasn J/mol J/(mol K) K
14,704 0.34 16,734 8.314 473.15
117

6.6 Simulation procedure

Fiber breakage during processing for every suspension of biofiber (AL, AS, WS) with
polypropylene was simulated according to the design in Table 6.2, which included three
processing variables (fiber mass concentration, RPM and temperature) at two levels each.
Replicates were obtained by carrying out ten repetitions of every simulation.

6.6.1 fiber concentration

The simulation used a 80 v.% filled capacity (208 cm3 ) for the internal mixer [42]. Fiber
volume fractions for the 10 and 30 wt.% fiber composites were = 0.06 and = 0.19,
according to:

Vf Mf p Mf Mp
= = ; Vf = ; Vp = (6.6)
Vp + Vf Mp f + Mf p f p

where Mf and Mp are the fiber and polymer mass concentration respectively. It was assumed
a melt polypropylene density, p , of 0.8 g/cm3 and a fiber density, f , of 1.5 g/cm3 . During
simulation the volume of individual fibers was computed using their length and width,
assuming they were perfect cylinders where the measured widths were the fiber diameters
[43; 98].

Every simulation started by first selecting the fiber type, AL, AS or WS, and then selecting
one set of processing conditions from Table 6.2. Then a sample (variable name fibers-in-
suspension) was created by adding fibers, one at a time until the fiber volume fraction was
reached. Each fiber was chosen randomly, with the possibility of repetition, from the pool
of experimental data. Every simulation had a different combination of fibers. Afterwards, a
value of Youngs modulus was assigned to every fiber in the sample. The elastic modulus for
each fiber was chosen randomly between the minimum and maximum values given by the
average and standard deviation of the experimentally estimated fibers Youngs modulus.
118

Table 6.2: Simulation design for each fiber type.


Fiber concentration RPM Temperature Final temperature
wt.% C C

AL AS WS
10 40 180 190 190 190
10 60 180 194 193 192
30 40 180 195 196 194
30 60 180 215 218 217
10 40 195 206 203 205
10 60 195 209 206 209
30 40 195 206 230 212
30 60 195 217 234 215

6.6.2 RPM

The mixing time for all samples was 6 min and was measured by the number of total
revolutions (variable name total-revolutions) according to:

total-revolutions = 6 RPM (6.7)

The mixing time processing loop variable was called revolution. revolution went from 1 to
total-revolutions in increments of 1.

6.6.3 Temperature

Processing temperature increases during experimental processes due to viscous heat dissipa-
tion [42]. In the simulation the variable temperature was set to increase from the set-point
temperature to the real final temperatures (see Table 6.2) reached in the experimental work
on the internal mixer (Chapter 3). The temperature increment was:
119

Final-temperature temperature
Temperature-increment = (6.8)
total-revolutions 1

6.6.4 Simulation processing

After the processing conditions and the sample had been set, the simulation processing loop
started. At every value of revolution the new temperature of the loop was calculated and
the simulation carried out the following:

Fibers that enter the high shear zone. A set of fibers that enter the high shear zone
(fibers-in-HSZ ) was chosen from fibers-in-suspension. fibers-in-HSZ had the same fiber
volume fraction as the sample. Fibers in the high shear zone were added one at a time
until the fiber volume fraction was reached. Each fiber was chosen randomly. The variable
fibers-in-HSZ was reset at every new value of revolution.

Fibers Youngs modulus during simulation. The elastic modulus of the fibers was
updated in every loop due to the increasing value of temperature. The effect of temperature
on the mechanical properties of solid wood has been well studied [197]. However, the use of
the models that describe the behavior of the Youngs modulus of wood with temperature
for the biofibers in this work would not be appropriate. This is due to the difference in size
scale between a piece of wood and a biofiber. Wood is a composite material made up of a
large number of biofibers, whose arrangement, orientation and composition varies between
wood pieces. The thermal conductivity of wood is low, and approximately about two to
four times that of common insulating materials [197]. This causes that thermal equilibrium
of a piece of wood with its environment happens over a period of time and the strength
value of the inner parts will be different than that for the outer parts. The shape and size of
a wood piece are considered important when analyzing the influence of temperature [197].

On the other hand, biofibers dimensions are small enough such that their thermal equi-
librium with the environment would happen at a much faster rate. Unfortunately, there is
only limited data available for the behavior of the biofibers elastic modulus at high tem-
120

peratures. Studies on the elastic modulus of isolated sisal fibers at 30, 80 and 100 C [198]
and wheat and flax straw at 25, 100 and 200 C [199] have been carried out. However, these
data only provide one point (at 200 C) that would be adequate for the temperatures used
in the simulation, as final temperature of the composites reached higher values in various
cases.

The effect of temperature on the biofibers Youngs modulus is one of inverse proportional-
ity. The elastic modulus decreases drastically at high temperature and this was explained
in terms of fiber weakening due to: a) differences in the thermal expansion of individual
components that would create stress concentrations or exacerbate fibers flaws [198] and b)
thermal degradation of the fiber components [199]. Due to the relationship between tem-
perature and the elastic modulus, as well as the consideration that biofibers reach thermal
equilibrium very fast, it would be reasonable to assume that the decrease of a biofibers
Youngs modulus happens at the same rate as their thermal degradation due to temper-
ature increase. A TGA thermogram shows fibers weight loss, due to fiber degradation,
against temperature. To account, in the simulation, for a continuous fiber weakening with
temperature, it was then assumed that the fibers Youngs modulus decreased at the same
rate as the percentage of fibers weight loss in the TGA thermogram. Prior to fiber thermal
degradation, the decrease was small and once degradation started a sharp decrease was
experienced.

Shear stress in the high shear zone. The shear stress, , in the drag flow of the high
shear zone was calculated as follows [59]:

2Rr RP M
= h ; h = (6.9)
60H

where h and H are the shear rate and height of the gap, respectively.

Breakage criteria. Every fiber in fibers-in-HSZ was analyzed according to the breakage
criteria to determine whether it was going to break or not. The breakage criteria used herein
followed the one published by Yamamoto and Matsuoka [134]. Yamamoto and Matsuoka
121

gave a fiber fracture condition using the ratio of the polymeric fluid shear stress, p , to
the fibers Youngs modulus, Ef : p /Ef . Their simulation generated threshold curves of
t /Et as a function of fiber aspect ratio, L/D, for fibers with a bending strength of 0.01Ef ,
0.1Ef , 0.2Ef and 0.3Ef . At any moment, for a given fiber aspect ratio, fiber breakage in
the simulation occurred if the following condition was satisfied:

p t
> (6.10)
Ef Et

We assumed that fibers were homogeneous and so their flexural strength was identical
to their tensile strength [200]. Experimental measurements of the tensile strength and
Youngs modulus of individual specimens of wheat straw published by Hornsby et al. [199]
and Panthapulakkal et al. [201], indicated that wheat straws tensile strength is between
0.01Ef to 0.02Ef . This range was used as a reference for all biofibers herein. However, the
threshold curve of t /Et vs L/D for fibers of bending strength of 0.01Ef does not cover
the full range of fiber aspect ratios. It was then decided to use the next threshold curve,
for a fiber of bending strength of 0.1Ef , which covers a larger range of fibers aspect ratio.
Yamamoto and Matsuoka [134] presented the threshold curves in a graphical form. The
data points for the 0.1Ef curve were extracted from the graph and fitted to the following
equation:

   
t L
log = 0.46 3.559 log (6.11)
Et D

Equation 6.11 is a decreasing curve, where fibers with longer aspect ratios have lower values
of t /Et .

Types of breakage. A fiber broke whenever it satisfied the condition of Equation 6.10.
Due to biofibers are non-homogeneous, with defects in their structures [201], fiber breakage
could happen anywhere along their length. Three different simulations, each one considering
one type of breakage was studied:
122

1. The fibers broke in half,

2. The fibers broke into two fibers of 1/3 and 2/3 of the initial fiber length,

3. Random breakage along the fiber length occurred.

The diameter of the fibers in the simulation remained constant. The variable fibers-in-
suspension was then updated by replacing the data of the two new fibers that were generated
for the information of the fiber that broke.

6.7 Technical description of the simulation

The simulation carried out is of a Monte Carlo type. The idea behind a Monte Carlo
simulation is that a complex process can be studied through the repetitive evaluation of
a large number of random samples, where these samples are derived from a distribution
of data (called a pseudo-population) that is representative of the true process population
[202].

The simulation was carried out using Matlab version 7.5.0.338 (R2007b) by MathWorks.
Matlab is a numerical calculation tool with programming capabilities. It provides a high-
level programming language. Matlab is short for matrix laboratory as its basic data elements
are matrices. It is has an extensive library of routines that greatly expand its functionality
for numerical calculations and data plotting with an large network of support group. Details
on Matlab can be found elsewhere [203].

A central part of a Monte Carlo simulation is the randomization of the sampling. This
study used the Matlab random function, RAND, which generates uniformly distributed
pseudo-random numbers. Every time a random number was required, the RAND subroutine
generated a new number.

The dimensions of length and width of the un-processed fibers as measured by the HiRes
FQA were used as the pseudo-population. When selected as an option, the HiRes FQA
123

provides a text file with the length and width of every single fiber that was measured
during a test. The range limits reported by the HiRes FQA are the following. Fiber lengths
reported were between 0.04 and 10 mm. Fiber widths reported were between 7 and 300 m.
Fibers that had a length longer that 0.04 mm but a width smaller than 7 m were reported
in the text file as having zero width. The aspect ratio of the fibers that show zero as their
width cannot be calculated. Due to the breakage criteria in the simulation depended on
a fibers aspect ratio (see section 6.6.4), these fibers were not included in the simulation.
The pool of unprocessed fibers for AL, AS and WS indicated in section 6.5.1 are the actual
number of fibers used for the simulation. The same range limits for fiber length and fiber
width reported above are the ones that the HiRes FQA used for the calculation of the
averages. Therefore the equipment did not include in the calculation of average widths
the widths reported as zero. To properly compare the results of the simulation with those
obtained with the HiRes FQA, the calculation of averages abided to the above ranges for
length and width. The information of length and width read for each fiber was store in
three matrices, each one corresponding to one type of fiber: AL, AS or WS. These matrices,
which were referred to as the fibers-to-mix matrices, provided the population of fibers used
in the simulation. After the fibers-to-mix matrices were created a third column was added.
The third column corresponded to the fibers aspect ratio, which was calculated using the
fibers length and fibers width. The fibers width was considered to be the fibers diameter.

The files of thermogravimetrical (TGA) data (see section 6.5.1) for each fiber (aspen or
wheat straw) were in text format and had the information of the initial weight of the
sample as well as the temperature and weight recorded during the tests. A variable (initial-
weight) equalled to the initial weight was defined for each fiber. Also two matrices were
created, one for aspen and one for wheat straw. Each matrix (TGA-variable) contained the
temperature and weight data as read from the TGA files.

The experimentally estimated average and confidence intervals of the Youngs modulus for
AL, AS and WS fibers (see section 6.5.1) were used to define two variables for each fiber
type. One was the lower-level of the interval, the other was the confidence-interval.
124

Polymer flow properties were also estimated experimentally and fitted to Equation 6.5
(section 6.5.2). The values of the shear stresses at any point during the simulation were
calculated using Equations 6.5 and 6.9, and Table 6.1.

The processing conditions used are reported in Table 6.2. Three matrices were created,
each for one fiber type (AL, AS, WS). Each matrix had four columns corresponding to:
fiber mass concentration, RPM, Temperature and Final temperature. The first three values
corresponded to the processing conditions, which are the same ones used in the analysis on
the internal mixer. The last column corresponded to the experimental final temperature
recorded by the equipment after processing was finished.

All previous data were the inputs for the simulation. In this work, the Monte Carlo simula-
tion was carried out on the occurrence of fiber breakage during compounding in an internal
mixer and the stochastic part was given by the random sampling carried out on the pop-
ulation of fibers. On the other hand, all processing conditions were carried out as nested
loops. At the onset of simulation, for-type nested loops established all the processing
conditions, according to:

Fiber selection The for loop selected AL, AS or WS in consecutive order. After a fiber
type was selected, the script described below was executed and after it had finished,
the next fiber type was selected.

Type of breakage The for loop selected the breakage type: in half, one-third and ran-
dom. After a breakage type was selected the script described below executed and after
it had finished, the next type of breakage was selected. During in half breakage,
every time a fiber broke, broke into 2 parts, each one half the length of the original.
During one-third breakage, every time a fiber broke, broke into 2 fibers, one fiber
was 1/3 of the original length, the other fiber was 2/3 of the original length. During
random breakage, every time a fiber broke, the RAND command generated a num-
ber between 0 and 1. One fiber was RAND Initial-length, the other one was the
remaining length. The value of RAND was different for every fiber in every occurrence
of breakage.
125

Temperature The for loop selected the temperatures 180 C or 195 C. After the first
temperature was selected the script described below was executed and after it had
finished the second temperature was selected.

RPM The for loop selected the mixing speed 40 or 60. After the first RPM was selected
the script described below executed and after it had finished the second RPM was
selected.

Fiber mass concentration The for loop selected the fiber mass concentration 10 or
30. After the first fiber mass concentration was selected the script described below
executed and after it had finished the second fiber mass concentration was selected.

Simulation of fiber breakage during processing At this point all processing condi-
tions had been established and the processing started. The temperature-increment
was calculated using Equation 6.8. Also, the volume of fibers (fiber-volume) that
corresponded to the fiber mass concentration and the suspension volume of 208 cm3
(80 v.% filled capacity of the internal mixer) were calculated using Equation 6.6.

1. Creation of the fibers-in-suspension matrix. The fibers that formed part of the
suspension that was processed under a set of processing conditions as estab-
lished by the for type loops were called fibers-in-suspension. Fibers in the
fibers-in-suspension matrix were chosen from the pool of data provided by the
fibers-to-mix matrix corresponding to the fiber type (AL, AS or WS). According
to a random number generated by the RAND subroutine, a fiber corresponding
to the random number was selected from the fibers-to-mix matrix and the fiber
information was copied into the fibers-in-suspension matrix. This process was
carried out, one fiber at a time, until the volume of fibers reached fiber-volume.
The volume of each fiber was also added to the fibers-in-suspension matrix. All
fibers in the fibers-to-mix matrix were able to be chosen more than once as they
had the same probability of been selected every time a fiber was required and
the random number selected. The fibers-in-suspension matrices were created at
this point of the simulation. Therefore, for every set of processing conditions a
new fibers-in-suspension matrix was created.
126

Additional information for every fiber was calculated and added at this point to
the fibers-in-suspension matrix. The information included:

Initial fibers Young modulus (EI ). For every fiber, a new random num-
ber was generated by the RAND subroutine. Using the random number a
Youngs modulus was assigned to the fiber according to:
EI = RAND confidence-interval + lower-level . Therefore, every fiber had
a different, randomly selected Youngs modulus every time the fibers-in-
suspension matrix was created.

Critical shear stress-to-fibers Youngs modulus ratio. Using Equation 6.11


[134]) the critical shear stress-to-fibers Youngs modulus ratio was calculated

2. Processing and fiber breakage. The volume of fibers that entered the high shear
zone (HSZ-fiber-volume) was calculated using Equations 6.1 to 6.3 and 6.6. The
mixing time of the suspension is simulated through the variable revolution, which
is given as a for loop (section 6.6.2). revolution takes values from 1 to total-
revolution in steps of 1. The closing of the loop revolution indicated that the
processing ended. After the revolution loop closed, all data was saved into a new
Matlab type file. Inside the revolution loop the script carried out the following:

Current temperature. The simulated temperature went from the set-point


to the final temperatures as shown in Table 6.2. At every new value of
revolution the current-temperature was calculated using Equation 6.8.

Polymer shear stress in the HSZ. The polymer shear stress in the high shear
zone (HSZ) was updated using Equations 6.5 and 6.9, and Table 6.1.

Current fibers Youngs modulus (EC ). The fibers Youngs modulus de-
pended on the current temperature and was updated at every new value of
revolution. An interpolation of the temperature-weight data in the TGA-
variable matrix was carried out that calculated the weight (new-weight) of
the sample that corresponded to the current temperature. Each current
fibers Youngs modulus was updated in the fibers-in-suspension matrix ac-
cording to: EC = EI new-weight/initial-weight.
127

A value of the current shear stress-to-current fibers Youngs modulus ratio


for each fiber was calculated.
fibers-in-HSZ matrix. According to a random number generated by the
RAND subroutine, a fiber corresponding to the random number was selected
from the fibers-in-suspension matrix and the fiber data was copied in the
fibers-in-HSZ matrix. This process was carried out, one fiber at a time,
until the volume of fibers reached HSZ-fiber-volume. Due to the fibers in
the fibers-in-suspension matrix are the fibers under processing, they can be
selected only once to enter the high shear zone at every value of revolution.
With every new value of revolution, all fibers in the fibers-in-suspension
matrix had the same probability to enter the high shear zone again.
Every fiber in the fibers-in-HSZ matrix was evaluated according to the break-
age criteria of Equation 6.10 to determine if it broke. If a fiber broke, two
new fibers were created whose length was determined by the type of breakage
under evaluation. One daughter fiber length information replaced the initial
fibers length in the fibers-in-suspension matrix and the second fiber length
was stored in a new row in the fibers-in-suspension matrix.
After fiber breakage ended, the fibers-in-HSZ matrix was cleared before the
new revolution started.
The data for the two new fibers was copied or updated to the fibers-in-
suspension matrix: fibers width, aspect ratio, fibers volume, critical shear
stress-to-fibers Youngs modulus ratio. The initial fibers Youngs modulus
of the two daughter fibers was assumed to be the same as the initial fibers
Youngs modulus of the initial fiber. If the new daughter fibers entered
the high shear zone the decay in modulus was calculated according to the
increased temperature as mentioned above.

This code represented on replicate. The code was run ten times to represent ten replicates.
This code is the one that carried out the simulation on fiber breakage and saving of the
data into Matlab type files. A section of this script is presented in Appendix D. Another
script was coded that extracted the data, calculated the averages and plotted the graphs.
128

6.8 Results and discussion

Simulations of biofiber breakage during processing in an internal mixer rendered data ma-
trices (fibers-in-suspension) with information on the final length, width and aspect ratio for
every fiber. Prediction results are compared to experimental data for the same type of fiber
composites that were processed in an internal mixer under identical processing conditions
as the simulations (Chapter 3). A typical normalized cumulative curve of the distribution
for fiber length is shown in Figure 6.4. The simulation corresponds to 10 wt.% AL fiber
composites processed at an RPM of 60 and temperature of 180 C for fibers that break in
half. Simulations predicted distributions of fiber lengths that were in good agreement with
experimental data. Fibers had a significant decrease in length during simulation.

Length distributions of biofiber composites are commonly characterized and compared


through their length-weighted average length, Lw, because it gives more weight to longer
fibers and reduces the effect of fines on the length values [119; 145]:

X
ni L2i
Lw = X (6.12)
ni Li

r
s2 ( Li )2 N0 ni Li (Li Lw)2
P P
sLw = ; e= P ; s2 = (6.13)
(Li )2 (N 0 1) Li
P
e

where i = 1, 2, M categories (bins in the distribution); n is the fiber count in the ith
category, L is the histogram class center length in the ith category, N 0 is the number of
non-zero weights, s is the weighted standard deviation of the sample and sLw is the standard
error of the weighted average [120]. In this work Lw is also referred to as the weight-average
fiber length.

Two RPMs were studied, 40 and 60, that give two shear rates, 205 and 307 s1 , respec-
tively. These shear rates fell into the power-law region of the viscosity-shear rate curve
used. Simulation temperatures increased from the set points, 180 and 195 C, to the fi-
nal temperatures that each batch reached in the experimental processing (see Table 6.2).
129

Figure 6.4: Normalized cumulative distribution of the predicted final weight-average fiber
length, Lw, for the simulation of 10 wt.% AL fiber composites at 60 RPM, 180 C and with
fibers broken in half.

Comparison between different simulations showed that predicted values of Lw varied with
RPM and temperature. Though the variation was small, around 3 percent, and in several
cases it was not statistically significant. In general, predicted Lw decreased with higher
RPMs and increased with the increase in temperature. The decrease of Lw with RPM is
consistent with experiments as higher RPMs increase the shear stresses of the system. The
increase of Lw with temperature in the simulation was contrary to experimental results. In
the experiments, the decrease of Lw with temperature was explained in terms of the fiber
weakening as processing temperatures reached the point of the biofibers thermal degrada-
tion (Chapter 3). The discrepancy in trend between simulation and experiments could be
due to the fact that the stresses in the suspensions would be higher than the ones predicted
as the resistance to flow (viscosity) and stresses increase with the amount of fibers in a
polymeric suspension [135; 166].
130

The comparison of the weight-average length, Lw, between the simulations and the ex-
perimental results from the internal mixer of Chapter 3 is shown in Figures 6.5 to 6.7.
The Monte Carlo simulation included changes in RPM, temperature and fiber concentra-
tion. Every fiber was treated as an isolated entity within the polymer matrix, where fiber
breakage depended solely on the effect of the fiberpolymer interaction. The simplicity
of the simulation did not prevent it, however, from getting results that are close to the
experimental values. Results for AS fibers were the closest, followed by AL fiber results.

If we consider that the stresses in a system increase with higher RPMs, higher fiber concen-
trations and longer fiber lengths [166]; it would be reasonable to assume that the stresses
in the real suspensions, from lower to higher, were: 10 wt.% 40 RPM < 10 wt.% 60 RPM <
30 wt.% 40 RPM < 30 wt.% 60 RPM. And comparison of stresses between the composites:
AS < AL < WS.

It can be noticed that the simulations had the tendency to over-predict Lw for suspensions
that had the highest stresses. This is especially noticeable for AL and WS fiber composites.
The over prediction had a tendency to increase with the increase of the stresses in the sys-
tem. For example, AL fiber composites over-prediction shows: 10 wt.% 40 RPM < 10 wt.%
60 RPM < 30 wt.% 40 RPM < 30 wt.% 60 RPM. The main reason for this behavior could be
due to the simulation did not include the effect of the interaction among fibers. Fiberfiber
interaction, which increases with fiber concentration, is a factor considered significant to the
occurrence of fiber breakage [110], and its effect is important in the semi-concentrated and
concentrate levels of fiber concentration. The criteria for suspensions in the concentrated
regime is [176]:

 
D
> (6.14)
L

The right side of Equation 6.14 is the inverse of the fiber aspect ratio. Using the aspect
ratio for fibers after simulation as well as for unprocessed fibers (Chapter 3), D/L values
ranged from 0.04 to 0.065. The fiber volume fraction for 10 and 30 wt.% fiber suspensions
were 0.06 and 0.19, respectively (Equation 6.6). Therefore, suspensions studied in this
131

work essentially fall into the concentrated regime. At these levels of fiber concentration,
the influence of the fiberfiber interaction on the extent of breakage would be significant.
Larger fiberfiber interactions and thus higher stresses would be expected by the increase of
RPMs and fiber concentration, which would be conducive to lower values of Lw, as observed
in the experiments.

Regarding the type of breakage (in half, one third, random), it seems that the point of
rupture where breakage occurs does not influence the final weight-average fiber length.

Figure 6.5: Comparison between prediction and experimental data of the final weight-
average fiber length, Lw, for AL fiber composites.
132

Figure 6.6: Comparison between prediction and experimental data of the final weight-
average fiber length, Lw, for AS fiber composites.
133

Figure 6.7: Comparison between prediction and experimental data of the final weight-
average fiber length, Lw, for WS fiber composites.
134

The simulation carried out in this work only considered breakage of fiber length but not
reduction of fiber width. However, as fibers broke, a change in the distribution of widths
occurred because the new reduced fibers were added into the system, changing the average
fiber width. Figures 6.8 to 6.10 show the final average fiber widths, Wn, for all the com-
posites. The agreement between the simulation and the experimental data is good and in
the case of AL and WS fiber composites, agreement seems to be better than for Lw. The
highest discrepancy between simulation and experiments was found for the composites: a)
AL10 wt.% at 40 RPM and 180 C; b) WS10 wt.% at 40 RPM both temperatures.

Simulation Wn values under-predicted experiments with low fiber concentration that were
processed at the low RPMs, and over-predicted the values of Wn of suspensions of 30 wt.%
fiber concentration. This behavior could be the result of a balance between two factors.
One factor would be the breakage condition in the simulation that made it easier for fibers
with higher aspect ratios to undergo length reduction. For the biofibers used in this work,
longer aspect ratios were easier to be found in thinner fibers. Therefore, the pool of smaller
fibers widths would have increased faster as fibers broke, bringing down the value of Wn and
giving results lower than experimental, as in 10 wt.% and 40 RPM composites. On the other
hand, fiberfiber interaction is also a viable factor for fiber width reduction and at 30 wt.%
fiber composites it would be the reason for simulations over-predicting experimental data.

The closeness between the simulation results and experiments on the final average width are
interesting because they could seem to indicate that no reduction of width occurs during
processing. However, reduction of biofiber width during processing is relevant because
breakage of fibers widthwise, through separation of the fiber bundles, has been widely
reported [88; 98]. A way to elucidate the importance of including fiber width breakage in
future simulations is by comparing the aspect ratio of the fibers.
135

Figure 6.8: Comparison between prediction and experimental data of the final average
fiber width, Wn, for AL fiber composites.
136

Figure 6.9: Comparison between prediction and experimental data of the final average
fiber width, Wn, for AS fiber composites.
137

Figure 6.10: Comparison between prediction and experimental data of the final average
fiber width, Wn, for WS fiber composites.
138

Figures 6.11 to 6.13 compare the final average aspect ratio of the fibers between the simula-
tion and experimental results. A discrepancy between them is found, most of the simulation
results were lower than experiments. During simulation, longer aspect ratio fibers were bro-
ken and the aspect ratio of the new fibers was always smaller than the original one because
fiber widths did not change. A closer agreement between simulation and experiments for
the fibers aspect ratio would have been obtained if fiber width breakage had been included
in the simulation. Predicted average aspect ratios for AL fiber composites were an excep-
tion as they showed a closer agreement between simulation and experiments. AL fibers
morphology before processing was between the ones of WS and AS fibers (Chapter 3) and
the play between the different factors that influence fiber breakage would have been the
result of the closer agreement found for AL fiber composites.

Figure 6.11: Comparison between prediction and experimental data of the final average
fiber aspect ratio, L/Dn, for AL fiber composites.
139

Figure 6.12: Comparison between prediction and experimental data of the final average
fiber aspect ratio, L/Dn, for AS fiber composites.
140

Figure 6.13: Comparison between prediction and experimental data of the final average
fiber aspect ratio, L/Dn, for WS fiber composites.
141

6.9 Conclusions

A Monte Carlo simulation was carried out to determine the level of biofiber breakage during
composite processing in an internal batch mixer. Two different batches of aspen, named
aspen long, AL, and aspen short, AS, fiber; and one batch of wheat straw were analyzed.
Polypropylene was used as the matrix for the composites. All input corresponding for the
fibers and polymer used experimental data. Simulation results were compared to experi-
mental results reported in Chapter 3. Simulation conditions included two RPMs (40 and
60), two fiber concentrations (10 and 30 wt.%) and two set-point temperatures (180 and
195 C). The simulation took into account the increase in temperature during composite
processing by including the final temperatures reached during the experimental process-
ing. The simulation assumed that fiber breakage occurred solely due to the effect of the
fiberpolymer interaction.

Simulation results for the weight-average fiber length, Lw, had, in general, good agreement
with experimental values. Prediction of Lw was better for AS fiber composites and it was
closer at lower fiber concentrations and lower RPMs. As these parameters increased the
predicted Lw values rose higher than experiments. The latter could the due to the exclusion
of the fiberfiber interaction in the simulation. This exclusion could also be the reason for
the higher predicted final average fiber width, Wn, compared to the experiments for higher
fiber concentration composites. In general the predicted Wn were also close to experimental
results. Fiber widths were not changed during the simulation but the final average fiber
width changed as new fibers were added into the system every time a fiber broke. Predicted
average fiber aspect ratios always decreased because fiber width was not changed during
simulation. However, the predicted AL fiber aspect ratios were close to experimental values
possibly due to result of the balance between all factors that influence fiber breakage.

The simulation assumptions are an oversimplification of the real conditions. However, the
complexity of real problems is many times better tackled by gradual steps. The simulation
herein represents a good basis to more complex works. The fiber-to-fiber interaction along
with a type of fiber width reduction would be interesting to analyze in further simulations.
Chapter 7
Conclusions and Recommendations

7.1 Concluding Remarks

Natural fiber breakage during composite processing with thermoplastic polymers is an un-
desired outcome if length breakage results in poorer reinforcement of the polymeric matrix.
In this study, some already known results were confirmed by experiments and analysis. The
study has also led to new interesting contributions to our understanding of this field of
research.

In this study, composites of thermomechanical fibers or aspen or wheat straw mixed with
a polypropylene matrix were used. It was shown that the dimensions of biofibers prior
to processing are important factors that can influence the level of fiber breakage. Their
importance on the final fiber width was demonstrated in the internal mixer study, while their
significance on final fiber length was more relevant during twin-screw extrusion experiments.

Another contribution to our understanding of biofiber breakage was related to the use of the
capillary rheometer to elucidate whether breakage occurred in elongational flow. Conver-
gent flows from large reservoirs into small openings, like that in a capillary rheometer, are
of great importance for dispersive mixing in processing equipment. They are found in the

142
143

flow before small gap regions between the screws and the barrel. Gap openings of approx-
imately 1 mm, as the capillary die used in this work, are common in composite processing
equipment in general. The flow of the suspensions through the capillary rheometer, while
did not give smooth pressure drops like those of pristine polymeric resins, would be more in
accordance to the flow of real suspensions through the small gaps in processing equipment
and this similarity gave more relevance of the results on fiber breakage obtained herein to
real situation in composite processing.

In this work, it was also shown that analyzing variables together by the multivariante
statistical method of projection to latent structures (PLS) allowed a better understanding on
how all factors (processing conditions, fiber initial morphology and processing equipment),
interact with one another to provide the conditions under which fiber breakage occurs.
For example, as the fiber volume concentration increased from the semi-concentrated to
the concentrated regime the effect on fiber breakage of the level of fiber in the suspension
changed in the twin-screw extruder (Chapter 5).

The simulation developed in this work had two key contributions: 1) a breakage model
developed for synthetic fibers showed to be helpful for the biofibers studied herein. 2)
10 wt.% biofiberpolypropylene suspensions fell within the concentrated regime, and yet, a
model developed for dilute suspensions reasonably predicted biofibers final weight-average
length. Turkovich and Erwin [110] were the first to reach the conclusion that concentrated
suspensions of synthetic fibers could be modelled via a dilute suspension mechanism, but
this had not been shown for natural fibers.

Other specific contributions are:

Elongational flow. It was demonstrated that breakage of natural fibers also happens
during elongational flow of the material prior to the entrance into a smaller region.
Fiber breakage in elongational flow was of more importance, in some cases, than the
breakage that happens in shear flow. Its effect, however, is not independent of other
factors but rather they all work together to determine the final fiber average length.
144

Temperature. The systems viscosity, and thus its stresses, decrease with temperature. On
the other side, fibers mechanical integrity also decreases if the temperature reaches
the fibers onset of thermal degradation. Fiber breakage will depend on the constant
play of the system stresses and fibers strength when processing reaches these high
temperatures.

Fiber concentration and fiber concentration regime. The importance of fiber con-
centration on fiber breakage was corroborated in all experiments. As fiber concentra-
tion increases so does breakage, this is most probable due to the increased interaction
among fibers. This was also concluded in the simulation, since the lack of inclusion
of the fiberfiber interaction caused simulations to over-predict final average lengths
at higher concentrations. An important contribution of this work is that it was found
that the effect of fiber concentration on biofiber breakage seems to depend on the
concentration regime and processing equipment. As fiber concentration reached the
concentrated fiber suspension regime, fibers seemed to be shielded from breakage,
most probably by the flocs that contain them.

Level of shear/stretch rates and RPM. Flow stresses are proportional to deformation
rates in a flow and the higher the stresses the higher the probability of reaching the
fiberss ultimate strength. It was obtained in the capillary rheometer study that
although higher shear/stretch rates are conducive to fiber breakage, the major increase
in breakage when RPMs are increased in a process could be due more to the increase in
the frequency that the material will pass through the gap regions than to the increased
value of stresses.

Processing equipment. The influence of RPM mentioned above is especially important


when the suspensions reach a concentrated level of fibers and more passes though
the gaps would be required to cause fiber breakage. And thus fiber breakage in the
twin-screw extruder in this work was found to be an equivalent between the capillary
rheometer study, which was analogous to one pass through a gap; and the analysis in
the internal mixer, where suspensions were mixed for 6 min.
145

Fiber initial morphology. From the internal mixer study it was found that fibers initial
morphology is correlated to the the level of fiber damage. The longer and thicker
the fibers the greater the reduction in both length and width. The effect of initial
fiber length on the final fiber length, however, was more significant in the twin-screw
extruder than in the internal mixer. In the internal mixer final fiber lengths depended
almost solely on processing conditions such as temperature, fiber concentration and
RPM.

Simulation. In this work, the importance of the fiberpolymer interaction on the breakage
of biofiber was corroborated as simulations gave results that were close to experiments.
Discrepancy between simulation and experiments mostly showed over-prediction of the
experimental data at higher concentration and RPMs.

In addition, the breakage of fibers that happens during capillary flow indicates that the
rheology of the suspension changes during rheological tests. This should be taken into
account when interpreting results from rheological characterization of semi-concentrated
and concentrated biofiber suspensions.

The results presented in this study are empirical in nature. Studies with other mixtures of
natural fibers and polymers would be helpful to ascertain the similarities and differences
within other systems. Nevertheless, the overall understanding from these experiments would
be highly valuable for composite researchers and manufacturers who seek to develop biofiber
composites with enhanced performance as reducing the decrease of fiber aspect ratio after
processing is of greatest value.

This study points out the important of giving proper attention to the feedstock. In general,
the internal mixer results point out the importance of using fibers with low thickness. Under
the same processing conditions, fiber length will be shortened to roughly the same final
values but thinner fibers would end up having the lowest thickness after processing, helping
reduce the decrease in final fiber aspect ratio. In addition, the TSE results indicated that
the feeding of the polymer have a high positive influence on the final fiber length, indicating
that only a small increase of its feeding would help preserving length. Finally, the capillary
146

rheometer study showed that the number of times the mixture passes through the narrow
gaps, due to an increase in RPM, would be more influential on the level of breakage than
the increase in shear stresses that result from the same increase in mixing speed. This result
warrants to explore the use of processing conditions or mixing elements that would provide
better mixing with less passes through the gaps, even if the level of stresses increases during
processing. Finally, this study opens up the idea to start focusing on the elongational flow
generated in convergent sections as a viable mean to decrease fiber breakage, either through
equipment design or polymer rheology.

7.2 Recommendations for Further Work

Furthering the knowledge on the different factors that influence biofiber breakage during
compounding is important. Some areas to explore that depend on the polymers, fiber and
processing are the following.

Polymer. The significance of the interaction between the polymer melt and fibers during
processing goes beyond the direct breakage of the fibers by the polymer. Polymer
rheology as well as its mechanical properties strongly depend on polymer structure and
composition. For example, linear polymers (e.g. LLDPE, L-PP) have better tensile
strength, stiffness and higher puncture resistance than branched polymers. However,
polymers with long chain branches (e.g. LDPE, LCB-PP) show a more pronounced
decrease in viscosity with shear rate, show higher elasticity and have significantly
higher strain hardening behavior than a linear polymer. Lower viscosities at higher
shear rates mean lower stresses and less power to extrude or inject the material.
And strain hardening leads to a more homogeneous deformation of the melt under
elongational flows. It would be interesting to study how different polymer rheological
properties influence the deformation of flocs during flow, affecting the contact between
fibers and fiber breakage. Also polymer blends could be studied as there is evidence
of rheological synergistic effects when blending thermoplastics such as LLDPE with
LDPE, and recently L-PP/LCB-PP.
147

Natural fibers. Results in this work were essentially empirical. Comparison between the
results of the fibers in this work with other biofibers, either chemically or mechani-
cally extracted, will bring us to a closer understanding of the influence of the fibers
type. Additionally, it is important to expand the database of individual specimens
of biofibers mechanical properties at wide ranges of temperatures, including when
possible temperatures above the onset of thermal degradation.

Additives. No additives or agents commonly used to better the flow of the suspensions
or increase the compatibility between biofibers and polymer were used in this work.
Study of fiber breakage in formulations closer to the ones in industry, such as for-
mulations with maleic anhydride grafted polymers or silane coupling agents, is a
recommended step to follow.

Equipment. Dispersive mixing in twin-screw extruders is important to disaggregate fiber


flocs and potentially cause fiber breakage. Dispersive mixing elements have various
designs that subject the suspension to a combination of shear and elongational flows.
Study of fiber breakage under different dispersive mixing elements or combination of
them is required.

Simulation. A necessary step to explore in further simulations would be the inclusion of


the interaction fiber-to-fiber along with a type of fiber width reduction.
List of References

[1] Suddell BC, Evans WJ. Natural fiber composites in automotive applications. In: Mo-
hanty AK, Misra M, Drzal LT, editors. Natural fibers, biopolymers and biocomposites.
Taylor & Francis Group Boca Raton; 2005. .

[2] Mueller DH, Krobjilowski A. New Discovery in the Properties of Composites Rein-
forced with Natural Fibers. J Ind Textiles. 2003;33(2):111130.

[3] Faruk O, Bledzki AK, Fink HP, Sain M. Biocomposites reinforced with natural fibers:
20002010. Prog Polym Sci. 2012;37(11):15521596.

[4] Begum K, Islam M. Natural fiber as a substitute to synthetic fiber in polymer com-
posites: a review. Research J Eng Sci. 2013;2278:9472.

[5] Satyanarayana KG, Arizaga GG, Wypych F. Biodegradable composites based on


lignocellulosic fibersAn overview. Prog Polym Sci. 2009;34(9):9821021.

[6] Van de Velde K, Kiekens P. Biopolymers: overview of several properties and conse-
quences on their applications. Polymer Testing. 2002;21(4):433442.

[7] Koronis G, Silva A, Fontul M. Green composites: a review of adequate materials for
automotive applications. Compos, Part B Eng. 2013;44(1):120127.

[8] Averous L, Moro L, Dole P, Fringant C. Properties of thermoplastic blends: starch


polycaprolactone. Polymer. 2000;41(11):41574167.

148
149

[9] Nagarajan V, Mohanty AK, Misra M. Sustainable green composites: value addi-
tion to agricultural residues and perennial grasses. ACS Sustainable Chem Eng.
2013;1(3):325333.

[10] Ma X, Yu J, Kennedy JF. Studies on the properties of natural fibers-reinforced


thermoplastic starch composites. Carbohydrate Polymers. 2005;62(1):1924.

[11] Lauke B, Fu SY. Strength anisotropy of misaligned short-fibre-reinforced polymers.


Compos Sci Technol. 1999;59(5):699708.

[12] Fu SY, Hu X, Yue CY. Effects of fiber length and orientation distributions on the
mechanical properties of short-fiber-reinforced polymers a review. Mater Sci Res Int.
1999;5(2):7483.

[13] Fu SY, Lauke B. The Elastic Modulus of Misaligned Short-Fiber-Reinforced Polymers.


Compos Sci Technol. 1998;58(3-4):389400.

[14] Fu SY, Lauke B. An analytical characterization of the anisotropy of the elas-


tic modulus of misaligned short-fiber-reinforced polymers. Compos Sci Technol.
1998;58(12):19611972.

[15] Callister WD. Materials Science and Engineering: An Introduction. sixth ed. John
Wiley and Sons; 2003.

[16] De Bruijn JCM. Natural Fibre Mat Thermoplastic Products from a Processors Point
of View. Appl Compos Mater. 2000;7:415420.

[17] English B, Stark N, Clemons C. Weight Reduction: Wood versus Mineral Fillers
in Polypropylene. In: Proc. Conference on Woodfiber-Plastic Composites; 1997. p.
237244.

[18] Pickering KL, Ji C. The Effect of Poly[methylene(polyphenyl isocyanate)] and


Maleated Polypropylene Coupling Agents on New Zealand Radiata Pine Fiber-
Polypropylene Composites. J Reinf Plast Comp. 2004;23(18):20112024.
150

[19] Kazayawoko M, Balatinecz JJ, Matuana LM. Surface Modification and Adhesion
Mechanisms in Woodfiber-Polypropylene Composites. J Mater Sci. 1999;34:6189
6199.

[20] Matuana LM, Balatinecz JJ, Sodhi RNS, Park CB. Surface Characterization of Esteri-
fied Cellulosic Fibers by XPS and FTIR Spectroscopy. Wood Sci Technol. 2001;35:191
201.

[21] Sain M, Suhara P, Law S, Bouilloux A. Interface Modification and Mechanical


Properties of Natural Fiber-Polyolefin Composite Products. J Reinf Plast Comp.
2005;24(2):121130.

[22] Hristov VN, Vasileva ST, Krumova M, Lach R, Michler GH. Deformation mechanisms
and mechanical properties of modified polypropylene/wood fiber composites. Polym
Composites. 2004;25(5):521526.

[23] Krzysik AM, Youngquist JA, Myers GE, Chahyadi IS, c Kolosick P. Wood-Polymer
Bonding in Extruded and Nonwoven Web Composite Panels. In: Conner AH, Chris-
tiansen AW, Myers GE, Others, editors. Wood Adhesives, 1990: Proceedings of a
Symposium Sponsored by the Usda Forest Service, Forest Products Laboratory and
the Forest Products Research So. Forest Products Society; 1991. p. 183189.

[24] Mali J, Sarsama P, Suome-Lindberg L, Metsa-Kortelainen S, Peltonen J, Vilkki M,


et al.. Woodfibre-Plastic Composites; 2003. Report, VTT Technical Research Centre
of Finland, Finland.

[25] Valadez-Gonzalez A, Cervantes-Uc JM, Olayo R, Herrera-Franco PJ. Chemical Mod-


ification of Henequen fibers with an Organosilane Coupling Agent. Compos, Part B
Eng. 1999;30:321331.

[26] Pickering KL, Abdalla A, Ji C, McDonald AG, Franich RA. The Effect of Silane
Coupling Agents on Radiata Pine Fibre for Use in Thermoplastic Matrix Composites.
Compos, Part A Appl Sci Manuf. 2003;34:915926.
151

[27] Ashori A. Woodplastic composites as promising green-composites for automotive


industries! Bioresour Technol. 2008;99(11):46614667.

[28] Chen Y, Sun L, Chiparus O, Negulescu I, Yachmenev V, Warnock M. Kenaf/ramie


composite for automotive headliner. J Polym Env. 2005;13(2):107114.

[29] Cengiz TG, Babalk FC. The effects of ramie blended car seat covers on thermal
comfort during road trials. Int J Ind Ergonom. 2009;39(2):287294.

[30] Pandey JK, Ahn S, Lee CS, Mohanty AK, Misra M. Recent advances in the application
of natural fiber based composites. Macromol Mater Eng. 2010;295(11):975989.

[31] Alves C, Ferr


ao P, Silva A, Reis L, Freitas M, Rodrigues L, et al. Ecodesign of
automotive components making use of natural jute fiber composites. J Cell Plast.
2010;18(4):313327.

[32] Jewett D. DaimlerChrysler goes natural for large body panel. Automotive Industries.
2000;180:9.

[33] Rohatgi V, Adur A, Shih K, Botros M, Previty R, Castle G. Cellulose - polymer


composites and related manufacturing methods; 2003. US Patent US 20030021915
A1.

[34] Bledzki A, Sperber V, Faruk O. Natural and wood fibre reinforcement in polymers.
Rapra Review Report. 2002;152:13.

[35] Brooks JG, Goforth BD, Goforth CL. Recycled thermoplastic polymers and cellulose
fiber particles encasulated in thermoplastic polymers; used as building material; re-
sistance to rotting, insects, moisture absorption and warping. Google Patents; 1998.
US Patent 5,759,680.

[36] Cooper T. Slower consumption reflections on product life spans and the throwaway
society. J Ind Ecol. 2005;9(1-2):5167.

[37] Dicker MP, Duckworth PF, Baker AB, Francois G, Hazzard MK, Weaver PM. Green
composites: a review of material attributes and complementary applications. Compos,
Part A Appl Sci Manuf. 2014;56:280289.
152

[38] Corbiere-Nicollier T, Laban BG, Lundquist L, Leterrier Y, M


anson JA, Jolliet O. Life
cycle assessment of biofibres replacing glass fibres as reinforcement in plastics. Resour
Conserv Recy. 2001;33(4):267287.

[39] Piorkowska E, Rutledge GC. Handbook of polymer crystallization. John Wiley &
Sons; 2013.

[40] Gabbott P. Principles and applications of thermal analysis. John Wiley & Sons; 2008.

[41] Donth EJ. The glass transition: relaxation dynamics in liquids and disordered ma-
terials. vol. 48 of Springer Series in Materials Science. Springer Science & Business
Media; 2001.

[42] Rauwendaal C. Polymer Mixing: A Self-Study Guide. 1st ed. Hanser Gardner Pub-
lications; 1998.

[43] Park JM, Quang ST, Hwang BS, DeVries KL. Interfacial evaluation of modified
Jute and Hemp fibers/polypropylene (PP)-maleic anhydride polypropylene copoly-
mers (PP-MAPP) composites using micromechanical technique and nondestructive
acoustic emission. Compos Sci Technol. 2006;66(15):26862699.

[44] Netravali AN. Ramie Fiber Reinforced Natural Plastics. In: Wallenberger FT, Weston
NE, editors. Natural fibers, plastics and composites. Kluwer Academic Publishers;
2004. .

[45] Tomczak F, Satyanarayana KG, Sydenstricker THD. Studies on lignocellulosic fibers


of Brazil: Part IIIMorphology and properties of Brazilian curaua fibers. Compos,
Part A Appl Sci Manuf. 2007;38(10):22272236.

[46] Mohanty AK, Misra M, Drzal LT. Natural Fibers, Biopolymers and Biocomposites.
Taylor & Francis Group Boca Raton; 2005.

[47] Hubbe MA, Rojas OJ, Lucia LA, Sain M. Cellulosic nanocomposites: a review.
BioResources. 2008;3(3):929980.

[48] Beck CB. An introduction to plant structure and development: plant anatomy for
the twenty-first century. 2nd ed. Cambridge University Press; 2010.
153

[49] Cruz-Ramos CA. Natural fiber reinforced thermoplastics. In: Clegg DW, Collyer AA,
editors. Mechanical properties of reinforced thermoplastics. Elsevier applied science
publishers; 1986. p. 6581.

[50] Philip BM, Abraham E, B D, Pothan LA, Thomas S. Plant FiberBased Composites.
In: Thakur VK, editor. Green composites from natural resources. CRC Press; 2013. .

[51] Shah DU. Developing plant fibre composites for structural applications by optimising
composite parameters: a critical review. J Mater Sci. 2013;48(18):60836107.

[52] Oksman K, Mathew AP, L


angstrom R, Nystrom B, Joseph K. The influence of fibre
microstructure on fibre breakage and mechanical properties of natural fibre reinforced
polypropylene. Compos Sci Technol. 2009;69(11):18471853.

[53] Tadmor Z, Gogos CG. Principles of polymer processing. 2nd ed. John Wiley & Sons;
2006.

[54] Vlachopoulos J. Introduction to Plastics Processing; 2001. Lecture Notes, Dept. of


Chem. Eng., McMaster University, Canada.

[55] Bird RB, Armstrong RC, Hassager O. Dynamics of polymeric liquids. Volume 1: fluid
mechanics. 2nd ed. John Wiley & Sons; 1987.

[56] Cheremisinoff NP. Polymer Mixing and Extrusion Technology. 1st ed. Marcel Dekker
Incorporated; 1987.

[57] White JL, Bumm SH. Polymer Blend Compounding and Processing. In: Isayev AI,
editor. Encyclopedia of Polymer Blends, Volume 2: Processing. John Wiley & Sons;
2011. .

[58] Cheng JJ, Manas-Zloczower I. Hydrodynamic analysis of a Banbury mixer. Polym


Eng Sci. 1989;29(11):701708.

[59] Kuroda MM, Scott CE. Blade geometry effects on initial dispersion of chopped glass
fibers. Polym Composites. 2002;23(5):828838.
154

[60] Freakley P, Idris WW. Visualization of flow during the processing of rubber in an
internal mixer. Rubber Chem Technol. 1979;52(1):134145.

[61] Shenoy AV. Rheology of filled polymer systems. Springer Science & Business Media;
1999.

[62] Rauwendaal C. Polymer extrusion. 4th ed. Hanser Gardner Publications; 2001.

[63] Wolcott MP, Englund K. A technology review of wood-plastic composites. 1999;.

[64] Rauwendaal CJ. Analysis and experimental evaluation of twin screw extruders. Polym
Eng Sci. 1981;21(16):10921100.

[65] White JL, Kim EK. Twin screw extrusion: technology and principles. Hanser; 1991.

[66] Cheng H, Manas-Zloczower I. Study of mixing efficiency in kneading discs of co-


rotating twin-screw extruders. Polym Eng Sci. 1997;37(6):10821090.

[67] Rauwendaal C. New dispersive mixers based on elongational flow. Plastics, Additives
and Compounding. 1999;1(4):2123.

[68] Bravo V, Hrymak A, Wright J. Study of particle trajectories, residence times and flow
behavior in kneading discs of intermeshing co-rotating twin-screw extruders. Polym
Eng Sci. 2004;44(4):779793.

[69] Elemans P, Van Wunnik J. The effect of feeding mode on dispersive mixing efficiency
in single-screw extrusion. Polymer Engineering & Science. 2001;41(7):10991106.

[70] Wilczy
nski K, Lewandowski A, Wilczy
nski KJ. Experimental study for starve-fed
single screw extrusion of thermoplastics. Polym Eng Sci. 2012;52(6):12581270.

[71] Macosko CW. Rheology: principles, measurements, and applications. 1994;.

[72] Cogswell F. Converging flow of polymer melts in extrusion dies. Polym Eng Sci.
1972;12(1):6473.

[73] Cogswell F. Measuring the extensional rheology of polymer melts. Trans Soc Rheol.
1972;16(3):383403.
155

[74] Kwag C, Vlachopoulos J. An assessment of Cogswells method for measurement of


extensional viscosity. Polym Eng Sci. 1991;31(14):10151021.

[75] Rajagopalan D. Computational analysis of techniques to determine extensional vis-


cosity from entrance flows. Rheol Acta. 2000;39(2):138151.

[76] Shroff RN, Cancio L, Shida M. Extensional flow of polymer melts. Trans Soc Rheol.
1977;21(3):429446.

[77] Covas J, Carneiro O. Assessing the convergent flow analysis as a technique for char-
acterizing the extensional flow of polymer melts. Polym Testing. 1990;9(3):181194.

[78] Bailey R, Kraft H. A Study of Fibre Attrition in the Processing of Long Fibre
Reinforced Thermoplastics. Int Polym Process. 1987;2:94101.

[79] Wu JY, Han WK, Chiang CC, Huang CC, Lee MS, Hu AT. . In: Proc. SPE ANTEC;
1991. p. 2032.

[80] Kottyan RE, Rosenthal J. . In: Proc. SPE ANTEC; 1991. p. 2028.

[81] Bijsterbosch H, Gaymans RJ. Polyamide 6 - Long Glass Fiber Injection Moldings.
Polym Composites. 1995;16(5):363369.

[82] Ramani K, Bank D, Kraemer N. Effect of Screw Design of Fiber Damage in Extrusion
Compounding and Composite Properties. Polym Composites. 1995;16(3):258266.

[83] Shon K, Liu D, White JL. Experimental studies and modeling of development
of dispersion and fiber damage in continuous compounding. Int Polym Process.
2005;20(3):322331.

[84] Arzondo LM, Perez CJ, Carella JM. Injection Molding of Long Sisal Riber-Reinforced
Polypropylene: Effects of Compatibilizer Concentration and Viscosity on Fiber Ad-
hesion and Thermal Degradation. Polym Eng Sci. 2005;45(4):613621.

[85] Joseph PV, Mathew G, Joseph K, Thomas S, Pradeep P. Mechanical properties of


short sisal fiber-reinforced polypropylene composites: Comparison of experimental
data with theoretical predictions. J Appl Polym Sci. 2003;88(3):602611.
156

[86] Alvarez V, Iannoni A, Kenny JM, Vazquez A. Influence of twin-screw process-


ing conditions on the mechanical properties of biocomposites. J Compos Mater.
2005;39(22):20232038.

[87] Joseph PW, Joseph K, Thomas S. Effect of processing variables on the mechanical
properties of sisal-fiber-reinforced polypropylene composites. Compos Sci Technol.
1999;59(11):16251640.

[88] Alvarez VA, Terenzi A, Kenny JM, Vazquez A. Melt rheological behavior of
starch-based matrix composites reinforced with short sisal fibers. Polym Eng Sci.
2004;44(10):19071914.

[89] Karmaker AC, Youngquist JA. Injection molding of polypropylene reinforced with
short jute fibers. J Appl Polym Sci. 1996;62(8):11471151.

[90] Mutje P, Girones J, L


opez A, Llop MF, Vilaseca F. Hemp strands: PP composites
by injection molding: effect of low cost physico-chemical treatments. J Reinf Plast
Comp. 2006;25(3):313327.

[91] George J, Bhagawan SS, Prabhakaran N, Thomas S. Short Pineapple-Leaf-Fiber-


Reinforced Low-Density Polyethylene Composites. J Appl Polym Sci. 1995;57(7):843
854.

[92] van den Oever MJA, Bos HL. Critical fibre length and apparent interfacial
shear strength of single flax fibre polypropylene composites. Adv Compos Letters.
1998;7(3):8185.

[93] Sojoudiasli H, Heuzey MC, Carreau PJ. Rheological, morphological and mechan-
ical properties of flax fiber polypropylene composites: influence of compatibilizers.
Cellulose. 2014;21(5):37973812.

[94] El-Sabbagh AMM, Steuernagel L, Meiners D, Ziegmann G. Effect of extruder elements


on fiber dimensions and mechanical properties of bast natural fiber polypropylene
composites. J Appl Polym Sci. 2014;131(12).
157

[95] Berzin F, Vergnes B, Beaugrand J. Evolution of lignocellulosic fibre lengths along the
screw profile during twin screw compounding with polycaprolactone. Compos, Part
A Appl Sci Manuf. 2014;59:3036.

[96] Beaugrand J, Berzin F. Lignocellulosic fiber reinforced composites: influence of com-


pounding conditions on defibrization and mechanical properties. J Appl Polym Sci.
2013;128(2):12271238.

[97] Gamon G, Evon P, Rigal L. Twin-screw extrusion impact on natural fibre morphology
and material properties in poly (lactic acid) based biocomposites. Ind Crops Prod.
2013;46:173185.

[98] Iannace S, Ali R, Nicolais L. Effect of processing conditions on dimensions of sisal


fibers in thermoplastic biodegradable composites. J Appl Polym Sci. 2001;79(6):1084
1091.

[99] Le Baillif M, Oksman K. The effect of processing on fiber dispersion, fiber length,
and thermal degradation of bleached sulfite cellulose fiber polypropylene composites.
J Thermoplast Compos Mater. 2009;22(2):115133.

[100] Davies GC, Bruce DM. Effect of environmental relative humidity and damage on the
tensile properties of flax and nettle fibers. Text Res J. 1998;68(9):623629.

[101] Mott L, Shaler SM, Groom LH. A technique to measure strain distributions in single
wood pulp fibers. Wood Fiber Sci. 1996;28(4):429437.

[102] Hughes M. Defects in natural fibres: their origin, characteristics and implications for
natural fibre-reinforced composites. J Mater Sci. 2012;47(2):599609.

[103] Bos H, Van Den Oever M, Peters O. Tensile and compressive properties of flax fibres
for natural fibre reinforced composites. J Mater Sci. 2002;37(8):16831692.

[104] Baley C. Analysis of the flax fibres tensile behaviour and analysis of the tensile
stiffness increase. Compos, Part A Appl Sci Manuf. 2002;33(7):939948.

[105] Li TQ, Wolcott MP. Rheology of HDPE-Wood Composites. I. Steady State Shear
and Extensional Flow. Compos, Part A Appl Sci Manuf. 2004;35(3):303311.
158

[106] Maschmeyer RO, Hill CT. Rheology of Concentrated Suspensions of Fibers in Tube
Flow. III. Suspensions with the Same Fiber Length Distribution). Trans Soc Rheol.
1977;21(2):195206.

[107] Yilmazer U, Cansever M. Effects of Reprocessing on the Fiber Length and Mechanical
Properties of Nylon-6/Glass Fiber Composites. In: Proc. SPE ANTEC; 2001. p. 59.

[108] Le Duc A, Vergnes B, Budtova T. Polypropylene/natural fibres composites: analysis of


fibre dimensions after compounding and observations of fibre rupture by rheo-optics.
Compos, Part A Appl Sci Manuf. 2011;42(11):17271737.

[109] Huq AMA, Azaiez J. Effects of Length Distribution on the Steady Shear Viscosity
of Semiconcentrated Polymer-Fiber Suspensions. Polym Eng Sci. 2005;45(10):1357
1368.

[110] Turkovich R, Erwin L. Fiber fracture in reinforced thermoplastic processing. Polym


Eng Sci. 1983;23(13):743749.

[111] Switzer III LH, Klingenberg DJ. Rheology of Sheared Flexible Fiber Suspensions via
Fiber-Level Simulations. J Rheol. 2003;47(3):759778.

[112] Chaouche M, Koch DL. Rheology of non-Brownian Rigid Fiber Suspensions with
Adhesive Contacts. J Rheol. 2001;45(2):369382.

[113] Montgomery DC, Runger GC. Applied statistics and probability for engineers. John
Wiley & Sons; 2010.

[114] Czarnecki L, White JL. Shear Flow Rheological Properties, Fiber Damage,and Masti-
cation Characteristics of Aramid-, Glass-, and Cellulose-Fiber-Reinforced Polystyrene
Melts. J Appl Polym Sci. 1980;25(6):12171244.

[115] Hine PJ, Lusti HR, Gusev AA. Numerical Simulation of the Effects of Volume Frac-
tion, Aspect Ratio and Fibre Length Distribution on the Elastic and Thermoelastic
Properties of Short Fibre Composites. Compos Sci Technol. 2002;62(10):14451453.

[116] Fu SY, Yue CY, Hu X, Mai YW. Characterization of fiber length distribution of
short-fiber reinforced thermoplastics. J Mater Sci Letters. 2001;20:3133.
159

[117] Joslin CG, Stell G. Effective Properties of Fiber-Reinforced Composites: Effects of


Polydispersity in Fiber Diameter. J Appl Phys. 1986;60(5):16111613.

[118] Takao Y, Taya M. The Effect of Variable Fiber Aspect Ratio on the Stiffness and
Thermal Expansion Coefficients of a Short Fiber Composite. J Compos Mater.
1987;21(2):140156.

[119] OpTest Equipment Inc . Operation Manual. Fiber Quality Analyzer Code LDA02.
900 Tupper, Hawkesbury, Ontario, Canada K6A 3S3; 1999.

[120] Heckert N, Filliben J. NIST Handbook 148: Dataplot Reference Manual, Volume
2: Let Subcommands and Library Functions. National Institute of Standards and
Technology Handbook Series; 1999.

[121] Wolf HJ. Screw plasticating of discontinuous fiber filled thermoplastic: Mechanisms
and prevention of fiber attrition. Polym Composites. 1994;15(5):375383.

[122] Kumar RP, Nair KCM, Thomas S, Schit SC, Ramamurthy K. Morphology and Melt
Rheological Behaviour of Short-Sisal-Fibre-Reinforced SBR Composites. Compos Sci
Technol. 2000;60:17371751.

[123] Joseph K, Kuriakose B, Premalatha CK, Thomas S. Melt Rheological Behaviour of


Short Sisal Fibre Reinforced Polyethylene Composites. Plast Rubber Compos Process.
1994;21(4):237245.

[124] Xiao K, Tzoganakis C. Rheological Properties of HDPE-Wood Composites. In: Proc.


SPE ANTEC; 2003. p. 975979.

[125] Li H, Law S, Sain M. Process Rheology and Mechanical Property Correlationship of


Wood Flour-Polypropylene Composites. J Reinf Plast Comp. 2004;23(11):11531158.

[126] Ayora M, Rios R, Quijano J, Marquez A. Evaluation by Torque-Rheometer of Suspen-


sions of Semi-Rigid and Flexible Natural Fibers in a Matrix of Poly(Vinyl Chloride).
Polym Composites. 1997;18(4):549560.

[127] Tucker III CL, Advani SG. Processing of short-fiber systems. In: Advani SG, editor.
Flow and Rheology in Polymer Composites Manufacturing. Elsevier; 1994. p. 147202.
160

[128] Rajabian M, Dubois C, Grmela M. Suspensions of Semiflexible Fibers in Polymeric


Fluids: Rheology and Thermodynamics. Rheol Acta. 2005;44(5):521535.

[129] Joung CG, Phan-Thien N, Fan XJ. Direct Simulation of Flexible Fibers. J Non-
Newtonian Fluid Mech. 2001;99(1):136.

[130] Schmid CF, Switzer LH, Klingenberg DJ. Simulations of Fiber Flocculation: Effects
of Fiber Properties and Interfiber Friction. J Rheol. 2000;44(4):781809.

[131] Moses KB, Advani SG, Reinhardt A. Investigation of Fiber Motion Near Solid Bound-
aries in Simple Shear Flow. Rheol Acta. 2001;40(3):296306.

[132] Forgacs OL, Mason SG. Particle motions in sheared suspensions X. Orbits of flexible
threadlike particles. J Colloid Sci. 1959;14:473491.

[133] Salinas A, Pittman JFT. Bending and Breaking Fibers in Sheared Suspensions. Polym
Eng Sci. 1981;21(1):2331.

[134] Yamamoto S, Matsuoka T. Dynamic simulation of flow-induced fiber fracture. Polym


Eng Sci. 1995;35(12):10221030.

[135] Guo R, Azaiez J, Bellehumeur C. Rheology of fiber filled polymer melts: Role of fiber-
fiber interactions and polymer-fiber coupling. Polym Eng Sci. 2005;45(3):385399.

[136] Selke SE, Wichman I. Wood fiber/polyolefin composites. Compos, Part A Appl Sci
Manuf. 2004;35(3):321326.

[137] Bledzki A, Reihmane S, Gassan J. Thermoplastics reinforced with wood fillers: a


literature review. Polym Plast Tech Eng. 1998;37(4):451468.

[138] Kumar KV, Safiulla M, Ahmed AK. An Experimental Evaluation of Fiber Re-
inforced Polypropylene Thermoplastics for Aerospace Applications. J Mech Eng.
2014;43(2):9297.

[139] Karian H. Handbook of polypropylene and polypropylene composites, revised and


expanded. 2nd ed. Marcel Dekker; 2009.
161

[140] Mohanty A, Drzal L, Misra M. Engineered natural fiber reinforced polypropylene com-
posites: influence of surface modifications and novel powder impregnation processing.
J Adhes SciTechnol. 2002;16(8):9991015.

[141] Maschmeyer RO, Hill CT. Rheology of Concentrated Suspensions of Fibers in Tube
Flow. II. An Exploratory Study). Trans Soc Rheol. 1977;21(2):183194.

[142] Irvine GM. The significance of the glass transition of lignin in thermomechanical
pulping. Wood Sci Technol. 1985;19(2):139149.

[143] Rowell RM, Sanadi AR, Caulfield DF, Jacobson RE. Utilization of Natural Fibres in
Plastic Composites: Problems and Opportunities, Lignocellulosic-Plastics Compos-
ites. Department of Forestry, University of Wisconsin; 1997.

[144] Facca AG, Kortschot MT, Yan N. Predicting the elastic modulus of natural fibre
reinforced thermoplastics. Compos, Part A Appl sci manuf. 2006;37(10):16601671.

[145] Robertson G, Olson J, Allen P, Chan B, Seth R. Measurement of fiber length, coarse-
ness, and shape with the fiber quality analyzer. Tappi J. 1999;82(10):9398.

[146] Wold S, Sj
ostr
om M, Eriksson L. PLS-regression: a basic tool of chemometrics.
Chemom Intell Lab Syst. 2001;58(2):109130.

[147] Aastveit AH, Martens H. ANOVA interactions interpreted by partial least squares
regression. Biometrics. 1986;p. 829844.

[148] Marquez A, Quijano J, Gaulin M. A calibration technique to evaluate the power-


law parameters of polymer melts using a torque-rheometer. Polym Eng Sci.
1996;36(20):25562563.

[149] Petrie CJS. The Rheology of Fibre Suspensions. J Non-Newtonian Fluid Mech.
1999;87(2-3):369402.

[150] Kerekes RJ, Schell CJ. Effects of fiber length and coarseness on pulp flocculation.
Tappi J. 1995;78(2):133139.
162

[151] Farnood RR, Loewen SR, Dodson C. Estimation of intra-floc forces. Appita J.
1994;47(5):391396.

[152] Huber P, Carre B, Petit-Conil M. The influence of TMP fibre flexibility on flocculation
and formation. BioResources. 2008;3(4):12181227.

[153] Soszynski R, Kerekes R. Elastic interlocking of nylon fibers suspended in liquid. Part
1. Nature of cohesion among fibers. Nord Pulp Pap Res J. 1988;3(4):172179.

[154] Switzer LH, Klingenberg DJ. Flocculation in simulations of sheared fiber suspensions.
Int J Multiphase Flow. 2004;30(1):6787.

[155] Sudduth RD. A generalized model to predict the viscosity of solutions with suspended
particles. III. Effects of particle interaction and particle size distribution. J Appl
Polym Sci. 1993;50(1):123147.

[156] Joshi SV, Drzal L, Mohanty A, Arora S. Are natural fiber composites environmentally
superior to glass fiber reinforced composites? Compos, Part A Appl Sci Manuf.
2004;35(3):371376.

[157] Gunning MA, Geever LM, Killion JA, Lyons JG, Higginbotham CL. Melt processing
of bioplastic composites via twin screw extrusion and injection molding. Polym Plast
Tech Eng. 2014;53(4):379386.

[158] Guo Q, Cheng B, Kortschot M, Sain M, Knudson R, Deng J, et al. Performance


of long Canadian natural fibers as reinforcements in polymers. J Reinf Plast Comp.
2010;29(21):31973207.

[159] Kumar RP, Nair KM, Thomas S, Schit S, Ramamurthy K. Morphology and melt rheo-
logical behaviour of short-sisal-fibre-reinforced SBR composites. Compos Sci Technol.
2000;60(9):17371751.

[160] Loukus J, Halonen A, M G. Elongational flow in multiple screw extruders. In: Proc.
SPE ANTEC; 2004. .

[161] Lee WK, George HH. Flow visualization of fiber suspensions. Polym Eng Sci;(2):146
156.
163

[162] Thomasset J, Carreau PJ, Sanschagrin B, Ausias G. Rheological properties of long


glass fiber filled polypropylene. J Non-Newtonian Fluid Mech. 2005;125(1):2534.

[163] Van Der Wal D, Goffart D, Klomp E, Hoogstraten H, Janssen L. Three-dimensional


flow modeling of a self-wiping corotating twin-screw extruder. Part II: The kneading
section. Polym Eng Sci. 1996;36(7):912924.

[164] Wang SQ, Drda PA. Stick-slip transition in capillary flow of polyethylene. 2. Molecular
weight dependence and low-temperature anomaly. Macromolecules. 1996;29(11):4115
4119.

[165] Hristov V, Vlachopoulos J. Effects of polymer molecular weight and filler particle size
on flow behavior of wood polymer composites. Polym Composites. 2008;29(8):831
839.

[166] Metzner A. Rheology of suspensions in polymeric liquids. J Rheol. 1985;29(6):739


775.

[167] Rides M. Rheological characterisation of filled materials: a review. National Physical


Laboratory; 2005.

[168] Gibson A, Williamson G. Shear and extensional flow of reinforced plastics in injection
molding. II. Effects of die angle and bore diameter on entry pressure with bulk molding
compound. Polym Eng Sci. 1985;25(15):980985.

[169] Gibson A. Die entry flow of reinforced polymers. Composites. 1989;20(1):5764.

[170] Murty KN, Modlen G. Experimental characterization of the alignment of short fibers
during flow. Polym Eng Sci. 1977;17(12):848853.

[171] Crowson R, Folkes M, Bright P. Rheology of short glass fiber-reinforced thermoplastics


and its application to injection molding I. Fiber motion and viscosity measurement.
Polym Eng Sci. 1980;20(14):925933.

[172] Van den Oever M, Snijder M. Jute fiber reinforced polypropylene produced by con-
tinuous extrusion compounding, part 1: Processing and ageing properties. J Appl
Polym Sci. 2008;110(2):10091018.
164

[173] Karnis A, Goldsmith H, Mason S. The kinetics of flowing dispersions: I. Concentrated


suspensions of rigid particles. J Colloid Interface Sci. 1966;22(6):531553.

[174] Hatzikiriakos SG, Kazatchkov IB, Vlassopoulos D. Interfacial phenomena in the cap-
illary extrusion of metallocene polyethylenes. J Rheol. 1997;41(6):12991316.

[175] Mark H. Encyclopedia of polymer science and technology, concise. 3rd ed. John Wiley
& Sons; 2007.

[176] Dinh SM, Armstrong RC. A rheological equation of state for semiconcentrated fiber
suspensions. J Rheol. 1984;28(3):207227.

[177] Pervaiz M, Sain MM. Carbon storage potential in natural fiber composites. Resour
Conserv Recy. 2003;39(4):325340.

[178] Chatterjee S, Hadi AS. Regression analysis by example. 4th ed. John Wiley & Sons;
2013.

[179] Kapp S. A comparison of process configurations for compounding woodfiber-plastic


composites. In: Proc. SPE ANTEC; 2005. p. 21222126.

[180] Cassel C, Hackl P, Westlund AH. Robustness of partial least-squares method for
estimating latent variable quality structures. J Appl Statist. 1999;26(4):435446.

[181] Sundararajakumar RR, Koch DL. Structure and properties of sheared fiber suspen-
sions with mechanical contacts. J Non-Newtonian Fluid Mech. 1997;73(3):205239.

[182] Wold S. PLS for multivariate linear modeling. In: Van de Waterbeemd H, editor.
Chemometric Methods in Molecular Design. Methods and Principles in Medicinal
Chemistry. Weinheim, Germany: Verlag- Chemie; 1994. p. 195218.

[183] Saheb DN, Jog J. Natural fiber polymer composites: a review. Adv Polym Tech.
1999;18(4):351363.

[184] Holbery J, Houston D. Natural-fiber-reinforced polymer composites in automotive


applications. JOM. 2006;58(11):8086.
165

[185] Wolcott M, Smith P. Opportunities and challenges for wood-plastic composites in


structural applications. In: Proceedings of Progress in Woodfibre-Plastic Composites-
2004 Toronto, ON; 2004. .

[186] Netravali AN, Chabba S. Composites get greener. Materials today. 2003;6(4):2229.

[187] Dittenber DB, GangaRao HV. Critical review of recent publications on use of natural
composites in infrastructure. Compos, Part A Appl Sci Manuf. 2012;43(8):14191429.

[188] Godavarti S. Thermoplastic wood fiber composites. In: Mohanty AK, Misra M, Drzal
LT, editors. Natural fibers, biopolymers and biocomposites. Taylor & Francis Group
Boca Raton; 2005. .

[189] Arbelaiz A, Fernandez B, Ramos J, Retegi A, Llano-Ponte R, Mondragon I. Me-


chanical properties of short flax fibre bundle/polypropylene composites: Influence of
matrix/fibre modification, fibre content, water uptake and recycling. Compos Sci
Technol. 2005;65(10):15821592.

[190] Forgacs O, Mason S. Particle motions in sheared suspensions: IX. Spin and deforma-
tion of threadlike particles. J Colloid Sci. 1959;14(5):457472.

[191] Bereaux Y, Charmeau JY, Moguedet M. Modelling of fibre damage in single screw
processing. Int J Multiphase Flow. 2008;1(1):827830.

[192] Kabanemi KK, Hetu JF. Effects of bending and torsion rigidity on deforma-
tion and breakage of flexible fibers: A direct simulation study. J Chem Phys.
2012;136(7):074903.

[193] Manas-Zloczower I. Mixing in high intensity batch mixers. In: Rauwendaal C, editor.
Mixing in polymer processing. CRC Press; 1991. p. 323376.

[194] Hernandez JP, Raush T, Rios A, Strauss S, Osswald TA. Theoretical analysis of fiber
motion and loads during flow. Polym Composites. 2004;25(1):111.

[195] Bai Y, Sundararaj U, Nandakumar K. Nonisothermal modeling of heat transfer inside


an internal batch mixer. AIChE J. 2011;57(10):26572669.
166

[196] Michaeli W. Extrusion Dies for Plastics and Rubber: Design and Engineering Com-
putations. 3rd ed. Hanser Publishers; 2003.

[197] Green DW, Winandy JE, Kretschmann DE. Mechanical properties of wood. In: Wood
handbook: wood as an engineering material. Gen. Tech. Rep. FPLGTR113. Madison,
WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory.;
1999. p. 76120.

[198] Chand N, Hashmi S. Mechanical properties of sisal fibre at elevated temperatures. J


Mater Sci. 1993;28(24):67246728.

[199] Hornsby P, Hinrichsen E, Tarverdi K. Preparation and properties of polypropylene


composites reinforced with wheat and flax straw fibres: part I fibre characterization.
J Mater Sci. 1997;32(2):443449.

[200] Laws V. The relationship between tensile and bending properties of non-linear com-
posite materials. J Mater Sci. 1982;17(10):29192924.

[201] Panthapulakkal S, Zereshkian A, Sain M. Preparation and characterization of wheat


straw fibers for reinforcing application in injection molded thermoplastic composites.
Bioresour Technol. 2006;97(2):265272.

[202] Thomopoulos NT. Essentials of Monte Carlo simulation: Statistical methods for
building simulation models. Springer Science & Business Media; 2012.

[203] Attaway S. Matlab: a practical introduction to programming and problem solving.


Butterworth-Heinemann; 2013.
167

Appendices

Appendix A Chapter 3 processed samples data

Table A.1: Data for all AL samples processed in the internal mixer
Temperature RPM Fiber concentration Lw Wn L/Dn

C wt.% mm m
180 40 10 0.90 ( 0.012 ) 46.0 ( 0.86 ) 20.0 ( 0.35 )
180 40 10 0.75 ( 0.010 ) 47.4 ( 0.87 ) 17.6 ( 0.30 )
180 40 10 0.79 ( 0.010 ) 46.5 ( 0.87 ) 18.4 ( 0.31 )
180 40 30 0.51 ( 0.004 ) 35.5 ( 0.61 ) 16.4 ( 0.21 )
180 40 30 0.64 ( 0.006 ) 36.6 ( 0.59 ) 18.7 ( 0.24 )
180 40 30 0.63 ( 0.005 ) 35.8 ( 0.53 ) 18.3 ( 0.21 )
180 60 10 0.70 ( 0.008 ) 39.1 ( 0.62 ) 19.0 ( 0.29 )
180 60 10 0.70 ( 0.006 ) 38.3 ( 0.47 ) 19.0 ( 0.22 )
180 60 10 0.69 ( 0.007 ) 38.3 ( 0.57 ) 18.7 ( 0.26 )
180 60 30 0.52 ( 0.005 ) 32.8 ( 0.62 ) 18.1 ( 0.25 )
180 60 30 0.49 ( 0.004 ) 35.4 ( 0.68 ) 16.2 ( 0.23 )
180 60 30 0.47 ( 0.004 ) 33.6 ( 0.59 ) 16.0 ( 0.20 )
195 40 10 0.73 ( 0.009 ) 38.5 ( 0.76 ) 19.6 ( 0.24 )
195 40 10 0.76 ( 0.009 ) 38.6 ( 0.68 ) 20.1 ( 0.23 )
195 40 10 0.85 ( 0.012 ) 42.3 ( 0.83 ) 21.3 ( 0.28 )
195 40 30 0.58 ( 0.005 ) 35.5 ( 0.61 ) 18.0 ( 0.26 )
195 40 30 0.58 ( 0.006 ) 35.5 ( 0.64 ) 17.7 ( 0.27 )
195 40 30 0.60 ( 0.006 ) 36.0 ( 0.62 ) 18.2 ( 0.27 )
195 60 10 0.68 ( 0.008 ) 38.4 ( 0.69 ) 19.3 ( 0.34 )
195 60 10 0.70 ( 0.008 ) 38.6 ( 0.66 ) 19.2 ( 0.32 )
195 60 10 0.57 ( 0.006 ) 36.5 ( 0.67 ) 16.4 ( 0.31 )
195 60 30 0.47 ( 0.005 ) 33.4 ( 0.76 ) 16.4 ( 0.28 )
195 60 30 0.47 ( 0.005 ) 33.8 ( 0.69 ) 16.1 ( 0.26 )
195 60 30 0.40 ( 0.004 ) 34.1 ( 0.66 ) 15.2 ( 0.23 )

Value in parenthesis is the standard error of the average.


168

Table A.2: Data for all AS samples processed in the internal mixer
Temperature RPM Fiber concentration Lw Wn L/Dn

C wt.% mm m
180 40 10 0.87 ( 0.008 ) 34.9 ( 0.32 ) 24.9 ( 0.18 )
180 40 10 0.91 ( 0.009 ) 32.9 ( 0.28 ) 27.4 ( 0.18 )
180 40 10 0.76 ( 0.006 ) 32.6 ( 0.28 ) 22.8 ( 0.15 )
180 40 30 0.69 ( 0.005 ) 29.5 ( 0.26 ) 23.0 ( 0.12 )
180 40 30 0.64 ( 0.005 ) 29.9 ( 0.27 ) 21.8 ( 0.11 )
180 40 30 0.72 ( 0.006 ) 28.9 ( 0.26 ) 24.3 ( 0.13 )
180 60 10 0.74 ( 0.006 ) 29.0 ( 0.25 ) 24.4 ( 0.15 )
180 60 10 0.69 ( 0.005 ) 28.9 ( 0.24 ) 23.7 ( 0.17 )
180 60 10 0.72 ( 0.005 ) 31.1 ( 0.27 ) 23.4 ( 0.12 )
180 60 30 0.53 ( 0.005 ) 28.8 ( 0.41 ) 19.1 ( 0.14 )
180 60 30 0.61 ( 0.005 ) 27.9 ( 0.24 ) 21.6 ( 0.13 )
180 60 30 0.56 ( 0.004 ) 28.9 ( 0.31 ) 20.1 ( 0.12 )
195 40 10 0.79 ( 0.011 ) 29.3 ( 0.40 ) 26.8 ( 0.31 )
195 40 10 0.77 ( 0.010 ) 29.9 ( 0.40 ) 25.3 ( 0.30 )
195 40 10 0.74 ( 0.010 ) 28.7 ( 0.40 ) 25.6 ( 0.31 )
195 40 30 0.75 ( 0.008 ) 27.4 ( 0.27 ) 25.7 ( 0.25 )
195 40 30 0.71 ( 0.008 ) 28.1 ( 0.36 ) 25.0 ( 0.24 )
195 40 30 0.66 ( 0.007 ) 27.9 ( 0.29 ) 23.6 ( 0.22 )
195 60 10 0.67 ( 0.005 ) 28.3 ( 0.24 ) 23.4 ( 0.13 )
195 60 10 0.66 ( 0.006 ) 27.7 ( 0.24 ) 23.6 ( 0.14 )
195 60 10 0.62 ( 0.005 ) 28.2 ( 0.30 ) 21.8 ( 0.14 )
195 60 30 0.53 ( 0.005 ) 29.2 ( 0.43 ) 19.2 ( 0.19 )
195 60 30 0.50 ( 0.005 ) 28.8 ( 0.42 ) 19.3 ( 0.17 )
195 60 30 0.54 ( 0.005 ) 29.1 ( 0.38 ) 19.3 ( 0.19 )

Value in parenthesis is the standard error of the average.


169

Table A.3: Data for all WS samples processed in the internal mixer
Temperature RPM Fiber concentration Lw Wn L/Dn

C wt.% mm m
180 40 10 1.13 ( 0.021 ) 56.8 ( 1.27 ) 21.5 ( 0.48 )
180 40 10 1.11 ( 0.023 ) 63.5 ( 1.37 ) 20.6 ( 0.51 )
180 40 10 1.20 ( 0.027 ) 67.0 ( 1.54 ) 20.2 ( 0.46 )
180 40 30 0.74 ( 0.012 ) 46.5 ( 1.07 ) 20.0 ( 0.38 )
180 40 30 0.71 ( 0.011 ) 43.7 ( 1.02 ) 19.5 ( 0.38 )
180 40 30 0.67 ( 0.009 ) 42.2 ( 0.82 ) 19.2 ( 0.33 )
180 60 10 0.78 ( 0.014 ) 46.6 ( 1.13 ) 19.9 ( 0.48 )
180 60 10 0.90 ( 0.016 ) 55.9 ( 1.17 ) 19.4 ( 0.49 )
180 60 10 0.80 ( 0.013 ) 48.9 ( 1.04 ) 19.4 ( 0.43 )
180 60 30 0.55 ( 0.006 ) 38.5 ( 0.77 ) 18.9 ( 0.22 )
180 60 30 0.49 ( 0.006 ) 39.7 ( 0.87 ) 17.1 ( 0.23 )
180 60 30 0.53 ( 0.007 ) 39.2 ( 0.91 ) 18.2 ( 0.26 )
195 40 10 1.04 ( 0.022 ) 54.4 ( 1.27 ) 21.2 ( 0.69 )
195 40 10 1.15 ( 0.026 ) 60.1 ( 1.49 ) 22.9 ( 0.87 )
195 40 10 0.97 ( 0.020 ) 53.9 ( 1.29 ) 21.6 ( 0.78 )
195 40 30 0.73 ( 0.012 ) 35.2 ( 0.94 ) 23.7 ( 0.43 )
195 40 30 0.67 ( 0.009 ) 36.7 ( 0.81 ) 22.3 ( 0.35 )
195 40 30 0.71 ( 0.009 ) 37.7 ( 0.77 ) 22.4 ( 0.35 )
195 60 10 0.69 ( 0.013 ) 42.4 ( 1.05 ) 19.9 ( 0.56 )
195 60 10 0.72 ( 0.015 ) 40.0 ( 1.19 ) 21.2 ( 0.60 )
195 60 10 0.65 ( 0.012 ) 42.1 ( 1.09 ) 17.8 ( 0.54 )
195 60 30 0.40 ( 0.006 ) 31.4 ( 1.05 ) 17.9 ( 0.28 )
195 60 30 0.51 ( 0.006 ) 36.6 ( 0.80 ) 17.8 ( 0.25 )
195 60 30 0.50 ( 0.006 ) 35.1 ( 0.77 ) 17.7 ( 0.25 )

Value in parenthesis is the standard error of the average.


170

Appendix B Chapter 4 processed samples data

Table B.4: Lw (mm) data of AL samples processed in the capillary rheometer


16-mm length die zero-length die
500 s1 1500 s1 500 s1 1500 s1

10 wt.% fiber
190 C 0.902 ( 0.008 ) 0.778 ( 0.008 ) 0.959 ( 0.010 ) 0.905 ( 0.008 )
0.833 ( 0.009 ) 0.861 ( 0.010 ) 0.878 ( 0.011 ) 0.870 ( 0.009 )
0.960 ( 0.011 ) 0.888 ( 0.009 ) 0.870 ( 0.009 ) 0.966 ( 0.010 )
200 C 0.762 ( 0.008 ) 0.673 ( 0.008 ) 0.991 ( 0.011 ) 0.779 ( 0.010 )
0.805 ( 0.008 ) 0.703 ( 0.007 ) 0.937 ( 0.009 ) 0.785 ( 0.008 )
0.856 ( 0.009 ) 0.786 ( 0.009 ) 0.954 ( 0.011 ) 0.750 ( 0.008 )
210 C 0.607 ( 0.006 ) 0.724 ( 0.007 ) 0.990 ( 0.008 ) 0.998 ( 0.009 )
0.631 ( 0.008 ) 0.751 ( 0.008 ) 0.896 ( 0.010 ) 1.039 ( 0.011 )
0.603 ( 0.009 ) 0.658 ( 0.008 ) 0.936 ( 0.009 ) 0.940 ( 0.012 )

30 wt.% fiber
190 C 0.542 ( 0.006 ) 0.529 ( 0.007 ) 0.567 ( 0.006 ) 0.531 ( 0.006 )
0.543 ( 0.005 ) 0.532 ( 0.006 ) 0.539 ( 0.006 ) 0.564 ( 0.006 )
0.531 ( 0.007 ) 0.511 ( 0.006 ) 0.518 ( 0.006 ) 0.569 ( 0.007 )
200 C 0.538 ( 0.006 ) 0.535 ( 0.006 ) 0.658 ( 0.009 ) 0.532 ( 0.006 )
0.492 ( 0.005 ) 0.481 ( 0.006 ) 0.625 ( 0.008 ) 0.579 ( 0.006 )
0.518 ( 0.006 ) 0.519 ( 0.006 ) 0.649 ( 0.010 ) 0.603 ( 0.005 )
210 C 0.467 ( 0.006 ) 0.434 ( 0.005 ) 0.466 ( 0.005 ) 0.456 ( 0.005 )
0.462 ( 0.005 ) 0.452 ( 0.005 ) 0.484 ( 0.006 ) 0.506 ( 0.005 )
0.489 ( 0.006 ) 0.458 ( 0.006 ) 0.412 ( 0.005 ) 0.417 ( 0.005 )

Value in parenthesis is the standard error of the average.


171

Appendix C Chapter 5 processed samples data

Table C.5: Data for all AL samples processed in the twin-screw extruder
Screws RPM PP feed (RPM) Fiber feed (RPM) Temperature Lw

C mm

120 6 60 190 0.78 ( 0.013 )


120 6 60 190 0.75 ( 0.011 )
120 6 60 190 0.77 ( 0.009 )
200 6 60 190 0.61 ( 0.006 )
200 6 60 190 0.65 ( 0.009 )
200 6 60 190 0.64 ( 0.007 )
280 6 60 190 0.76 ( 0.008 )
280 6 60 190 0.65 ( 0.012 )
280 6 60 190 0.77 ( 0.008 )
120 9 60 190 0.90 ( 0.009 )
120 9 60 190 0.94 ( 0.008 )
120 9 60 190 0.96 ( 0.011 )
200 9 60 190 0.89 ( 0.008 )
200 9 60 190 0.93 ( 0.006 )
200 9 60 190 0.93 ( 0.006 )
280 9 60 190 0.62 ( 0.007 )
280 9 60 190 0.64 ( 0.010 )
280 9 60 190 0.63 ( 0.008 )
120 9 15 190 1.10 ( 0.009 )
120 9 15 190 1.11 ( 0.013 )
120 9 15 190 1.08 ( 0.011 )
200 9 15 190 0.90 ( 0.007 )
200 9 15 190 0.94 ( 0.009 )
200 9 15 190 1.01 ( 0.008 )
280 9 15 190 0.87 ( 0.007 )
280 9 15 190 0.88 ( 0.013 )
280 9 15 190 0.92 ( 0.012 )
Continued on next page

Value in parenthesis is the standard error of the average.


172

Table C.5 continue from previous page


Screws RPM PP feed (RPM) Fiber feed (RPM) Temperature Lw

C mm

120 6 40 190 0.89 ( 0.012 )


120 6 40 190 0.91 ( 0.016 )
120 6 40 190 0.92 ( 0.013 )
200 6 40 190 0.88 ( 0.009 )
200 6 40 190 0.92 ( 0.010 )
200 6 40 190 0.88 ( 0.012 )
280 6 40 190 0.68 ( 0.009 )
280 6 40 190 0.71 ( 0.011 )
280 6 40 190 0.70 ( 0.007 )
120 9 60 200 0.84 ( 0.009 )
120 9 60 200 0.88 ( 0.009 )
120 9 60 200 0.80 ( 0.008 )
200 9 60 200 1.02 ( 0.010 )
200 9 60 200 0.94 ( 0.013 )
200 9 60 200 0.99 ( 0.012 )
300 9 60 200 0.86 ( 0.008 )
300 9 60 200 0.85 ( 0.011 )
300 9 60 200 0.84 ( 0.009 )

Value in parenthesis is the standard error of the average.


173

Table C.6: Data for all AS samples processed in the twin-screw extruder
Screws RPM PP feed (RPM) Fiber feed (RPM) Temperature Lw

C mm

120 6 40 190 0.96 ( 0.007 )


120 6 40 190 0.98 ( 0.009 )
120 6 40 190 0.97 ( 0.011 )
120 6 40 200 0.96 ( 0.007 )
120 6 40 200 1.03 ( 0.011 )
120 6 40 200 1.02 ( 0.010 )
120 6 40 210 0.94 ( 0.008 )
120 6 40 210 1.01 ( 0.007 )
120 6 40 210 1.04 ( 0.006 )
120 3 40 190 0.85 ( 0.005 )
120 3 40 190 0.89 ( 0.005 )
120 3 40 190 0.86 ( 0.007 )
120 3 40 200 0.84 ( 0.007 )
120 3 40 200 0.88 ( 0.007 )
120 3 40 200 0.79 ( 0.009 )
120 3 40 210 0.89 ( 0.010 )
120 3 40 210 0.91 ( 0.013 )
120 3 40 210 0.94 ( 0.009 )
120 6 10 190 0.99 ( 0.009 )
120 6 10 190 0.89 ( 0.008 )
120 6 10 190 0.93 ( 0.011 )
120 6 10 200 0.82 ( 0.008 )
120 6 10 200 0.95 ( 0.011 )
120 6 10 200 0.85 ( 0.007 )
120 6 10 210 0.80 ( 0.008 )
120 6 10 210 0.84 ( 0.007 )
120 6 10 210 0.82 ( 0.009 )
Continued on next page

Value in parenthesis is the standard error of the average.


174

Table C.6 continue from previous page


Screws RPM PP feed (RPM) Fiber feed (RPM) Temperature Lw

C mm

80 3 40 190 0.87 ( 0.010 )


80 3 40 190 0.91 ( 0.011 )
80 3 40 190 0.86 ( 0.008 )
80 3 40 200 0.79 ( 0.008 )
80 3 40 200 0.91 ( 0.014 )
80 3 40 200 0.92 ( 0.007 )
80 3 40 210 0.83 ( 0.008 )
80 3 40 210 0.86 ( 0.008 )
80 3 40 210 0.85 ( 0.005 )
80 6 40 190 1.03 ( 0.012 )
80 6 40 190 1.03 ( 0.008 )
80 6 40 190 1.05 ( 0.011 )
80 6 40 200 0.95 ( 0.010 )
80 6 40 200 0.96 ( 0.009 )
80 6 40 200 0.98 ( 0.008 )
80 6 40 210 1.08 ( 0.007 )
80 6 40 210 1.09 ( 0.009 )
80 6 40 210 1.19 ( 0.015 )
80 6 10 190 1.05 ( 0.005 )
80 6 10 190 1.03 ( 0.006 )
80 6 10 190 1.02 ( 0.009 )
80 6 10 200 1.00 ( 0.011 )
80 6 10 200 0.90 ( 0.007 )
80 6 10 200 0.93 ( 0.010 )
80 6 10 210 0.99 ( 0.007 )
80 6 10 210 1.05 ( 0.007 )
80 6 10 210 1.08 ( 0.008 )

Value in parenthesis is the standard error of the average.


175

Appendix D Listing of a section of the source code

1 %###########################################################################
2 % L O O P S E C T I O N HERE THE BEGINNING OF THE LOOPS ARE SET
3 %###########################################################################
4 % ========> LOOP FOR THE TYPE OF BREAKAGE BEGIN
5 % There are 3 types of breakage: type 1: Breakage in half
6 % type 2: Breakage one third
7 % type 3: Breakage random
8 for breakage type = 1:1:3
9 if breakage type == 1
10 breakage directory = ['Breakage in half ',replicate];
11 elseif breakage type == 2
12 breakage directory = ['Breakage one third ',replicate];
13 else
14 breakage directory = ['Breakage random ',replicate];
15 end
16

17 % LOOP FOR THE TEMPERATURE BEGIN


18 for temperature = vector temperature
19 if temperature == 180
20 temperature directory = '180';
21 temperature index sim = 0;
22 else
23 temperature directory = '195';
24 temperature index sim = 4;
25 end
26

27 % LOOP FOR THE RPM BEGIN


28 for RPM = vector RPM
29 if RPM == 40
30 RPM directory = '40';
31 RPM index sim = 0;
32 else
33 RPM directory = '60';
176

34 RPM index sim = 1;


35 end
36

37 % Calculation of shear rate


38 shear rate = pi * 2*radius blade * RPM /...
39 (high shear zone height * 60);
40

41 % Total number of revolutions of blade 1, which is the one attached


42 % directly to the shaft
43 revolutions = RPM * mixing time;
44 % creation of the vector that is going to be used for the loop
45 total revolutions = initial RPM:increment RPM:revolutions;
46

47

48 % LOOP FOR FIBER MASS CONCENTRATION BEGIN


49 for fiber mass concentration = vector fiber mass concentration
50 if fiber mass concentration == 10
51 fiber mass concentration directory = '10';
52 fiber mass concentration index sim = 1;
53 else
54 fiber mass concentration directory = '30';
55 fiber mass concentration index sim = 3;
56 end
57

58 %####################################################################
59 %% S E C T I O N 3 [CREATION OF VARIABLES] B E G I N
60 %####################################################################
61 % Create the FINAL TEMPERATURE variable.......... Begin
62 Final Temperature row = fiber mass concentration index sim...
63 + RPM index sim + temperature index sim;
64 conditions vector = [fiber mass concentration ...
65 RPM temperature];
66

67 if fiber type == 1
68 % FTM [=] Final Temperature Matrix
69 conditions vector FTM = AL Final Temperature...
70 (Final Temperature row,1:3);
177

71

72 if sum(conditions vector == conditions vector FTM) = 3


73 error('myToolbox:myFunction:fileNotFound',...
74 'The conditions vector is not correct')
75 end
76

77 Final Temperature = AL Final Temperature...


78 (Final Temperature row,4);
79

80 elseif fiber type == 2


81 % FTM [=] Final Temperature Matrix
82 conditions vector FTM = AS Final Temperature...
83 (Final Temperature row,1:3);
84

85 if sum(conditions vector == conditions vector FTM) = 3


86 error('myToolbox:myFunction:fileNotFound',...
87 'The conditions vector is not correct')
88 end
89

90 Final Temperature = AS Final Temperature...


91 (Final Temperature row,4);
92

93 else
94 % FTM [=] Final Temperature Matrix
95 conditions vector FTM = WS Final Temperature...
96 (Final Temperature row,1:3);
97

98 if sum(conditions vector == conditions vector FTM) = 3


99 error('myToolbox:myFunction:fileNotFound',...
100 'The conditions vector is not correct')
101 end
102

103 Final Temperature = WS Final Temperature...


104 (Final Temperature row,4);
105 end
106

107 % Create the FINAL TEMPERATURE variable.......... End


178

108 % Create the variable 'temperature increment' this variable


109 % will indicate the increment of temperature with every revolution.
110 temperature increment = (Final Temperature ...
111 temperature) / (revolutions 1);
112

113 % Volume of each sample lot


114 sample volume = total mixer capacity* degree of fill/100;
115 polymer mass concentration = 100fiber mass concentration;
116 suspension density = ((fiber mass concentration + ...
117 polymer mass concentration) * fiber density...
118 * polymer density)/ (polymer mass concentration...
119 * fiber density + fiber mass concentration...
120 * polymer density);
121

122 % Volume of fibers in the suspension or sample volume.


123 fiber suspension volume = (suspension density polymer density)...
124 * sample volume / (fiber density polymer density);
125

126 % ####################################################################
127 % S E C T I O N 3 [CREATION OF VARIABLES] E N D
128 % ####################################################################
129

130 %######################################################################
131 % S E C T I O N 4 [CREATION OF FIBERS IN SUSPENSION MATRIX] B E G I N
132 % #####################################################################
133 % Residual volume (mm3). The initial residual volume is the total
134 % volume of fibers in suspension
135 Vr = fiber suspension volume;
136

137 % Preallocation of the 'fibers in suspension' matrix


138 % STEP A) Allocation of the 'fibers in suspension' matrix
139 clear fibers in suspension
140 fibers in suspension = zeros(30e6,11) .* NaN;
141

142 % Set a counter and initialize it.


143 counter fibers in suspension = 0;
144
179

145 while(Vr > 0)


146 counter fibers in suspension = ...
147 counter fibers in suspension + 1;
148

149 % Select a row randomly from 'fibers to mix' matrix


150 r = floor(rand*TotalRows)+1;
151

152 % Select the fiber information (row) from


153 %the matrix 'fibers to mix'
154 fiber info = fibers to mix(r,:);
155

156 % fiber volume of the last fiber added to


157 % 'fibers in suspension' matrix
158 fiber volume = pi*fiber info(1,2)2*fiber info(1,1)/4;
159

160 % Row of values to input in the


161 % fiber in suspension matrix
162 row fiber info = [fiber info,0,fiber volume];
163

164 % Store the fiber information in the matrix


165 % 'fibers in suspension'
166 % column 1: length (mm). column 2: width (mm).
167 % column 3: aspect ratio. column 8: fiber volume
168 fibers in suspension(counter fibers in suspension,[1:4,8]) = row fiber info;
169

170 % Update value of the Residual volume:


171 % Residual volume (Vr) = Residual volume
172 % volume of last fiber added
173 Vr = Vr fiber volume;
174 end
175

176 % Store the (Col 11) fiber modulus information


177 eval(['fibers in suspension(1:counter fibers in suspension,11)...
178 = rand(counter fibers in suspension,1) .*' interval name...
179 '+' lower level name ';']);
180

181 % Store (Col 7) the critical shear stress to fiber modulus ratio
180

182 fibers in suspension(1:counter fibers in suspension,7) = ...


183 0.46 3.559 .* log10(fibers in suspension...
184 (1:counter fibers in suspension,3));
185

186 fibers in suspension copy = fibers in suspension...


187 (1:counter fibers in suspension,1:3);
188

189 % #########################################################################
190 % S E C T I O N 4 [CREATION OF FIBERS IN SUSPENSION MATRIX] E N D
191 % #########################################################################
192

193 %########################################################################
194 %% S E C T I O N 5 [BREAKAGE DURING MIXING] B E G I N
195 %########################################################################
196 % Volume that passes through one high shear zone after one rotation
197 % The high shear zone runs through 80% of each chamber,
198 high shear zone one rotation volume = 0.8 * pi*...
199 (radius chamber2 radius blade2) *...
200 high shear zone length;
201

202 % Blade 1 has 3 high shear zones.


203 blade 1 high shear volume = 3 *...
204 high shear zone one rotation volume;
205

206 % Blade 2 has 3 high shear zones and they cover 2/3 of the volume
207 blade 2 high shear volume = 2 * ...
208 high shear zone one rotation volume;
209

210 % Total high shear volume for one RPM


211 one rotation high shear volume = ...
212 blade 1 high shear volume + ...
213 blade 2 high shear volume;
214

215 % calculation of the volume of fibers that are high sheared


216 % after one rotation.
217 % Variable name: 'Vfhs' [=]
218 %volume of fibers that pass the high shear zone
181

219 Vfhs = (suspension density polymer density) * ...


220 one rotation high shear volume / ...
221 (fiber density polymer density);
222

223 % Breakage begins


224 % 'rows fibers in suspension' indicates the number
225 % of rows with information
226 rows fibers in suspension = counter fibers in suspension;
227

228 % initialize the current temperature variable


229 current temperature = temperature;
230

231 for k = total revolutions


232 %****************************************************
233 % Begin Calculation of the data that depend on the
234 % temperature change
235

236 % Calculation of the shift factor, a T


237 T0 K = T0 + 273.15;
238 temperature K = current temperature + 273.15;
239 a T = exp(E/R*(1/temperature K 1/T0 K));
240

241 % Calculation of eta


242 eta=a T *eta zero*shear rate...
243 (power law index1);
244

245 % Calculation of shear stress


246 shear stress = eta * shear rate;
247

248 % calculation of the current fiber modulus


249 eval(['interpolation row = find(' TGA variable '...
250 (:,1) <= current temperature);']);
251

252 interpolation lower limit = interpolation row(end);


253

254 eval(['x1 = ' TGA variable '(interpolation lower limit,1);']);


255 eval(['y1 = ' TGA variable '(interpolation lower limit,2);']);
182

256 eval(['x2 = ' TGA variable '(interpolation lower limit + 1,1);']);


257 eval(['y2 = ' TGA variable '(interpolation lower limit + 1,2);']);
258

259 new y = (current temperature x1) * (y2 y1) / (x2 x1) + y1;
260 percent decrease = (initial weight new y) / initial weight;
261

262 % (Col 5) Current fiber modulus


263 fibers in suspension(1:rows fibers in suspension,5) =...
264 (1 percent decrease) .* fibers in suspension...
265 (1:rows fibers in suspension,11);
266

267 % (Col 6) Log10 of the shear stresstocurrent fiber modulus ratio:


268 fibers in suspension(1:rows fibers in suspension,6) = ...
269 log10(shear stress ./ fibers in suspension...
270 (1:rows fibers in suspension,5));
271

272 % End Calculation of the data that depend on the


273 % temperature change
274 %****************************************************
183

You might also like