You are on page 1of 206

ME480/580: Materials Selection

Lecture Notes for Week One


Winter 2012

MATERIALS SELECTION IN THE DESIGN


PROCESS
Reading: Ashby Chapters 1, 2, and 3.
Reference: Kenneth G. Budinski, Engineering Materials: Properties and Selection
Fifth edition, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1996.

HISTORICAL CONTEXT
Ashby does a nice job of setting the historical context of the development of materials
over the years with the cover illustration for Chapter One.
Note: Change from NATURAL materials (on left) toward MANUFACTURED materials
(on right) toward ENGINEERED materials (near future). We have an increasingly large
number of materials to deal with, on the order of 160,000 at present!
Other books for general reading on the history and development of materials science and
engineered materials are:
J. E. Gordon, The New Science of Strong Materials, or Why You Don't Fall Through the
Floor, Princeton University Press, Princeton, NJ.
M. F. Ashby and D. R. H Jones, Engineering Materials Parts 1, 2, and 3, Pergamon
Press, Oxford, UK.
F. A. A. Crane and J. A. Charles, Selection and Use of Engineering Materials,
Butterworths, London, UK.
P. Ball, Made to Measure: New Materials for the 21st Century, Princeton University
Press, 1997.

MATERIALS PROPERTIES
Before we can discuss the appropriate selection of materials in design, we have to have a
foundation of what we mean by "materials properties". Both Budinski and Ashby provide
lists of these in the texts. For example:

Week One: Page 1

W. Warnes: Oregon State University

This is a pretty complete list. Budinski discusses these in his chapter 2, and Ashby has his
own definitions and discussion in chapter 3 in which he breaks the materials down into
six categories: METALS, POLYMERS, CERAMICS, ELASTOMERS (which Budinski
groups with plastics), GLASSES (which Budinski groups with ceramics), and HYBRIDS
(or composite materials). In all, about 120,000 different materials with property values
ranging over 5 orders of magnitude!
The importance of these chapters is that unless you have a clear idea of how a property
value is measured (see Homework One), you cannot properly use the property for
calculations in mechanical design. To these properties, we will add two other important
materials properties: PRICE, and EMBODIED ENERGY (and other environmental
materials properties.)
MATERIALS IN THE DESIGN PROCESS
Different authors have different ideas about how the design process should work.
Budinski's design strategy is found in figure 18-1.

Week One: Page 2

W. Warnes: Oregon State University

BUDINSKI FIGURE 18.1 NOTES:


1. Calculations in the first block! Analysis is important! (NOTE: ALGEBRA is a
critical skill for success in this class, as is UNIT ANALYSIS. Youve been
warned!)
2. Analysis appears multiple times throughout design process.
3. Materials selection is in the last step.
4. Iteration?

Week One: Page 3

W. Warnes: Oregon State University

In my opinion, Ashby uses a better design strategy, especially in terms of materials


selection. He breaks the design flow path into three stages, called CONCEPTUAL
DESIGN, EMBODIMENT DESIGN, and DETAIL DESIGN (see figure 2.1).
CONCEPTUAL DESIGN:
All options are kept open.
Consideration of alternate working principles.
Assess the functional structure of your design.
EMBODIMENT DESIGN:
Use the functional structure to ANALYSE the operation.
Sizing of components.
Materials down-selection.
Determination of operational conditions.
DETAIL DESIGN:
Specifications written for components.
Detailed analysis of critical components.
Production route and cost analysis performed.

How does materials selection enter into Ashby's process? (Figure 2.5)
Materials selection enters at EVERY STAGE, but with differing levels of
CONSTRAINT and DETAILED INFORMATION.
CONCEPTUAL DESIGN:
Apply PRIMARY CONSTRAINTS (eg. working temperature, environment, etc.).
(Budinski figure 18-2 has a good list of primary constraints to consider.)
100% of materials in, 10-20% candidates come out.
EMBODIMENT DESIGN:
Develop and apply optimization constraints.
Need more detailed calculations and
Need more detailed materials information.
10-20% of materials in, 5 candidate materials out.

Week One: Page 4

W. Warnes: Oregon State University

DETAILED DESIGN:
High degree of information needed about only a few materials.
May require contacting specific suppliers of materials.
May require specialized testing for critical components if materials data does not
already exist.
CLASS APPROACH
Two different philosophies have been presented here:
Budinski: get familiar with a set of basic materials from each category, about seventyfive in total, and these will probably handle 90% of your design needs (see Figure 18-8).
Ashby: look at all 160,000 materials initially, and narrow your list of candidate materials
as the design progresses using some technique to narrow your choices.
Materials selection has to include not only properties, but also SHAPES (what standard
shapes are available, what shapes are possible), and PROCESSING (what fabrication
route can or should be used to produce the part or raw material, eg. casting, injection
molding, extrusion, machining, etc.). It can also include ENVIRONMENTAL IMPACT.
The point is that the choice of materials interacts with everything in the engineering
design and product manufacturing process (see Ashby Figure 2.6).
In the remainder of this course we will develop a systematic approach to dealing with all
these interactions and with looking at the possibilities of all 160,000 of these materials
based on the use of MATERIALS SELECTION CHARTS as developed by Ashby.
Flow of the course:

Optimization of selection without considering shape effects.


Optimization under multiple constraints.
Optimization of selection considering shape effects.
Considerations of environmental impact.
Optimization of material process selection.

SELECTION CHARTS (Ashby chapter 4)


1. Materials don't exhibit single-valued properties, but show a range of properties, even
within a single production run (see Ashby, Figure 4.1 for example.)
EXAMPLES: The elastic modulus of copper varies over a few percent
depending on the purity, texture, grain size, etc.
Week One: Page 5

W. Warnes: Oregon State University

The mechanical strength of alumina (Al2O3) varies by more than a factor


of 100 depending on its porosity, grain size, etc.
Metal alloys show large changes in their mechanical and electrical
properties depending on the heat treatments and mechanical working they
have experienced.
NOTE: Because the properties of materials may vary over large ranges, it
will be critical to be able to interpret property data using SEMI-LOG and
LOG-LOG plots. If you arent comfortable with logarithmic math and
making and reading log axes on plots REVIEW IT!
2. Performance is seldom limited by only ONE property.
EXAMPLE: in lightweight design, it is not just strength that is important,
but both strength and density. We will need to be able to compare
materials based on several properties at once.
Because of these facts, we can produce charts such as this selection chart from Ashby:

There is a tremendous amount of information and power in these charts. First of all, they
provide the materials property data as "balloons" in an easy to compare form. Secondly,
other physical information can be displayed on these charts.

Week One: Page 6

W. Warnes: Oregon State University

EXAMPLE: the longitudinal wavespeed of sound in a material is given by the equation

" E%
V =$ '
# !&
log (V ) =

1/2

Rewrite this equation by taking the base-10 logarithm of both sides to get:
1
# log ( E ) ! log ( " ) %& or log ( E ) = 2 log (V ) + log ( " ) .
2$

This is an equation of the form Y = A + BX, where:


Y = log(E),
A = constant = 2log(V) = y-axis intercept at X = 0,
B = slope = 1, and
X = log().
This appears as a line of slope = 1 on a plot of log(E) versus log(). Such a line connects
materials that have the same speed of sound (constant V).
NOTE: X = 0 means what for the value of density?

Week One: Page 7

W. Warnes: Oregon State University

EXAMPLE: The selection requirement for a particular minimum weight design (derived
next time) is to maximize the ratio of

E 1/2
1
= C = constant, which leads to log ( E ) = log ( C ) + log ( ! ) , or
!
2

log ( E ) = 2 log ( C ) + 2 log ( ! ) ,


Y

+ BX.

This is a straight line of slope = 2 on a plot of log(E) versus log().


Such a line connects materials that will perform the same in a minimum weight design,
that is, all the materials on this line have the same value of the constant, C.

End of File.

Week One: Page 8

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week One
Winter 2012

PERFORMANCE INDICES
Reading: Ashby Chapters 4 and 5.

Materials Selection begins in conceptual design by using PRIMARY CONSTRAINTS non-negotiable constraints on the material imposed by the design or environment.
Examples might include "must be thermally insulating", or "must not corrode in
seawater".
These take the form of "PROPERTY > PROPERTYcritical", and appear as horizontal or
vertical lines on the selection charts.

NOTE: Don't go overboard on primary constraints. They are the easiest to apply and
require the least thought and analysis, but they can often be engineered around, for
example, by active cooling of a hot part, or adding corrosion resistant coatings.
After initial narrowing, you should develop PERFORMANCE INDICES.
DEFINITIONS:
PERFORMANCE:

Week One: Page 9

W. Warnes: Oregon State University

OBJECTIVE FUNCTION:

CONSTRAINT:

PERFORMANCE INDEX:

In the following, we will assume that performance (P) is determined by three factors:
FUNCTIONAL REQUIREMENTS (carry a load, store energy, etc.)
GEOMETRICAL REQUIREMENTS (space available, shape, size)
MATERIALS PROPERTIES

What we want to do is OPTIMIZE our choice of materials to maximize the performance


of the design subject to the constraints imposed on it, so that we will try to make
P ! Pmax .
We will further assume that these three factors are SEPARABLE, so that the performance
equation can be written as:

If this is true, then maximizing performance will be accomplished by independently


maximizing the three functions fn1, fn2, and fn3.
fn1 is the place where creative design comes in.

Week One: Page 10

W. Warnes: Oregon State University

fn2 is where geometry can make a difference.

fn3 is the part we're most interested in. When the factors are separable, the materials
selection doesn't depend on the details of fn1 or fn2! This means we don't have to know
that much about the design to make intelligent materials choices.
Our first step in this class will be to maximize performance by only considering fn3
(selection of materials without shape effects). Later on we'll look at adding in the effect
of shape on performance by maximizing the product of fn2 x fn3.
EXAMPLE ONE: Design a light, strong tie rod.
The design requirements are:

to be a solid cylindrical tie rod


length L
load F, which may include a safety factor
minimum mass

Let's start by doing this the "old" way:

OLD WAY: PART ONE


1) CALCULATIONAL MODEL to use in the analysis
(pretty simple for this example).

2) We know an equation for the failure strength of a tie rod:

Week One: Page 11

W. Warnes: Oregon State University

We know F, and we can always look up f, so we can find the right cross sectional area,
A. In the past, this part has always been made in our company from STEEL, which we
know if a good high strength material, so we can look up in a database somewhere the
f(steel). Now we know what the smallest area will be:

3) And now we can find the mass of the rod, the measure of performance:

From our analysis, we can see that by choosing a higher strength steel, we can use a
smaller A and thereby reduce our mass. Our recommendation: use a high strength steel.

OLD WAY: PART TWO


A new engineer comes along and she says "Wait...the design constraint says minimum
mass, and your analysis shows that we can ALSO lower the mass by going to a lower
density material. Let's use a high strength Al alloy instead of steel."

From the materials properties I found that the MASS (Aluminum) / MASS (Steel) = 60%
which means we can get an increase in performance of 1.67 times. Our recommendation:
use a high strength aluminum.
What's wrong with these two approaches? Nothing really. They both rely on established
tradition in the company, and the use of "comfortable" materials. They both also
ASSUME a material essentially at the outset.
Week One: Page 12

W. Warnes: Oregon State University

ASHBY APPROACH (LINEAR OPTIMIZATION THEORY)


Looking at this list of requirements, we start with
1) CALCULATIONAL MODEL to use in the analysis

2) Determine the MEASURE OF PERFORMANCE (MOP), P.


In this case, we have been told that the goal is to get a part that has a minimum mass.

NOTE: P is defined so that the larger it is the better our performance; we want to
maximize P. This is our OBJECTIVE FUNCTION. NOTE also that Ashby defines P to
be either minimized or maximized, just so long as you keep track of which one it is. We
could also write the MOP as

3) IDENTIFY the parameters in our analytical model and MOP:


L=
A=
F=
=
4) Write an equation for the CONSTRAINED variables: (we have to safely carry the load
F)

Week One: Page 13

W. Warnes: Oregon State University

5) Rewrite the constraint equation for the free variable and substitute this into the MOP:

6) Regroup into the three functional groups fn1, fn2, and fn3

To maximize P we want to choose a material that maximizes the ratio


#!f &
%$ " (' = M = MATERIALS PERFORMANCE INDEX.

NOTE: We don't need to know anything about F, or A, to choose the best material for the
job!
EXAMPLE TWO: Design a light, stiff column.
The design requirements are:

slender cylindrical column


length L fixed
compressive load F
minimum mass

1) CALCULATIONAL MODEL to use in the analysis

Week One: Page 14

W. Warnes: Oregon State University

2) MEASURE OF PERFORMANCE (MOP), P.


Minimum mass again.

3) IDENTIFY the parameters in our analytical model and MOP:


L=
A=
F=
=
4) CONSTRAINT equation: (no Euler buckling of this column)

where n is a constant that depends on the end conditions, and E is the Youngs Modulus.
(NOTE: There are a number of convenient mechanics equations in the appendix in the
back of Ashby, appendix B, which I will use almost exclusively. You may use any
analytical equations you like as long as you understand them!)
5) Rewrite the constraint equation for the free variable and substitute this into the MOP:

Week One: Page 15

W. Warnes: Oregon State University

6) Regroup into fn1, fn2, and fn3

7) PERFORMANCE INDEX = MMAX =

RECIPE FOR OPTIMIZATION


1) Clearly write down the design assignment/goal.
2) Identify a model to use for calculations.
3) Determine the measure(s) of performance with an equation (weight, cost,
energy content, stiffness, etc.)
4) Identify the FREE, FIXED, PROPERTY, and CONSTRAINT parameters.
5) Develop an equation for the constraint(s).
6) Solve the CONSTRAINT equation for the FREE parameters and substitute into
the MOP.
7) Reorganize into the fn1, fn2, fn3 functions to find M.
NOTES:
i) M is always defined to be maximized in order to maximize performance.
ii) A full design solution is not needed to find M! You can do a lot of materials
optimization BEFORE your design has settled into specifics.
End of File.

Week One: Page 16

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week One
Winter 2012

MATERIALS OPTIMIZATION WITHOUT SHAPE


Reading: Ashby Chapter 5, 6.

RECIPE FOR OPTIMIZATION


1) Clearly write down the design assignment/goal.
2) Identify a model to use for calculations.
3) Determine the measure(s) of performance with an equation (weight,
cost, energy content, stiffness, etc.)
4) Identify the FREE, FIXED, PROPERTY, and CONSTRAINT params.
5) Develop an equation for the constraint(s).
6) Solve the CONSTRAINT equation for the FREE parameters in the
MOP.
7) Reorganize into fn1, fn2, fn3 functions to find M.
NOTES:
i) M is always defined to be maximized in order to maximize
performance.
ii) A full design solution is not needed to find M! You can do a lot of
materials optimization BEFORE your design has settled into specifics.
EXAMPLE THREE: Mirror support for a ground based telescope. Typically these
have been made from glass with a reflective coating--the glass is used only as a stiff
support for the thin layer of silver on the top surface. Most recent telescopes have
diameters in the 8-10 m range, and are typically limited by the mirror being out of
position by more than one wavelength of the light it is reflecting (). The design
requirements are that the mirror be large, and that it not sag under it's own weight by
more than 1- when simply supported. Since the mirror will need to be moved around to
point it in the right direction, it needs to be very light weight.
DESIGN ASSIGNMENT:

Circular disk shaped mirror support


Size = 2r
lightweight
deflects () under own weight by less than .

Week One: Page 17

W. Warnes: Oregon State University

MODEL:

MOP: minimize mass

PARAMETERS:
r=
t=
=
=
NOTE: For clearly explaining this design process to others, lets start using a table to list
the design requirements. This will become much more important to you in the design
project at the end of the term, so start practicing it now.
Design Requirements for Telescope Mirror Supports
Function
Support reflective surface for ground-based optical telescope
Constraints
Radius, r, specified (and large!)
Stiff enough to not deform under own weight by more than
Objective
Minimize mass
Free Parameters
Thickness, t.

Week One: Page 18

W. Warnes: Oregon State University

CONSTRAINT EQUATION: (use the helpful solutions in appendix B of Ashby or a


mechanics textbook)
For a simply supported disk under its own weigh, the center deflection is:

(NOTE: The less-than-or-equal sign is a good way to identify a constraint parameter.)

APPLY TO MOP: Solve for the free parameter, t.

CAUTION: m (mass) appears in the constraint equation, so its now on both sides of the
equation. We have to solve for m so we can plug it into the objective function, P.

Since the MOP is minimum mass, Pmax =

Week One: Page 19

1
:
m

W. Warnes: Oregon State University

E
E 1/ 3
; maximizing any of these will
NOTE: This is equivalent to an M = 3 , or M =
!
!
maximize our performance. For reasons that will become clear later, it is always best to
use the performance index that comes directly from the optimization analysis, in this

" E%
case, M = $ 3 '
#! &

1/2

APPLYING PERFORMANCE INDICES TO SELECTION CHARTS


Use the telescope mirror support as an example. We use Ashby's CHART 1 (E versus ).
We could apply PRIMARY CONSTRAINTS and say that, in order for the design to
work, the modulus must be E > 20 [GPa], and the density, < 2 [Mg/m3].

Our selection region will be in the upper left, and we end up with expensive candidate
materials such as CFRP.

Week One: Page 20

W. Warnes: Oregon State University

" E%
OR: we can use the performance index from above: M = $ 3 '
#! &

1/2

E 1/2
= 3/2 :
!

which gives us a line of slope = 3 on a log(E) versus log() plot (Chart 1).
How do we plot this on the chart? Start with an X-Y point, say
X = log(density) = log() = log(0.1 [Mg/m3]) = -1
Y = log(modulus) = log(E) = log(0.1 [GPa]) = -1
Now, for every decade unit in X we go up three decade units in Y (slope = 3).
one unit in X gives X = log() = 0, which gives = 1.0 [Mg/m3], and
three units in Y gives Y = log(E) = 2, which gives E = 100 [GPa].
This is a line of slope = 3. Ashby helps us out with some guide lines for common design
criteria.

Week One: Page 21

W. Warnes: Oregon State University

NOTES:
1) This line connects materials with the SAME PERFORMANCE INDEX for this design
(same value of M).
What are the units of the performance index? It will be different for every design
situation, but for the telescope example:

Let's just use M =

[GPa ]1/2

. It will always be easiest to leave the units of the


3/2
! Mg $
#"
m 3 &%
performance index in the scale units of the plot of materials properties.
Look at our line-- it passes through the point E = 0.1 [GPa], = 0.1 [Mg/m3].

It also passes through the point


E = 100 [GPa,] = 1 [Mg/m3].

This means that all materials on this line will perform the same, and should be considered
as equal candidates for the job.

Week One: Page 22

W. Warnes: Oregon State University

2) As we move the line (keeping the slope the same), we change the value of the
performance index (M), and thus the PERFORMANCE of the material in this design. For
example, if we move the line to the lower right to the point
E = 1000 [GPa], = 5 [Mg/m3], then

These materials do not perform as well as the first set of materials with
!
$
# GPa1/2 &
M = 10 #
3/2 & .
Mg
#
&
#"
&%
m3

3) As we move to larger E and smaller , does M increase or decrease?

Remember, we have to keep the line slope equal to three, or we won't have an equi-index
line.
We want high performance, so we keep shifting the line to the upper left until we only
have a small set of materials above the line -- THESE are the CANDIDATE materials for
this design.
We find a lot of materials that perform AS WELL AS OR BETTER THAN the
composites!
4) As M changes, what does that mean?

so a material with an M = 4 weighs HALF that of a material with an M = 2, but TWICE


an M = 8. By maximizing M, we minimize the mass... just what our design calls for.
Week One: Page 23

W. Warnes: Oregon State University

5) We stated earlier that we can use any of these ratios for M:

M=

E 1/2
E
E 1/ 3
or
M
=
or
M
=
.
! 3/2
!3
!

Let's check to see if that makes sense:

E
!3

E 1/2
! 3/2

E 1/ 3
!

CHART:

CHART:

CHART:

SLOPE:

SLOPE:

SLOPE:

UNITS OF M:

UNITS OF M:

UNITS OF M:

The VALUE of the M will be different in each case, the UNITS of M will be different,
but because the SLOPE and the SELECTION CHARTS are the same, the MATERIALS
SELECTED WILL BE THE SAME!

Week One: Page 24

W. Warnes: Oregon State University

6) Reality Check Number One:


This optimization procedure has given us the BEST PERFORMING MATERIALS given
our stated objective (measure of performance) and constraint. But are the answers
sensible? How do we know?
Look back at the derivation. All but one of the parameters are known (FIXED) or are
determined by the optimization process (MATERIAL PROPERTIES). To check the
design, it is important to use the materials that have been suggested to determine the
value of the free parameter, t in this case, to see if it is indeed sensible.
For a first check, let's compare the relative thicknesses needed for the different materials
to function in the design:
From the derivation, we know:

Solve for t to get the free parameter as a function of the other parameters in the design:

The relative thickness of two competing materials is given by:

NOTE: The advantage of comparing the relative thickness is that a lot of the design
parameters cancel out, so that we dont need to know a lot about the details of the design
to look at the relative values of the free parameter.
Week One: Page 25

W. Warnes: Oregon State University

For several candidate materials, we have the following property data (obtained from the
Ashby chart #1):

Glass
Composites
Wood Products
Polymer Foams

E [GPa]
100
30
4
0.1

[Mg/m3]
2.2
1.5
0.8
0.2

( / E) [Mg/m3GPa]
0.022
0.05
0.2
2

( / E)1/2 [Mg/m3GPa]1/2
0.148
0.224
0.447
1.4

Compare these materials against a "standard" material; for instance, glass has been
commonly used in this application.
Material
Composites
Wood Products
Polymer Foam

t / t(glass), or how much thicker than glass the mirror must be.

NOTE that all four of these materials PERFORM the samethey have the same value of M and the same
performance (MASS). But, because they have different properties, they have different values of the free
parameter needed to make them work.

7) Reality Check Number Two:


So, we know the relative thicknesses, but what about the actual thicknesses? To find
these, we need to have values for all of the FIXED and CONSTRAINT parameterswe
need to know more about the design. Let's pick some reasonable values:
r=
g = 9.8 [m/s2]
=
From the analysis, we know t =

Material
Glass
Composites
Wood Products
Polymer Foam
Week One: Page 26

t = Mirror Thickness [m]

W. Warnes: Oregon State University

WOW! These are HUGE!!! What went wrong?


Two important points here. FIRST, the optimization process tells you the best materials
for the job. It doesn't guarantee that your design will work. It is quite possible that the
design cannot be built to work using existing materials. If this is the case, what are your
options?
GIVE UP, or REDESIGN.
SECOND, the design requirements, calculational model, or constraint equations may be
wrong or too simple to accurately describe the design. Your options are:
GIVE UP, CHECK YOUR ASSUMPTIONS, or REDESIGN.

In this case we know an 8 [m] mirror has been constructed from glass that is only about 1
[m] thick and that it works. How could we redesign to reduce the thickness needed for the
mirror?
One option:

Now the model must change, perhaps to a simple beam like this, or something more
complex.

(NOTE - for the simple beam model shown above, the performance index turns out to be
the same, which yields the same materials for the selection process. Changing to a more
realistic model or design changes the constants in the equations, but not the best choice of
materials. Woooo...cool!)
End of File.

Week One: Page 27

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Two
Winter 2012

MATERIALS SELECTION OPTIMIZATION


WITHOUT SHAPE EFFECTS -- II
Reading: Ashby Chapters 5 and 6.

EXAMPLE: Materials for Flywheels


DESIGN ASSIGNMENT: Design a flywheel to store as much energy per unit weight as
possible and not fail under centripetal loading.
MODEL: Solid disk of diameter 2R and thickness t rotating with angular velocity .

MOP: Maximize energy per unit mass

Kinetic energy of spinning disk:

Mass of flywheel:

Week Two: Page 1

W. Warnes: Oregon State University

PARAMETERS:
R=
=
t=
Design Requirements for Flywheels
Function
Store as much energy per unit mass as possible without failure
Constraints
Size? Possibly
No failure when spinning with angular velocity of
Objective
Maximize energy/mass stored
Free Parameters
Radius, R
Thickness, t
To complete our list of constraint and materials property parameters, we'll need to look at
the "no failure" constraint. Basically, we'll keep increasing the rotational velocity until
the flywheel comes apart. What is the maximum stress in the flywheel?

Our constraint is that the maximum stress must be less than the yield strength, so

CONSTRAINT EQUATION: rewrite in terms of the free parameters as

APPLY TO MOP:

Week Two: Page 2

W. Warnes: Oregon State University

APPLY TO SELECTION CHART: Given our performance index, we probably want to


use a selection chart like log() versus log(), and look at a line of slope = 1.

MATERIALS SELECTION: We want to


consider materials above and to the left of our
line, as these have larger values of / . What
materials do we get?
First, let's examine the units of M, and then
make a table from data in the selection chart:

MATERIAL

M[ MPa/(Mg/m3) ]

CERAMICS
CFRP
GFRP
Be alloys
Steels
Ti alloys
Mg alloys
Al alloys
Woods
Lead
Cast Iron

Week Two: Page 3

W. Warnes: Oregon State University

WOW! Why did we not select lead and cast iron in favor of low density materials?
Our design requirement was maximum U/m, which led us to increase up to the failure
constraint--a strength limited design.
For lead and cast iron flywheels, the design statement is different. If we just want to
maximize U, then the
MOP =

(The message is: the design statement is critical to getting the right answer!)
Look at the list again: The best performer is CERAMICS-- BUT, we need to check on
the measurement of f for ceramics used in Ashby's chart. In the description of the
selection chart it says that f means:

0.2% offset tensile yield strength for METALS


non-linear stress point for POLYMERS
compressive crushing strength for CERAMICS

The flywheel is in tensile loading, so ceramics are not such good performers. The best
performers are:
Materials for Flywheels
Material
M = (! f / " ) #$ MPa / ( Mg / m 3 ) %&
CFRP
200-500
GFRP
100-400
Be alloys
300
Other alloy systems
100-200
Lead
Cast Iron

3
8-10

Comment
Excellent choice, expensive
Cheaper alternative to carbon
Manufacturing toxicity issues
Many inexpensive and common
alternatives.
Steam-punk option
Classic choice for energy storage

To down-select, we need another constraint criterion (COST?). This brings up an


important issue about MULTIPLE CONSTRAINTS, which we'll postpone until a future
time.
End of File.
Week Two: Page 4

W. Warnes: Oregon State University

Lecture Notes for Week Two


Winter 2012

MATERIALS SELECTION OPTIMIZATION


WITHOUT SHAPE EFFECTS -- III
Reading: Ashby Chapter 5 and 6.

A DIFFERENT EXAMPLE: Spring design


We use springs for storing elastic energy. We usually want to maximize the
energy/volume, or the energy/mass. Stored elastic energy is found from the stress-strain
curve for the material as the work done by the applied stress:
ENERGY/VOL=

Because we are in the elastic region.


This is the area under the stress-strain curve
up to the yield stress, and gives us

[ENERGY/VOL]axial =

Leaf springs and torsion bars are less efficient in storing energy than axially loaded
springs because not all the material is loaded to the yield point, so

[ENERGY/VOL]torsion =

[ENERGY/VOL]leaf =

Week Two: Page 5

W. Warnes: Oregon State University

In all of these cases, the performance index will be M1-MAX=

Look at the selection chart of modulus versus strength:

BUT!! This time better materials appear to the LOWER RIGHT (increasing and
decreasing E).
We find lots of conventional materials for springs (elastomers, steels) but also many
others:

Ceramics- good in compression


Glass- often used in high precision instrumentation
Composites-look interesting

WHAT ABOUT ENERGY/MASS SPRINGS?


Week Two: Page 6

W. Warnes: Oregon State University

In this case, the performance index becomes M2-MAX =

What do we use for a selection chart? Since the mass is a key consideration in these
spring designs, we want to have represented in both the axes:

Now we can use selection chart 5:

The selection leads us to elastomers, ceramics and polymers, but the metals lose out
because of their high density.

Week Two: Page 7

W. Warnes: Oregon State University

This raises the question of "how do I know which selection chart to use?"

Two ideas to keep in mind at this point:


1. As already stated, the mass is important, so keeping in the axes is a good idea.
2. We don't have selection charts for the other ones.
But what if we did?

We can construct the other selection plots using the CES software, and the net result is
that, while the selection plots are somewhat different, the materials that pass the selection
are identical.
End of File.

Week Two: Page 8

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Two
Winter 2012

CASE STUDY: NATURAL BIOMATERIALS


Reading: Ashby chapter 12.9

BIOMIMETICS
Natural materials have continually been used by mankind in the development of new
engineering applications. Many natural materials continue into the present day as useful
materials, including wood, bamboo, and natural fibers, such as cotton, hemp, and silk.
Especially recently, as materials engineers have become increasingly facile at building
new materials from the ground up (composites, multilayers and heterostructures,
functionally gradient materials, quantum well structures, etc.), natural materials have
become a focus for developing materials for engineering applications. This field of
research has become known as biomimeticsusing natural materials as models for
new engineered materials.
Recognizing why biomimetics is such an exciting research area starts by looking at the
materials that nature has developed for its use. In almost all cases, natural materials are
composite, or hybrid, materials, often displaying structural features over large range of
dimensional scales. Ashby has collected the physical properties of many natural materials
in his book and in the CES software (which well look at next week). Chapter 12 has a set
of Selection Charts hidden away that focus on natural materials.

Week Two: Page 9

W. Warnes: Oregon State University

LOW-MASS ELASTIC MATERIALS (Figure 12.13)

This selection chart is used for selecting materials for applications involving stiffness per
unit mass. Weve already analyzed a couple of different applications which require lightstiff materials under different loading conditions, leading to the three guidelines shown
on the plot.
Lets put STEEL and Al on the figure, just for reference:

Steel
Al alloy

Density [kg/m3]
7900
2800

Youngs Modulus [GPa]


216
80

Strength [MPa]
1000
500

E
, beating steel by a factor of 3 to 4,
!
and pushing flax, hemp and cotton up pretty high. Woods, palm, and bamboo perform
"
E 1/2 %
very well in bending and buckling $ M =
.
! '&
#
Cellulose is the winner for tensile stiffness: M =

Week Two: Page 10

W. Warnes: Oregon State University

LOW-MASS HIGH-STRENGTH MATERIALS (Figure 12.14)

Again, natural materials show up nicely on this chart, with silk having the best strengthto-weight ratio.

Week Two: Page 11

W. Warnes: Oregon State University

ELASTIC ENERGY STORAGE MATERIALS (Figure 12.15)

For this chart, the best materials are those with large values of and small values of E: in
!2
the upper left corner. Spring materials are those with large values of M =
, while
E
!
elastic hinge materials are those with large values of M = .
E
Whats the best natural material for springs?
Whats best for elastic hinge applications?
Interesting!

Week Two: Page 12

W. Warnes: Oregon State University

TOUGH NATURAL MATERIALS (Figure 12.16)

Materials with large values of toughness are at the top of the chart (antler, bamboo, and
bone), good for impact loading.
The criterion for carrying a load safely when a crack is present is shown by the lines of
constant fracture toughness (the dashed lines at 45 degrees.)

Week Two: Page 13

W. Warnes: Oregon State University

CASE STUDY: IMPLANTABLE BIOMATERIALS


("Biomaterials-An Introduction", J. B. Park, R. S. Lakes, Plenum Press, 1992, and Biomedical
Engineering and Design Handbook, Volumes 1&2, Myer Kutz, editor, McGraw-Hill Co. Inc., 2009 )

I. HISTORY

1860's: Aseptic surgical techniques developed by Lister.


1890's: Bone repair using plaster of Paris.
early 1900's: Metallic plates used for bone fixation during skeletal repair.
Problems with corrosion and failure.
1930's: Development of stainless and Co-Cr alloys. First successful joint
replacements.
1940's: WWII pilots-PMMA shows low bioreactivity, and leads to development
of PMMA as an adhesive and skull bone replacement.
1950's: Blood vessel replacements.
1960's: Cerosium- an epoxy filled porous ceramic used as a direct bone
replacement.
1970's: Bioglasses.

II. DESIGN CONCERNS


1. Material properties (strength, fatigue, toughness, corrosion).
2. Design (load distribution, stress concentrators).
3. Biocompatibility (immune system, toxicity, inflammation, cancer).
Other effects on success rate include surgical technique, patient health, and patient
activity.
Relative importance of these issues changes with time:

Week Two: Page 14

W. Warnes: Oregon State University

III. A PARTIAL LIST OF BIOLOGICAL APPLICATIONS (from Park)

Week Two: Page 15

W. Warnes: Oregon State University

IV. BIOPOLYMERS
Five types and three reactivities:
1. WATER SOLUBLE (used in solution for lubrication, improving hydrophyllic
interactions at surfaces, reduce thrombogenesis)
2. HYDROGELS (poly hydroxyethylmethacrylate [PHEMA] used in soft contact
lenses, others used for drug delivery systems)
3. GELS (react in-situ to form soft structures; natural fibrin cross-links to form clots)
4. ELASTOMERS (principally silicones and polyurethanes, PUR)
a. Silicones more bio-inert than PUR and is oxygen permeable
b. Silicones are thermosets, while PUR are thermoplastics
c. PUR can be processed to larger range of properties
d. Silicones: artificial finger joints, blood vessels, heart valves, catheters,
implants (breast, nose, chin, ear)
e. PUR: pacemaker leads, angioplasty balloons, heart membranes
5. RIGID (main ones are Nylons [significant water absorbance issues], PET, PEEK,
PMMA, PVC [external uses], PP. PE [especially UHMWPE in hip and knee
prosthetics as a low friction and wear surface, tricky in a metal-PE wear couple
though][LDPE cant be autoclaved since Tm is too low so only outside body uses],
and PTFE [bioinert])
The reactivities are described as BIOINERT, BIOERODABLE, and
BIODEGRADABLE.
Week Two: Page 16

W. Warnes: Oregon State University

V. BIOLOGICAL CERAMICS
Four main types, or classes, determined by reaction rate in body:
1.
2.
3.
4.

INERT
POROUS INGROWTH
BIOACTIVE
RESORBABLE

Initial research on ceramic biomaterials was fueled by an interest in the chemical


inertness of them as a class, but over the past 25-30 years there has been a definite shift
toward the bioactive ceramics.

V.A. INERT BIOCERAMICS


Oxides (chemically stable), Carbon. In general these are characterized as: no change is
found in tissue, or the degradation product is easily handled by the body's natural
regulation process. In inert ceramics, the body tissue forms micron sized fibrous
membranes around the insert, and it is locked into place by mechanical interlocking of
the rough surfaces.
Week Two: Page 17

W. Warnes: Oregon State University

Typical inert ceramics are:

Al2O3 for joint prosthetics, dental applications. Alumina has a very low abrasion
rate, about 10X less than a PE/metal wear surface in joints replacements.
ZrO2 is an alternative to alumina, with a higher fracture toughness and even better
wear resistance.
LTI (low temperature isotropic) carbon for heart valves and coatings on some
prosthetics.
DLC-(Diamond-like Carbon) films because of their stability.

V.B. POROUS INGROWTH BIOCERAMICS


Surface preparation of the ceramics is a critical part of the functionality of the implant.
Made with a porous or roughened surface can allow essentially inert bioceramics to
establish a strong mechanical bond with natural tissue by allowing tissue growth into
pores and rough surfaces.
V.C. SURFACE REACTIVE BIOCERAMICS
Small amount of selective chemical reactivity with tissue leads to a CHEMICAL BOND
between the tissue and the implant. Implant is protected from further degradation due to
the reacted "passivation" layer.

BIOGLASS: Na2O-CaO-CaF2 P2O5 SiO2


APATITE: Ca10 (PO4) - 6OH2

Used for small bone replacements (low stress) and as coatings on other inserts to enhance
bonding.
Surface coatings often experience failure due to fatigue of the substrate, and the coatings
are not so good in tension.

Week Two: Page 18

W. Warnes: Oregon State University

V.D. RESORBABLE BIOCERAMICS


Materials that fill space and are taken up by the body with time, presumably to be
replaced with new bone growth. The goal is to provide a scaffold on which new healthy
tissue can grow and eventually replace the implanted ceramic.
Example:

VI. APPLICATION EXAMPLES


Corrosion Issues

Mixed Metals (Galvanic Corrosion)

Week Two: Page 19

W. Warnes: Oregon State University

Dental prosthetics: the most successful area of application of composite materials in the
bio-applications is in the dental prosthetic area, particularly involving ceramic
composites. (Dr. Kruzic: Research on mechanical failure in dental composites.)

Mechanical Design

Week Two: Page 20

W. Warnes: Oregon State University

Hip/Knee Prosthetics
Ti shaft in bone fixed by glue (PMMA) or cement. High density Al2O3 ball and socket
joint. Better than Ti on HDPE because no release of metallic and polymeric wear
particles (toxicity).
State of the art: replace the Ti with C fiber reinforced graphite, or with thermoplastic
matrix/carbon fiber composite and protective coating.

VII. FDA REGULATIONS


Biomaterials is one of the engineering areas most involved in government regulation. The
definitions are specific but not always obvious. For instance, an example given in the
Kutz book (vol. 2, p. 22, emphasis added by me):
How does this affect your morning toothbrushing? When you brush your teeth you are
using a MEDICAL DEVICEthe brush. The brush works in a mechanical manner on
your teeth to remove unwanted material. The toothpaste you use could be a COSMETIC
in that it is applied to the teeth to cleanse. However, if you choose a fluoride toothpaste
you are using a DRUG, since the fluoride is metabolized in the body in order to prevent
tooth decay. If you choose to use an oral rinse to reduce adhesion of plaque to your teeth
before you brush, you are using a MEDICAL DEVICE. The oral rinse loosens plaque
that is then removed by your brushing.
End of File.
Week Two: Page 21

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Three
Winter 2012

MULTIPLE CONSTRAINTS IN MATERIALS


SELECTION: OVERCONSTRAINED DESIGN I
Reading: Ashby Chapter 7 and 8.

Most design problems are more complex than those examples we've discussed so far.
Let's look at a more complex design:
EXAMPLE
DESIGN ASSIGNMENT:

Cantilever beam of square cross section and fixed length L.


Support an end load, F, without failing.
End deflection must be less than .
Minimum mass.

MODEL:

MOP: minimum mass:

PARAMETERS:
L:
F:
t:
:
:
Week Three: Page 1

W. Warnes: Oregon State University

Design Summary:
Design Requirements for Stiff, Strong Cantilever
Function
Support end load, F, without failure.
Constraints
Length, L, specified;
No plastic yielding in bending;
Stiff enough to not deform under end load by more than .
Objective
Minimize mass.
Free Parameters
Square cross section of t x t.
Okaylets tackle these two constraints as if they were each a separate constraint, using
the optimization recipe.
FIRST CONSTRAINT: No failure under the end load, F.

SUBSTITUTE INTO THE MOP:

WHAT ABOUT THE OTHER CONSTRAINT (ON DEFLECTION)?

SUBSTITUTE INTO MOP:

Uh-oh... we've got two constraints, and now we have two materials performance indices
(M) and they're DIFFERENT! What do we do?
Week Three: Page 2

W. Warnes: Oregon State University

This type of design is called an OVERCONSTRAINED design-- That is, we have more
constraints than free parameters. Most materials selection problems are OVERCONSTRAINED. There are several ways we can deal with multiple constraints in the
selection process, by using DECISION MATRICES, MULTIPLE SELECTION
STAGES, ACTIVE CONSTRAINTS, COUPLING EQUATIONS, and PENALTY
FUNCTIONS.

I. DECISION MATRICES
Commonly used and presented in other design classes. One version comes from Crane
and Charles (see syllabus for reference).
In simplest form, a matrix is developed with the DESIGN REQUIREMENTS along the
columns and the CANDIDATE MATERIALS along the rows:
I.A.

Materials are rated in a GO-NO GO fashion as either acceptable (a), under-value (U),
overvalue (O), or excessive (E).
PROBLEMS:
1.
2.
3.
Next best (but still not very good) approach is to inject some quantitative measure by
replacing U, a, O, E with numbers 1-5 (increasing is better).

Week Three: Page 3

W. Warnes: Oregon State University

I.B.

This provides a quantifiable selection criterion, but


PROBLEMS:
1.
2.
To eliminate concern #2, we could add WEIGHTING FACTORS,
I.C.

but this just adds another level of subjectivity. How can you back-up the assertion that the
rigidity is 2.5 times more important than cracking resistance?
One significant improvement we can add here is to use PERFORMANCE INDICES
rather than materials properties. This essentially takes us beyond primary constraints into
the realm of optimization. For each constraint or design goal, we develop an M value to
use as one of the columns:

Week Three: Page 4

W. Warnes: Oregon State University

I.D.

Crane and Charles convert these to dimensionless numbers (relative values) by dividing
by the largest property value, and then sum these to determine the overall rating of the
material.
This is better (we're selecting based on performance indices) but now we're back to
treating all of these with the same importance. The last act of Crane and Charles is to
apply weighting factors to the performance indices:
I.E.

This is pretty good except that it is STILL SUBJECTIVE because there is no justification
for the weighting factors that are used.
The difficulty with most of the decision matrix approaches is simply this subjectivity.
There are some schemes for improving that, and Dr. Ullman's group at OSU has been
studying the design methodology and has developed an approach that has resulted in a
computer program called the Engineering Decision Support System (EDSS).

http://www.cs.orst.edu/~dambrosi/edss/info.html

Week Three: Page 5

W. Warnes: Oregon State University

II. MULTIPLE SELECTION STAGES


A second approach to the multiple constraints problem is the use of selection stages. Each
of the constraints is used to develop a performance index as we did in the earlier
example:
M1, M2, M3, ..., Mn.
These are rank ordered in order of importance (uh-oh...) from most important to least. We
use the first performance index on the appropriate selection chart, and select a large
enough (uh-oh...) group of materials to leave something for the other constraints to work
with. Repeat with the other performance indices.
(Why the "uh-oh"s? How do we decide on the rank ordering? Subjective decision. How
do we decide on the number of materials to leave in the pool at each stage? Subjective
decision again.)
II.A. EXAMPLE: Multiple Stage Selection for a Precision Measurement System
(micrometer).
There are several design goals we want to meet with this design:
1.
2.
3.
4.

minimize the measurement uncertainty due to vibrations of the stiff structure,


minimize the distortions of the structure due to temperature effects,
keep the hardness high for good wear properties, and
keep the cost low.

Design Requirements for Precision Measurement Tool


Function
Choose material for an inexpensive high precision hand-held
micrometer
Constraints
Handheld limits size and mass--not too limiting.
Objective
Minimize vibrational errors;
Minimize thermal distortions due to temperature gradients;
High hardness for improved wear properties;
Minimize the cost.
Free Parameters
Choice of material
Let's tackle these one at a time-II.A.1. VIBRATIONS
We want to drive the natural frequency of the main structure as high as possible. The
useful approximations give us the natural frequency as:
Week Three: Page 6

W. Warnes: Oregon State University

II.A.2. THERMAL DISTORTION


The strain due to a change in the temperature of the structure is determined by

If we want to know how the thermal strain changes along the length of our structure due
to a temperature gradient, we take the derivative to find

We also know (for a 1-D heat flow approximation) that the heat flux is given by

" d! %
To minimize the thermal distortion $ T ' for a given heat flow, we need to maximize
# dx &

II.A.3. HIGH HARDNESS


We can treat the hardness, H, as a direct function of the yield strength:

Week Three: Page 7

W. Warnes: Oregon State University

II.A.4. COST
Finally to keep the cost low, we want to maximize

So, to summarize, we have FOUR design goals, each of which gives us a different
performance index:
Minimize vibrations:

Minimize thermal distortion:

Maximize hardness:

Minimize cost:

With the multiple stage selection approach we will take each of these individually and
make a series of selection charts.

Week Three: Page 8

W. Warnes: Oregon State University

II.B. First Selection Stage: We will use Ashby's chart 1, with a slope of 1, and the
selection area above and left of the line:

We don't want to eliminate too many materials, otherwise, there'd be nothing left for the
other criteria to do.
Rank ordered list of materials that "passed" this selection stage, from highest performers
to lowest:
Ceramics
Be
CFRP
Glasses/WC/GFRP
Woods/Rock, Stone, Cement/Ti, W, Mo, steel, and Al alloys.

Week Three: Page 9

W. Warnes: Oregon State University

II.C. Second Selection Stage: We will use Ashby's chart 10, with a slope of 1, and the
selection area below and right of the line:

Rank ordered list of materials that "passed" this selection stage, from highest performers
to lowest:
Ceramics
Invar
SiC/W, Si, Mo, Ag, Au, Be (pure metals)
Al alloys
Steel
Notice that there is some overlap between materials that passed the first stage and those
that passed the second. Thats good.

Week Three: Page 10

W. Warnes: Oregon State University

II.D. Third Selection Stage: We will use Ashby's chart 15, and apply the last two
constraints as primary constraints. We want to search in a selection area in the upper left
of the chart:

Rank ordered list of materials that "passed" this selection stage, from highest performers
to lowest:
Glasses
Steel/Stone
Al alloys/Composites
Mg, Zn, Ni, and Ti alloys/Ceramics
Compare these in a table:
First Selection Stage
Ceramics
Be
CFRP

Second Selection Stage


Ceramics
Invar
SiC/(pure metals)
W, Si, Mo, Ag, Au, Be

Glasses/WC/GFRP

Al alloys

Woods/Rock, Stone,
Cement/Ti, W, Mo, steel,
and Al alloys

Steel

Week Three: Page 11

Third Selection Stage


Glasses
Steel/Stone
Al alloys/Composites
Mg, Zn, Ni, and Ti alloys/
Ceramics

W. Warnes: Oregon State University

The candidate materials that make it through all three stages are STEELS and Al
ALLOYS.
We might want to relax the selection criteria a bit to take another look at ceramic
materials, which appear in two of the lists.
The main advantage of this multiple stage selection process is that the assumptions are
simple and clearly stated regarding the rank ordering of the performance indices. The
disadvantage is that it is still subjective in determining the rank ordering and the position
of the selection lines on each of the charts.
The quantitative approach to multiple constraints combines the decision matrices and
selection stages with coupling equations and/or penalty functions. These are topics well
look at next.
End of File.

Week Three: Page 12

W. Warnes: Oregon State University

ME480/580: Materials Selection


Tutorial Overview Notes on CES-Edupack
Winter 2012

INTRODUCTION TO CES (CAMBRIDGE


ENGINEERING SELECTOR)-EDUPACK
SOFTWARE FOR WINDOWS

Nomenclature:
Throughout these notes references to buttons or icons that should be clicked will be given
in BOLD and pull-down menu items will be given in ITALICS .
1) Log onto your Engineering account.
2) Once in Windows, open the MIME Apps and start the CES-EduPack 2010 program.
(We still have the older version, CES Selector 3.1, on-line. Dont use it by mistake!)

INSIDE CES:
You will see the WELCOME screen when you startup, and a Choose Configuration
window.
There are three levels of material and process database information in this version of
the software:
Level 1: about 70 materials and 70 processes, with a limited set of property data;
Level 2: about 100 materials and 110 processes, with an extended set of property
data;
Level 3: about 3000 materials with a comprehensive data set for each.
1) For now, choose English -- Level 1 until you are used to the program. Later on well
switch to Level 3 to use all the information for doing problems and the design project.
2) You should now be in the main program control window. At the top on the left, you
should see the database you are using, along with a pull down menu for the TABLE
Week Three: Page 13

W. Warnes: Oregon State University

(MaterialUniverse), and SUBSET (Edu Level 1). You will also see a toolbar with several
icons along the top of the main window. These are how you will interact with CES.

CES INFORMATION:
There is a large amount of on-line help and database information available in CES.
1) You should automatically be in the browsing tool, but if not click on the BROWSE
tab to see the information in the materials database.
2) Double click on a folder to open it. Eventually you'll work your way through the
hierarchy to an individual material record. Take a look at the materials record. This is the
database information that has been developed for each material in the database.
3) You can change the database to browse by choosing a different TABLE or SUBSET
from the pull-down menus. Try several different tables to see what they offer.
4) You may also change databases by clicking on the CHANGE button. If you change to
Level 3, you will find SEVEN different TABLES, and a larger number of SUBSETS.
Within the Level 3 MaterialsUniverse, for example, there are a number of SUBSETS,
including All Bulk Materials, Ceramics, Foams, Magnetic Materials, Metals, Polymers,
and Woods. You might use these to narrow down a selection process to a smaller class of
materials.
5) You can also SEARCH the database using the SEARCH button. Nuf said.
6) Reference material is also available on-line, as well as an on-line help function. Click
on the HELP button or the menu item. The "CES InDepth" is an on-line reference book
about CES and the selection process we have been using in class. In fact, all of the
appendices form the textbook can be found in here (if you know where to look!)
7) There are also video tutorials and getting started guides that you can access if you want
to learn more about the capabilities of the program.
8) For the last thing to do on this part, click on the TOOLS button and select OPTIONS.
Click on the UNITS tab to set the preferred units of the data. Choose the currency you
want to use for cost analysis here. This also allows you to set the units for the selection
charts. Choose SI (consistent) for the unit system. (HINT: using USD [$] instead of
Myanmar Kyat would probably be a good idea).

MAKING A SELECTION CHART (the cool stuff):


Week Three: Page 14

W. Warnes: Oregon State University

1) Click on the SELECT tab to start. (Alternatively, you can choose the NEW PROJECT
menu item in the File Menu.) You will need to choose a database and subset to use in the
selection project.
2) Now click on the NEW GRAPH STAGE icon on the toolbar or from the SELECT
menu. (The toolbar buttons are, from left to right, NEW GRAPH STAGE, NEW LIMIT
STAGE, NEW TREE STAGE.)
3) You should get a window with the "Graph Stage Wizard" title. Make sure that the XAXIS tab is selected, and then use the ATTRIBUTES pull-down menu to choose the
material property to plot on the x-axis. Choose YIELD STRENGTH (ELASTIC LIMIT)
from the pull down list.
3) Click on the Y-AXIS tab to set the material property for the Y-axis.
4) Select YOUNG'S MODULUS from the ATTRIBUTES menu.
5) Click OK.
6) You should now have a new window labeled "Stage: 1" with a graph of your selection
chart, showing on the right side of the screen, along with a new tool bar row with about
16 icons on it.

CHANGING AND USING A SELECTION CHART:


1) Click on the STAGE PROPERTIES icon (the first icon on the left of the new tool
bar.)
2) You can now change the axes of the active stage. Change the SCALES to be LINEAR
in both X and Y. Click OK. Now you know why the data is usually plotted on a log-log
plot.
3) Click on any bubble on the chart to find out what the material is. Drag the pop-up label
around, and it should leave a connecting line behind pointing to the bubble. Doubleclicking on a bubble brings up the materials data sheet for that material.
4) Delete the label by selecting it with the mouse and pushing the DELETE key.
5) Change the axes back to log-log.
6) There are three types of selection tools you can use: point-line, gradient-line, and box.
These are the icons that follow the CURSOR icon.
7) For simple or primary constraints, you should use the BOX selection tool. Click on the
BOX button. Then click on a point in the selection chart and drag the mouse to enclose a
Week Three: Page 15

W. Warnes: Oregon State University

set of materials in the box. Note that the STATUS BAR (at the bottom left of the screen)
gives you the X,Y location of your cursor in the plot units. Note also that any material
bubble that is partly inside the selection box is colored, while the others are greyed-out.
The colored bubbles have been selected by the selection process, and now show up as a
list in the RESULTS section on the left side.
8) Now click on the GRADIENT-LINE selection tool button. Type in the slope of the
selection criterion (line slope) you want (use 1 for now), and click OK.
9) Click on some X-Y position to position the line, and the line will be drawn for you at
that location. Notice that the STATUS BAR shows you the value of the selection
criterion for the line position you have chosen, along with the value of M for that line (be
wary of the units, though!). The final step is to click either ABOVE or BELOW the line
to tell the program which region is the selection region. Again, colored materials have
passed this selection, and greyed materials have failed.
10) Moving the cursor onto the selection line allows you to reposition the selection line
for higher or lower M values. If you want to change the slope, you can start over by
clicking the GRADIENT-LINE selection tool button again.
11) Note that you may have only ONE selection criterion operating at a time on a single
selection chart. If you want more than one criterion for a particular set of x-y axes, you
need to make-up additional STAGES with the same axes and apply the other selection
criteria to those.
12) The RESULTS section in the left of the window shows you the materials passing
your selection criterion. You can modify the results section by using the pull-down menu
to choose what results to view. This is especially helpful when using multiple stages.
13) Finally, you can save this set of selection criterion to disk and recall it later using the
SAVE PROJECT menu item. In the FILE menu

A MULTIPLE STAGE EXAMPLE:


We want to do a materials selection for a high quality precision measuring system,
essentially a top line micrometer (we did this one in class as our example of a multiple
stage selection process). After extensive analysis, we have found that we need a material
that will produce a LOW THERMAL DISTORTION (M1 = / ), LOW VIBRATION
(M2 = (E / )1/2), maximize the HARDNESS (M3 = H), and minimize the cost
(M4 = 1/C ).
1) First you will need to start with a clean project. In the FILE menu click on the NEW
PROJECT item. We will use the EduLevel 1: Materials database for this example.
Make sure this is set up in the selection data section.
Week Three: Page 16

W. Warnes: Oregon State University

2) Stage 1 will deal with M1: Click NEW GRAPHICAL STAGE selection, and in the
X-Axis properties choose the THERMAL CONDUCTIVITY property. For the Y-Axis
properties choose THERMAL EXPANSION COEFFICIENT, and click on the OK
button.
3) Now click on the GRADIENT-LINE selection tool button. Type in the slope of the
selection criterion you want (use 1), and click OK. Locate the point for = 10 [W/m-C],
and =1 X 10-6 [1/C]. (Remember that you can use the Status Bar at the bottom of the
window to tell you the X-Y position of the cursor.) Click BELOW the line (since we
want large and small ).
4) Now that you have a selection criterion on the graph, click on the STAGE
PROPERTIES icon. A new tab is available that lets you change the details of your
selection- slope, side of the line, and exact location! Use this to place your selection line
in exactly the same position that I have used (X = 10, Y = 1). If youve done everything
the same as I have, you should see SEVEN candidate materials in results list.
5) Stage 2 will deal with M2: Click NEW GRAPHICAL STAGE, and in the X-Axis
properties choose the DENSITY. In the Y-Axis properties choose YOUNG'S
MODULUS, and click on the OK button.
6) Now click on the GRADIENT-LINE selection tool button. Type in the slope of the
selection criterion you want (use 1), and click OK. Locate the point for E = 2 x 109 [Pa],
and = 100 [kg/m3]. Click ABOVE the line (since we want large E and small ).
7) If you click on the STAGE PROPERTIES button while in Stage 2 you can choose to
turn off or on the display to show the RESULT INTERSECTION, those materials that
have passed all the stages so far. If you only want to see the materials that pass, choose to
HIDE FAILED RECORDS. (I don't recommend this at the beginning!). Your results list
should show SIX materials now that pass both selection criterion.
8) Stage 3 will deal with M3 and M4: Click NEW GRAPHICAL STAGE, and for the XAxis properties we have to do something fancy. There is not a property listed for COST,
but there are properties PRICE [USD/kg] and [kg/m3]. First, for the x-axis, click on the
ADVANCED button. You should see a hierarchical list of all the materials properties
available. Click on the GENERAL PROPERTIES in the pull-down menu and you will
see a list of the general properties. By choosing properties from the list and using the
math function buttons, you can set up quite complicated materials selection axes. Wow!
Isn't this cool? Select PRICE and multiply it by DENSITY to get the X-axis to be the
[USD/volume] you need for minimum cost design. Click OK. We should also change the
name of the axis to at least include the UNITS!!!! (something like MATERIAL COST
[$/m^3]) so we know what we are looking at in the selection chart.
9) In the Y-Axis properties choose HARDNESS-VICKERS, and click on the OK button.

Week Three: Page 17

W. Warnes: Oregon State University

10) Now click on the BOX selection tool button. Use the box to select the materials with
a MATERIAL COST less than 10,000 [USD/m3], and a HARDNESS greater than 1 x 109
[Pa].
11) Go to the RESULTS window and check your results. You can view the selection
criteria you have used here, as well as the materials that have passed each stage. If you
have done this problem the same way I have, you will end up with three materials
passing: Al Alloys, Silicon, and Silicon Carbide.
NOTES:
You may only search one database at a time. To change databases:
1) Click the CHANGE button in the Selection Data section and select the database you
want to search.
2) Then choose the subset of materials you want to Select From You can fiddle with
this, for instance, by choosing to look only at ceramics or metals.
Once you have developed a selection stage, changing databases does not change your
selection stage(s) or selection criteria. CES will automatically run through the selection
process using the new database whenever you change databases. It's easy to search the
other databases this way. My advice is to always start off with the ALL BULK
MATERIALS subset in the LEVEL 3 database, and use the others as your design
develops.
(If you do this now, you should have 14 candidate materials from the three stage
selection, using the Level 3 database.)
End of File.

Week Three: Page 18

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Case Study
Winter 2012

CASE STUDIES IN MATERIALS SELECTION:


POLYMER FOAMS
("Polymeric Foams", Klempner and Frisch, 1991; "Plastic Foams" Frisch and Saunders, 1973)

Why look at foams? EXAMPLE: Simply supported beam in bending- minimum mass (or
cost). Assume b, L, are fixed, h is free, and the center deflection under load, F, is limited.

Use Rule of Mixtures to determine foam properties, e.g. 90% air foam:

SOthis means the foam material, which is 90% nothing with no properties, has a
performance nearly FIVE TIMES the solid polymer beam ( or, looked at another way, for
hf = 2hs you can get the same deflection with 80% less mass!).
One can also laminate the surface of foams with a high strength layer to drive
strength/weight ratio even farther up.

Week Three: Page 19

W. Warnes: Oregon State University

Foams also have energy absorption properties due to the compressibility of the gas in the
cells.
New materials classFoamed metals (Al and steel) behave exactly the same way! Foams
are nifty!
TYPES OF POLYMER FOAMS:

Gas-dispersed foams, using "blowing agents"


Syntactic foams, using hollow spheres of glass or plastic.
Open-cell vs. closed-cell.

POLYURETHANE FOAMS:
Most widely used. Depending on chemistry can vary their properties from flexible
cushions to rigid foams for structural applications, with density ranging from 0.0096-0.96
Mg/m3.
Can be made in a continuous process as a "bun" 2-8 feet wide X 1-5 feet thick X 10-60
feet long.

Can be processed as "integral skin" foams.


POLYSTYRENE FOAMS: Also very widely used in the form of extruded blocks. Formed
by:
1. Force volatile liquid (neopentane) into crystalline spheres of PS ( ~ 0.96 Mg/m3)
2. Pre-expansion done with steam, spheres expand to 0.016-0.16 Mg/m3.
3. Final-expansion in a mold with steam heat, spheres fuse together.
ABS FOAMS: Used in pallets, and as structural material in furniture.

Week Three: Page 20

W. Warnes: Oregon State University

SYNTACTIC FOAMS: Use hollow microspheres (30 micron diameter) of glass, ceramic,
or plastic for difficult to foam materials, such as epoxies.
ARCHITECTURAL USES OF FOAMS:
Besides insulating properties (PUR foams among the lowest thermal conduction
materials), can also be used as a primary structural material, as in this University of
Michigan study from the late 60's.
Major controlling factor: keeping within small elastic and creep deformation limits.
Looked at double-curved shells. Several different approaches:
POLYSTYRENE SPIRAL GENERATION

POLYURETHANE SPRAY APPLICATION

Week Three: Page 21

W. Warnes: Oregon State University

FOLDED PLATE STRUCTURES WITH POLYURETHANE/PAPER BOARDS

FILAMENT WINDING ON PUR BOARD

End of File.

Week Three: Page 22

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Four
Winter 2012

OVERCONSTRAINED DESIGN PART II:


ACTIVE CONSTRAINT METHOD
Let's look at an example with a single design objective (MOP), but several constraints, an
OVERCONSTRAINED problem. One way to approach it is to use a multiple selection
stage process, as we did for the precision micrometer example in the last lecture. A
difficulty with the approach is the ordering of the constraints and selection stages, and the
subjective placement of the selection line in the multiple stages. A more systematic
approach uses the "active constraint" approach. An example:
EXAMPLE: The support rod for an infrared-electronics cooling cryogenic fluid container
in a spacecraft is to be designed. The most important characteristic of this tie rod is that it
should carry a minimum amount of conductive heat into the cryogenic container. The
conductive heat flow equation tells us that the conductive heat flow along this support rod
is:

where C is a constant (the temperature gradient), is the thermal conductivity of the rod,
and A is the cross sectional area of the rod.
There are three constraints on the rod:
First, that the loading due to the mass of the cryogenic fluid and container should not
exceed the failure strength of the tie rod (ignore the mass of the rod).
Second, the deflection, , should be less than a critical value, max.
Third, the vertical frequency of vibration must be high enough to not affect the measurements being made. In other words, f should be larger than a critical frequency, fmin.

Week Four: Page 1

W. Warnes: Oregon State University

MODEL:
Assume a SOLID CYLIDER for the rod, A = ! r 2 .

MOP: minimum heat flow into the cryogen:

PARAMETERS
L=
=
A (or r) =
max =
fmin =
F=
q=
DESIGN SUMMARY:
Design Requirements for Space Cryogenic Support Rod
Function
Support end load, F, without conducting heat into cryogenic fluid.
Constraints
Length, L, specified;
No plastic yielding in bending;
Stiff enough to not deform under end load by more than ;
Fundamental vibration frequency larger than fmin.
Objective
Minimize conductive heat flow.
Free Parameters
Cross section of solid cylinder, A.

Week Four: Page 2

W. Warnes: Oregon State University

PERFORMANCE EQUATION ONE : Start with the load constraint:

PERFORMANCE EQUATION TWO : Now use the deflection constraint:

PERFORMANCE EQUATION THREE : And finally the vibration constraint: For a


vibrating rod with a mass at the end, the fundamental (lowest) frequency is

where K is the elastic stiffness, given by

Week Four: Page 3

W. Warnes: Oregon State University

and m is the mass of the cryogenic container, mc. Then

To perform the multiple selection stage process, we would set up two stages, one for M1
and one for M2, M3.
For the active constraint approach, we have to know more about the design, especially the
details of the values of the fixed and constraint parameters. First, write out the equations
for the MOP using each of the constraints:

PMAX(M1)=

PMAX (M2)=

PMAX (M3)=

Week Four: Page 4

W. Warnes: Oregon State University

If we know, or can estimate, the values of the fixed and constraint parameters, we can
calculate numerical values of the measures of performance for each material. Let's put
some numbers down for this design:
F (= mc*a) = 196 [N]
mc = 20 [kg]
max = 0.01 [m]
fmin = 5 [Hz] = 5 [1/s]
L = 0.1 [m]
C = temperature gradient = (300 [K])/(0.1 [m]) = 3000 [K/m]
Now we can set up a spreadsheet table of values for the material properties of the
materials we're interested in and calculate the measures of performance. My spreadsheet
in EXCEL looks something like this, and I pulled the rough values from the Ashby
selection charts:
VARIABLES:
L [m]
mc [kg]
F [N]
Delta Max [m]
f min [Hz]
C [K/m]

Material
CFRP
GFRP
Ti alloys
Al alloys
Steel
Polymer Foam

=
=
=
=
=
=

0.1
20
196
1.00E-02
5
3000

SIGMA
[MPa]
1500
1200
700
300
1000
6

E
[GPa]
200
90
100
70
220
0.3

LAMBDA
[W/m-K]
0.6
0.6
12
200
30
0.04

Load
P1 [1/W]
2.71E+01
2.34E+01
8.16E-01
2.78E-02
4.14E-01
1.03E+01

Deflection Vibration Minimum


P2 [1/W] P3 [1/W]
P [1/W]
2.74E+01 2.73E+01 2.71E+01
1.84E+01 1.83E+01 1.83E+01
9.69E-01 9.66E-01 8.16E-01
4.87E-02 4.85E-02 2.78E-02
5.75E-01 5.73E-01 4.14E-01
1.59E+01 1.59E+01 1.03E+01

For each individual material, we look at the SMALLEST value of P. WHY?


In order to satisfy all the constraints, we must satisfy the one that most limits our
performance. If we can satisfy that one (by choosing a particular value of r, the free
parameter), we will satisfy all of them,
In this example, the minimum performance for all of the materials is different for each
material -- the ACTIVE CONSTRAINT varies for the materials we have examined. If we
don't satisfy it, the design will fail.

Week Four: Page 5

W. Warnes: Oregon State University

Now, we can pick the material with the LARGEST value of the active constraint
performance (from the last column in this example) to be the optimal performer for the
design, in this case CFRP.
What have we learned by going through this active constraint analysis that we didn't
know before?
1) To become more objective and quantitative in the selection process for
OVERCONSTRAINED designs, we need to know more detailed information about the
design.
2) It's a lot more work and time to do all of the quantitative calculations, but...
3) We now know what the limiting constraints on the materials are. With the spreadsheet,
we can play some "what if" games with the fixed parameters-- how do the P's change if
you decrease the cutoff frequency, or increase the mass, or allow less deflection? These
trade-offs can be used to tune up the design and go back to your boss/client with a
quantitative reason to consider changing one of the fixed parameters. As the values of
these parameters change, at some point one of the other constraints will become the
active constraint for a given material.
The Last Step: REALITY CHECK: Let's plug back into the constraint equations to find
the value of the cylinder radius in each case:
Material
CFRP
GFRP
Ti alloys
Al alloys
Steel
Polymer Foam

SIGMA
[MPa]
1500
1200
700
300
1000
6

E
[GPa]
200
90
100
70
220
0.3

LAMBDA
[W/m-K]
0.6
0.6
12
200
30
0.04

Load
r1 [m]
2.55E-03
2.75E-03
3.29E-03
4.37E-03
2.92E-03
1.61E-02

Deflection
r2 [m]
2.54E-03
3.10E-03
3.02E-03
3.30E-03
2.48E-03
1.29E-02

Vibration
r3 [m]
2.54E-03
3.11E-03
3.03E-03
3.31E-03
2.48E-03
1.29E-02

Maximum
r [m]
2.55E-03
3.11E-03
3.29E-03
4.37E-03
2.92E-03
1.61E-02

End of File.

Week Four: Page 6

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Four
Winter 2012

OVERCONSTRAINED DESIGN PART III: COUPLING


EQUATIONS
Consider a design situation in which we have one design goal (MOP), two constraints,
and one free parameter. We are over-constrained in this situation.
By calculating a performance index analysis using the first constraint, we end up with:

With the second constraint, we have:

Now, the MOP is the same, so we can equate these two (we only have one design, which
will perform at a given level, P):

The relative weighting of the two performance indices is DETERMINED BY THE


DESIGN and not by our subjective judgments!
III.A. EXAMPLE: A Light Tie Rod
Week Four: Page 7

W. Warnes: Oregon State University

III.A.1. DESIGN ASSIGNMENT:

cylindrical tie rod of length L


minimum weight
support a load F
extension less than X

III.A.2. MODEL: Same as before

III.A.3. MOP: minimum mass:

III.A.4. PARAMETERS
L=
F=
X =
A=
=

DESIGN SUMMARY:
Design Requirements for Light Tie Rod
Function
Support end load, F, without conducting heat into cryogenic fluid.
Constraints
Length, L, specified;
No plastic yielding in tension;
Stiff enough to not deform under end load by more than x;
Objective
Minimize the mass of the rod.
Free Parameters
Cross section of solid cylinder, A.

Week Four: Page 8

W. Warnes: Oregon State University

III.A.5. PERFORMANCE EQUATION ONE (derived previously in the example from


Week One)

III.A.6. PERFORMANCE EQUATION TWO

III.A.7. DEVELOP THE COUPLING EQUATION:

Week Four: Page 9

W. Warnes: Oregon State University

Result
The best performing material will be one in which E/ is maximized, / is maximized,
and their ratio is held at L/X.
How do we apply this to a selection chart? We want to use a chart for this example like
chart 5.

We want both of our performance indices to be maximized, so we'll be looking at


materials in the upper right hand corner of the chart. For our particular design, we'll have
a given value of L/X that is determined by the constraints of the design. Lets say it is
100.
We will look at a straight line of slope 1 on the plot, and we want the line for which the
ratio of the performance indices is 100

Week Four: Page 10

W. Warnes: Oregon State University

By moving along this line of constant L/X, we improve our performance by increasing
the values of the performance index, and we simultaneously maintain the weighting
factor determined by the design.
Our best choice of material is...

We probably want to open up the search region a bit, to allow some materials other than
diamond in the mix, so for this case we can use a rectangular search region centered on
the line of L/X = 100.
By moving away from the line, we shift toward STIFFNESS DOMINATED designs (to
the upper left) or toward STRENGTH DOMINATED design (to the lower right).
NOTE: If you use coupling equations, you don't need to use multiple stage selection
processes, but you may have to generate your own Ashby Selection Charts!

MULTIPLE CONSTRAINT DESIGN: FULLY


DETERMINED DESIGNS
Look back at previous lectures -- we had an example of OVERCONSTRAINED design
with the cantilever beam. We had two constraints (no failure under an end load, F, and
deflection less than ), and only one free parameter (square cross section of t X t). We
ended up with two materials performance indices, M1 and M2, which we could use in a
two-stage selection process.
Alternatively, we could use a coupling equation to couple the two M values together and
do a one stage process, as just described (check out HW3...).
A last possibility is to revise the design statement to increase the number of free
parameters. This will give us two free parameters and two constraints--fitting the
definition of a FULLY DETERMINED design.
EXAMPLE: Light cantilever beam
DESIGN ASSIGNMENT: Let's change it slightly, from a square beam of t X t to a
rectangular beam of b X h.

Week Four: Page 11

W. Warnes: Oregon State University

Cantilever beam of rectangular cross section and length L.


Support an end load, F, without failing.
End deflection must be less than .
Minimum mass.

MODEL:

MOP: minimum mass:

PARAMETERS:
L:
F:
b:
h:
:
:
DESIGN SUMMARY:
Design Requirements for Light Cantilever Beam
Function
Support end load, F, without failure.
Constraints
Length, L, specified;
No plastic yielding in bending;
Stiff enough to not deform under end load by more than .
Objective
Minimize mass of beam.
Free Parameters
Cross section of solid rectangle, b x h.
Week Four: Page 12

W. Warnes: Oregon State University

We've got two free parameters (b and h) and two constraints!


CONSTRAINT ONE:

CONSTRAINT TWO:

SOLVE FOR b AND h:

Week Four: Page 13

W. Warnes: Oregon State University

SUBSTITUTE INTO MOP:

This type of design is called FULLY DETERMINED design. We can get a complete
solution (with one M value), because we have the same number of free parameters as
constraints.

MULTIPLE CONSTRAINT OPTIMIZATION


(A GENERAL DESCRIPTION)
The first part of the optimization process is writing out the following:
1. Measure(s) of Performance: quantitative functions to maximize the relative
success of different designs. (P)
2. Constraining Equation(s): functions that set acceptable limits on the behavior of
the design in use. (C)
3. Design-fixed Parameters: parameters that appear in the P and/or C equations that
are not changeable under the conditions of the design. (D)
4. Free Parameters: the additional parameters from P and/or C that are not fixed.
(F)
There are several possible scenarios:

Week Four: Page 14

W. Warnes: Oregon State University

I.A. SINGLE MOP DESIGNS


For designs with a single measure of performance, we can imagine several possibilities:
I.A.1. Zero Free Parameters
This is a pretty unusual situation, but it is conceivable when an existing
design is to be used and only requires a change in material. Not a lot of
opportunity here for optimization.

I.A.2. One Free Parameter


I.A.2.a. One Constraint Equation (1C1F)
With one C and one F we are FULLY DETERMINED, and the
constraint, C, is applied to the measure of performance, P, through
the free parameter, F, to develop a single performance index, M.

I.A.2.b. Two Constraints (2C1F)


Now the design is OVERCONSTRAINED. We treat each
constraint separately as in the 1C1F case. In so doing, we end up
with two performance indices:

Since we still have only one P, these two functions can be equated
to find a coupling equation (or a relative weighting factor) of
M1 / M2 :

Week Four: Page 15

W. Warnes: Oregon State University

I.A.2.c. Three (and more) Constraints


The design is definitely overconstrained. We start the same way
we did for the 2C1F design:

Using the pairs of performance index functions, we can determine


the relative weightings of these M's.

Week Four: Page 16

W. Warnes: Oregon State University

I.A.3. Two Free Parameters


I.A.3.a. One Constraint (1C2F)
In this case the design in UNDERCONSTRAINED. We need to
find a way of either fixing one of the parameters, or come up with
another constraint. In some cases we may be able to change
variables to reduce to 1F.
EXAMPLE: A minimum mass connecting rod of rectangular cross
section with heat flow larger than some value qo.

Convert from 2F to 1F using A = bh, since both constraint and


MOP depend only on the area A.
I.A.3.b. 2C2F
Fully determined design. Solve the two constraining equations for
the two unknowns (F1, F2) and plug into the P.

We end up with ONE performance index, M.


I.A.3.c. 3C2F
Overconstrained, and can be treated as three independent 2C2F
problems:

Week Four: Page 17

W. Warnes: Oregon State University

Since we are still talking about 1P designs, we can generate


coupling equations here as well:

I.A.4. General Results for 1P Designs


C<F
C=F
C>F

Underconstrained; need to add a constraint.


Fully determined; one performance index, M.
Overconstrained; multiple M's, coupling equation(s).

I.B. MULTIPLE MOP DESIGNS


The first step will be to rank order the P's. Remember how to determine
whether you are dealing with a P or a C:

If the feature is to be MINIMIZED or MAXIMIZED, then it is a P.


If the feature must be GREATER THAN or LESS THAN a
reference value, then it is a C.

With a rank ordered list of the P's, we can treat each one separately as a
single MOP problem:

Week Four: Page 18

W. Warnes: Oregon State University

NOTES:
1. The DESIGN will generally have a single set of constraints that
can be applied to all of the P's, but the free parameters may be
different in different P equations. In this example, P1 depends only
on F1 and F2, while P2 depends on F1 and F3.
2. If a constraint equation doesn't involve any of the F's in a particular
P, then the constraint can't be used to optimize this measure of
performance. In this example, C3 does not apply to P1.
3. You can't get coupling equations between M's determined from
different P's. In this example:

We can form coupling equations by coupling the three P2 equations, but


we can't find a coupling equation relating M(2)12 and M(1)12 because P1 is
not equal to P2.

Week Four: Page 19

W. Warnes: Oregon State University

WHAT NEXT?
What do you do with all of these performance indices and coupling
equations? Two choices:
Rank order the performance indices by order of importance and perform a
multiple stage selection process, or;
Get more information about the design and determine the active constraint
for each material in a tabular matrix, or:
Set up a decision matrix based on the performance index values for each
material. the decision matrix can be rank ordered, or can be set up with
weighting factors as determined from the coupling equations.
End of File.

Week Four: Page 20

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Five
Winter 2012

CALCULATING MASSIVELY OVERCONSTRAINED DESIGNS


It is often a good idea at the beginning of a design project to figure out how intense the
analysis is going to be by calculating the total number of performance indices and
coupling equations you are likely to end up with. Just a quick word about how to do this
as a combinatorial problem.
EXAMPLE: You have a design with

ONE measure of performance,


THREE free parameters, and
EIGHT constraints.

To find a materials performance index, we need to have a FULLY DETERMINED


design, so we'll want to take three of the eight constraints at a time to solve for the three
free parameters. How many combinations of the eight constraints do we have in sets of
three?

Now, the problem with this counting is that it counts the combination of constraint 1+2+3
as different from the combination of 1+3+2 and 3+2+1. We need to divide the total by the
number of combinations of three constraints (in any order) that we can have. This
overcounting factor is found by:

Week Five: Page 1

W. Warnes: Oregon State University

So, the total number of UNIQUE combinations of the eight constraints in sets of three is
56! WOW...that's 56 M-values in this problem!
BUT WAIT...THAT'S NOT ALL! How many combinations of the M-values can we have
in groups of two to create unique coupling equations do we have? Using the same
process, we get:

That's a LOT of coupling plots to make up...even with CES.


Here is a plot showing the rapid increase in the number of M-values and coupling
equations with the number of constraints for a single MOP, four free parameter design.
Ouch!

What can we do to make this better? One choice is to do what I told you not to
do...change one of the constraints into a measure of performance (for example, instead of
having a maximum allowable cost, set up a minimum cost measure of performance).
Week Five: Page 2

W. Warnes: Oregon State University

This changes the problem. We now have a design with

TWO measures of performance,


THREE free parameters, and
SEVEN constraints.

Using the combinatorial calculations above, we would have (for each of two MOPs)

This is still pretty ugly, but is a lot more tractable than the original problem. Of course,
we've replaced the original problem with needing to (subjectively) determine which MOP
is the most important.
The main message is that, when you are massively OVER-CONSTRAINED it is best to
try to reduce the number of constraints you have to being only one or two larger than the
number of free parameters you have, and the way to do this is to turn some of the
constraints into MOPs.
REMEMBER: If you have the same number of free parameters as constraints, you will
have only ONE M-value, no matter how many free parameters you have!

Week Five: Page 3

W. Warnes: Oregon State University

CES AND COUPLING CHARTS


You may have wondered by now how to do the coupling charts in CES and have CES tell
you the best materials in the RESULTS section. The answer is that you have to trick it,
and here's how.
FIRST: Make up a coupling chart the
way that you usually do, with M1 on the
Y-axis and M2 on the X-axis. This will
be STAGE ONE of a multiple stage
selection process. Find the correct
location of the coupling equation line
(slope of one and at the correct location
for the design parameters) and set the
line on the chart at the correct location.
This is shown schematically in the
figure to the right with the correct value
of the coupling constant shown as the
dotted line. The solid line is slightly
offset above the correct position for (NOTE: The position of the coupling line can
clarity in the last couple of plots.
be EXACTLY placed on the selection chart in
CES by looking (in the "Project" Menu) for
Now, click BELOW this line so that "Stage Properties". The dialog box has a tab
CES selects all the materials that touch for "Selection" that allows you to enter exact
the line or are below it.
values for where you want the line placed.)

SECOND: Make a COPY of the first


selection stage by clicking on it in the
PROJECT window, choosing COPY,
and PASTE. This will make an
identical version of your first stage as
STAGE TWO. In stage two, choose the
selection region to be the area ABOVE
the line.

Week Five: Page 4

W. Warnes: Oregon State University

By setting the conditions of Stage Two


to show only the subset of materials
that have passed BOTH stage one and
two, you will see only those materials
that touch the selection line (shown
schematically to the right). The
RESULTS window will also list only
these materials.

To select the BEST MATERIALS,


make one more copy of the original
stage, setting it as STAGE THREE. For
this stage, use a selection line that is
approximately at right angles to the
coupling equation line and select the
region ABOVE the line to be the active
region. By moving this selection line up
and down you can pick off the materials
that give you the MAXIMUM values of
M1 and M2, AND are on the coupling
line. The RESULTS window now lists
the materials passing all three stages.
You can move the third stage selection
back and forth to determine the rank
order of the materials as well.
If you have a design that has more than one coupling equation, you will have to make a
number of selection stage sets, three stages for each coupling equation. If you use the
copy and paste functions, though, this is not too tough to do, and you only have to fiddle
with the selection line in the third stage of each coupling equation set.
End of File.

Week Five: Page 5

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Five
Winter 2012

DEALING WITH CONLFICTING OBJECTIVES


Reading: Ashby Chapters 7 and 8.

So far we have talked about the myriad techniques for dealing with overconstrained
designs, using the active constraint method, or the coupling equation approach. In all of
these designs, we have been very careful to define only one measure of performance. But,
there are some design situations in which you find yourself with two or more design
objectives; multiple measures of performance. In many cases, these multiple objectives
are conflicting; you cant satisfy them both with the same material. This requires a
different approach taken from optimization theory called TRADE-OFF PLOTS, and
PENALTY FUNCTIONS.
NOTE: The biggest difference in our process and thinking from what we have done so far
is that the objective function must be defined such that we want to MINIMIZE it in order
to get the best performance. I have been careful to require everything to be defined in
terms of MAXIMIZING PERFORMANCE, but for this type of optimization analysis, we
need to define a minimizing function that will maximize performance. Ashby calls these
objective equations P, as before, so we just need to be careful that we know whether the P
requires maximizing or minimizing to bring success.
A SIMPLE EXAMPLE: Well go back to a previous problemthe simple cantilever. The
design statement has been:

MINIMUM MASS (our measure of performance),


Fixed length, L,
Square cross section, b X b,
Not fail plastically under end load F.

This is a FULLY DETERMINED design, and the only change we need to make from
previous analysis is that the measure of performance, Pmin, will be a minimizing function:

Week Five: Page 6

W. Warnes: Oregon State University

Going through the usual routine with the constraint gives us the following expression for
the measure of performance:

In the normal analysis, wed want to pull out the materials selection index, in this case,

and then use it to make a selection plot. (NOTE: Do we want to find materials that have
large or small values of M?)
For our design, we are also told that we must MINIMIZE THE COST. This is clearly a
second design objective, and the chance that the material that minimizes mass will also
minimize cost is pretty slight. Here we have a case of multiple and conflicting objectives.
We will go ahead and analyze the design using the second constraint, which we write as

where C is the material property of Price, having units of [USD/kg]. Pushing through
with the analysis (using the load constraint) gives us the following result for P:

Two objectives, two M-values, can we couple them? NO! The Ps are different, so we
cant set them equal to find a coupling equation.
To proceed, we need to know more about the design. As in other complex designs, we
need to know the values of the fixed parameters to carry on. Lets assume
F=

L=

Then the constant factor in the first measure of performance is

Week Five: Page 7

W. Warnes: Oregon State University

Plugging this into the measure of performance to be minimized, we get:

In this case, the constant factor in the second measure of performance is the same, so our
second measure of performance to be minimized is:

We can use CES to make a plot for us of the two MOPs, mass and cost. Using the
ADVANCED axis option, we can write out the equation

m = 113

!
= 113 * [Density] / [Yield Strength (Elastic Limit)] ^ 0.6667
" 2/3
f

Now the Y-axis will be the MASS of the beam, in [kg]. Similarly, we can set up the Xaxis to be the COST of the beam in [USD]. Schematically, the plot looks like this:

Okaytime for some optimization theory terminology. Each of the bubbles on this plot
is called a SOLUTION, because it represents, for a particular material, the cost and mass
of a beam that will satisfy the constraint.

Week Five: Page 8

W. Warnes: Oregon State University

Look at the bubble labeled Awe can see that there are many other solutions that have
either a smaller value of mass or a smaller cost, represented by the vertical and horizontal
lines on the plot. The materials between the lines have BOTH a smaller mass and lower
cost. A is said to be a DOMINATED solution, because there is at least one other solution
that outperforms it on BOTH performance metrics.
The B bubble, on the other hand, is a NON-DOMINATED solution, because there are no
other solutions that have both a smaller mass and a lower cost. But is B the OPTIMUM
solution?
Looking at the plot shows that there are, in fact a whole variety of solutions that are nondominated. We can draw a line through them all and we arrive at a boundary, which is
called the TRADE-OFF SURFACE, along which all the non-dominated solutions lie.
We can, at this point, use our expertise or intuition to choose the best materials from all
of the candidate, non-dominated, solutions, but there must be some quantitative way of
dealing with this. The answer is to develop PENALTY FUNCTIONS.

Week Five: Page 9

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Case Study
Winter 2012

CASE STUDIES IN MATERIALS SELECTION:


SHIPBUILDING
(REFERENCE: "Brittle Behavior of Engineering Structures", E. R. Parker, John Wiley
and Sons, NY, 1957.)

There are two basic parts to a


ship- the hollow HULL, and
the SUPERSTRUCTURE.
The hull is subjected to two
forces:
1) gravity due to the
mass of the ship and
the
cargo,
and
2) buoyancy of the
hull.
While these forces balance,
they
are
not
always
uniformly distributed, and
can be strongly affected by
cargo loading.
For shorter cargo ships,
"HOGGING" is common, as
the buoyancy in the center is
larger per unit length than it
is at the ends.
Longer ships tend to "SAG", even in still water, but the worst case comes from riding the
waves.
The hull is subjected to a large bending moment, and so tends to fail in panel buckling.
The superstructure is used as a panel stiffener to prevent hull buckling.

Week Five: Page 10

W. Warnes: Oregon State University

We can analyze this using simple beam bending:

The first performance measure will be to minimize the mass of the ship:

subject to the constraint of no failure:

The second performance measure is to minimize the deflection subject to the failure
constraint:

Week Five: Page 11

W. Warnes: Oregon State University

Since these are two DIFFERENT MOP's, we can't generate a coupling equation. Look at
potential materials using a multi-stage selection.
SELECTION STAGE 1) versus = CHART 2, slope = 1, upper left

CANDIDATE MATERIALS:

CFRP
GFRP
Steels
Ti alloys
Al alloys
Wood

Week Five: Page 12

W. Warnes: Oregon State University

SELECTION STAGE 2) E versus = CHART 4, slope = 1, upper left

CANDIDATE MATERIALS:

Al alloys
Steels
Ti alloys
CFRP, GFRP, Wood

PERFORMANCE:
CANDIDATE MATL
CFRP
GFRP
Steels
Ti Alloys
Al Alloys
Woods

[MPa]
700
400
1800
1000
430
110

[Mg/m3]

E [GPa]

M1

M2

1.6
1.6
7.8
4.2
2.6
0.6

30
20
220
100
60
1

440
250
230
240
165
185

0.043
0.050
0.122
0.100
0.140
0.009

For M1 the best performers are polymer composites, but they lose out to steel in M2 for
which they show deflections three times larger than the steels. Ti and Al look pretty
good, but they lose out when we throw cost into the equation. HIGH TENSILE
STRENGTH STEEL is the commonly used material, except in high performance weightdriven designs (racing yachts with CFRP).

Week Five: Page 13

W. Warnes: Oregon State University

STEEL SHIP PLATES AND FRACTURE


In the early 1900's, ship plates were completely riveted together. At the end of WWI a
push for faster construction times drove shipbuilders toward using substantially welded
ship plates, but as the war stopped, the money for development dried up. In 1921 a small
merchant ship (the FULLAGAR, 150 ft. long) was the first fully welded ship to hit the
water, and worked in England for many years.
At the start of WWII, the push came on to rapidly produce ships for the merchant marine
fleet to supply the war effort, and welding technology was again pushed. The approach
was a "cookie cutter" one, with a small number of ship plans, and many shipyards
producing the same design. The construction was begun in 1941, and in total,
2500 Liberty Ships
500 T-2 tankers
400 Victory ships
were constructed. Shortly after these ships entered service, they began breaking apart,
sometimes spectacularly! The rapid and massive scale-up required by the war meant that
unskilled laborers and inadequate welding practice were used, and blamed for what
happened.

Week Five: Page 14

W. Warnes: Oregon State University

Two major causes of the failures were found:


1) STRESS RAISERS: access holes through the decking plates and structural plates were
cut for ladderways and cargo loading. These were initially cut as rectangular holes. Many
cracks initiated at the corners of these holes. By changing the design to rounded holes,
many fewer failures were reported.
2) UNKNOWN EFFECTS: (at the time)
No correlation was found between failure and the tensile strength of the steel samples
taken from various parts of the failed ship plates. Loading at failure was typically around
700 [MPa], well within the design load.
Extensive study of the brittle fracture energy (toughness) using the Charpy impact test
found the following:

Week Five: Page 15

W. Warnes: Oregon State University

Ductile to Brittle Transition Temperature (DBTT) is arbitrarily set as about 15 [ft-lbs] of


fracture energy. ANSWER: the DBTT was too high (the steel was brittle at the
temperatures of the North Sea).

(NOTE: This is not the story of what happened to the Titanic, a fully riveted ship built
and sunk in 1912. Substandard wrought-iron rivets used in the hull, as opposed to the
good quality steel rivets used elsewhere in the ship, failed prematurely in the cold water
iceberg-collision. The riveted hull plates separated, bringing her down.)
End of File.

Week Five: Page 16

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Six
Winter 2012

CONFLICTING OBJECTIVES:
PENALTY FUNCTIONS
The penalty function is basically a way of combining the different measures of
performance into a single-valued figure of merit that quantitatively describes the
performance. This is done by defining a figure of merit, Z, such that

where the s are called EXCHANGE CONSTANTS. The exchange constants are
defined mathematically as the change in Z associated with an independent change in the
performance metric, P:

We often find that cost is one of the performance metrics, and it has been a standard
process in optimization theory to use money as the overriding figure of merit by letting
the units of the figure of merit be in dollars:

Lets look at how this shows up on our trade-off plot. For the cantilever beam example,
there are two performance metrics, minimum mass and minimum cost. The penalty
function is then written as:

and we want to find the one solution, or material, that minimizes the value of Z.
For a given value of Z, the penalty function is a linear equation relating the mass and the
cost as
Week Six: Page 1

W. Warnes: Oregon State University

For each constant value of Z, this equation plots as a line of slope = !

1
, so that the
"

trade-off plot looks schematically like this:

Each of the lines is a line of constant Z, with the same slope (or exchange constant, ),
and the material that is the best choice minimizes Z, as shown in the left figure. Since we
usually need to plot the materials data on log-log plots, the log-log version is shown on
the right, with the only real difference being the shape of the constant Z linesthe linear
equations now become curves in the log-log plot.
The main difficulty in performing the penalty function analysis comes down to finding
the exchange constants, . How does one decide how many dollars the designer is willing
to trade for a kg change in mass? Looking at the trade-off plot, we can see that the exact
value of the exchange constant is not really needed. Examine the figure on the next page:
we have the same solution (material) for any value of between 1 and 2.

Week Six: Page 2

W. Warnes: Oregon State University

Only when the exchange constant is outside that range do we end up with a different
solution being the best choice. That means we may be able to be a little sloppy in our
determination of without affecting our choice of material.

-1/1

-1/2

Lets go back to the example and see what, if anything, we gain by working through the
trade-off plots. The design involved a simple cantilever to support an end load, and we
required both minimum mass and minimum cost.
MULTI-STAGE SELECTION OF THE CANTILEVER (from Week Five Lectures)
We can use a two-stage selection process with one stage for each performance metric.
We have that

Which gives two M values,

Week Six: Page 3

W. Warnes: Oregon State University

FIRST STAGE: MINIMUM MASS; M 1 ( MIN ) =

!
:
" 2/3
f

Use ! f versus " (Y vs. X) with a slope of 3/2:

OKAYwe draw the line no problemnow, which side of the line do we want to look
at? In this case, we have defined the M based on minimizing the performance metric, P,
so that we want a small value of M. This pushes us towards the materials in the upper
left.
Looking at our candidate materials, we have (I used CES 2011 to get my list):

With the best material being


Week Six: Page 4

W. Warnes: Oregon State University

SECOND STAGE: MINIMUM COST; M 2 ( MIN ) =

C!
:
" 2/3
f

Use ! f versus C" (Y vs. X) with a slope of 3/2 (we must use CES for this one)

Candidate materials include:

The overlap of materials is

(NOTE that I used a first stage to eliminate ceramic and natural materials from the final
answer.)
Week Six: Page 5

W. Warnes: Oregon State University

PENALTY FUNCTION SELECTION OF THE CANTILEVER


We can use CES to generate a trade-off plot using the analysis we have already done. The
2/3
C'
X-axis will be set up as COST using ($) = !# 113 ( Nim ) m $& 2/3 , and the Y-axis as the
"
%(
f
2/3 $ '
!
MASS using m = # 113 ( Nim ) m & 2/3 . The plot in CES looks like this:
"
%(
f

(Note that I used a first stage to eliminate the ceramic and natural materials again, which
are greyed out. Be sure to CHECK YOUR UNITS!)
Based on the selection line I drew, the top four materials are:

Week Six: Page 6

W. Warnes: Oregon State University

(NOTE: Heres what the CES Trade-off plot looks like with LINEAR axes.)

What is the difference between the two procedures (multi-stage versus trade-off plot)?
1) We ended up with pretty nearly the same materials in both casesthats good.
2) We needed one less selection plot in using the trade-off plot, since it captured
everything on one plotthats better.
3) We also have a quantitative reason for our materials when we use the trade-off
plot and the penalty function because we can justify each material choice based
on the range of exchange constants over which it is the best choice. This is
definitely a good thing, and makes the penalty function approach the better
choice.

Week Six: Page 7

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Six
Winter 2012

OVERCONSTRAINED DESIGN: PART IV


A FINAL EXAMPLE
Let's try an example of a more complicated design. A microbrewery in Oregon wants to
redesign their shipping and storage containers for their fine malt products. Being
responsible members of society, they have a strong interest in "green" design; that is,
having their storage containers be as environmentally benign as possible. They have put
together the following list of requirements for us:

As environmentally benign as possible;


hold about 10 gallons of liquid;
reusable;
be able to be stacked three high while in storage;
capable of being refrigerated;
hold their refrigerated temperature over a long time period.

After conversation with our customers, we find that they really are committed to the
green design, even if it costs them more to build. This gives us our
MEASURE OF PERFORMANCE: (Use Ashby's approach of minimum embodied
energy, and make sure that the container can be reused, to get at the green design.)

MODEL: Initially, let's start off with a basic cylinder design. Also, let's assume we're
going to get a thin walled container.

Week Six: Page 8

W. Warnes: Oregon State University

Since it has to hold 10 gallons, this fixes our L and R, though we could get a little fancy
and set:

PARAMETERS:

CONSTRAINTS: Given the list of requirements we have, the constraints might be


broken out as
1: Not fail under a crushing load, F, equal to the weight of two full containers;
2: Keep the time to warm up the contents greater than some critical value; and
3: Not fail due to refrigeration.
Constraint 3 seems pretty trivial--not many materials that wouldn't pass that at the
temperatures of interest here (though if we were to go below freezing point, it might be a
different story due to thermal stresses between container and contents).
CONSTRAINT ONE: No fail under end load, F.

Week Six: Page 9

W. Warnes: Oregon State University

CONSTRAINT TWO: Time to warm up should be long. Let's look at this a bit. The main
problem is with the liquid refreshment warming up by some amount (call it in C) over
a time period of t seconds. Heat will have to flow through the walls of the container by
conduction. How much heat energy do we need to heat up the fluid by ?

Now, how long does it take to get that much heat energy in through the sides of the
container? Use 1-D heat flow equation:

Plug the total energy needed into the heat flow equation, and solve for the free parameter,
w.

Week Six: Page 10

W. Warnes: Oregon State University

SUB INTO THE PERFORMANCE EQUATION:

OKAY! We have an overconstrained problem (2C1F), so we end up with two materials


performance indices, M1 and M2. What do we do now?

Week Six: Page 11

W. Warnes: Oregon State University

MULTI-STAGE SELECTION
FIRST STAGE: Use M1 = /q* (NOTE: This is now Figure 15.10 in edition 4)

Slope =

Selection Region =

Week Six: Page 12

W. Warnes: Oregon State University

SECOND STAGE: Use M2 = 1 / (q* )


Chart =
Slope =

AN ASIDE: MAKING YOUR OWN FIRST ORDER SELECTION CHARTS


versus q* is a combination of chart 9 ( versus a) and chart 18 ( versus q*). We can
generate the selection chart we want by making a transparency overlay, tracking the
range of values we are interested in for each material, and transferring it to the new
selection chart:

With much grunting and sweating, we come up with a pretty fair selection chart for
versus q* by hand (using only the charts in the book!):

Week Six: Page 13

W. Warnes: Oregon State University

(Or, spend the money to use Ashby's program, CES, and have the computer do it for you
in about 5 minutes).
OKAY- back to the design. With our new selection chart of versus q*, we need to
know:
SELECTION REGION =

Now we can do the two stage selection process to get our materials choices (left as an
exercise for the reader...)
What about the active constraint? We want to look, for each material, at the MINIMUM
value of P as determined by the two performance equations we have:

PMAX-1 =

Week Six: Page 14

PMAX-2 =

W. Warnes: Oregon State University

Comparing these shows that we need to know the following design parameters to
determine the value of the measure of performance of each of the materials:

Wow! Basically, we need to know everything about the design! We might as well go
ahead and do the coupling equation approach, since the information that we need about
the design is the same. (We'll do this next time.)
What about a reality check? We need to calculate w (wall thickness) for each material:

w (strength) =

w(thermal) =

Again, we need to know a fair amount about the details of the design to finish the reality
check. We'll do this in the next section.

Week Six: Page 15

W. Warnes: Oregon State University

COUPLING EQUATION FOR THE BEER KEG


According to our understanding of 2C1F designs, we should be able to put together a
coupling equation. Set the two performances equal to each other to get:

This tells us we want a plot of

How do we get this? EITHER: use CES, OR make a table of all the materials you are
interested in, possibly by first looking at the two stage selection process.
How do we use the coupling equation? At this point, we need to put in some values for
the parameters in the coupling equation. Early in the design, we can make
approximations:
HEAT CAPACITY:

LOAD F: Customers tell us they want to stack three high. The bottom container needs to
support a weight equal to two full containers.

Week Six: Page 16

W. Warnes: Oregon State University

CONTAINER FIXED DIMENSIONS R AND L:

TEMPERATURE RISE, :

TIME TO HEAT UP, t:

TEMPERATURE GRADIENT, T: (between the temperature of the beverage and the


outside temperature)

Plug these values into the coupling equation, and this tells us that
M1 / M2 =

What does this look like on the coupling plot?

Week Six: Page 17

W. Warnes: Oregon State University

How do we find the place to put the line? Its all about UNITS!

NOTE: Use the STAGE PROPERTIES option in CES to set the line exactly where you
want it.

Plot the M-values on the two axes (M1 versus M2) and set the line at the value of
9.2 x 10-3 [MPa-W / m-K]. Choosing only materials that lie along this line, we want the
ones with the highest values of M1 and M2 (on the line and to the upper right). This
selection gives us:

Week Six: Page 18

W. Warnes: Oregon State University

Best materials (in rank order) come out to be: (there are only SIX)
Ultra Low Density Wood (Transverse),
Low Density Rigid Polymer Foam
Very Low Density Rigid Polymer Foam
Medium Density Flexible Polymer Foam
Carbon Foam (Reticulated, Vitreous)
Low Density Flexible Polymer Foam
What about the active constraint? We want to look, for each material, at the MINIMUM
value of P as determined by the two performance equations we have. Since there are only
six, let's compare all the candidate materials:
SIGMA
Material
ULD Wood (Trans)
LD Rigid Poly Foam
VLD Rigid Poly Foam
MD Flex Poly Foam
Carbon Foam
LD Flex Poly Foam

[MPa]
0.65
1
0.17
0.374
0.255
0.16

RHO

LAMBDA

Strength

Thermal

Minimum

[Mg/m^3] [MJ/kg] [W/m-K] P1 [1/kJ] P2 [1/kJ] P [1/kJ]


0.15
5.15
0.037
1.95E-03 7.48E-04 7.48E-04
0.0655
160
0.0315 2.21E-04 6.48E-05 6.48E-05
0.028
150
0.0305 9.37E-05 1.67E-04 9.37E-05
0.0865
140
0.0595 7.15E-05 2.97E-05 2.97E-05
0.05
167.5
0.075
7.05E-05 3.41E-05 3.41E-05
0.055

140

0.0495

4.81E-05 5.61E-05 4.81E-05

This time, we see that, for some materials, the thermal constraint is limiting the
performance and for others it is the strength constraint. Selecting our minimum
performance for each material gives us the final column. The materials with the largest
value of minimum performance is ULD Wood, with VLD Rigid Poly Foam coming in
second. The order is a little different from the rank order found from the coupling
Week Six: Page 19

W. Warnes: Oregon State University

equation, but that is due to the (single) values of the material properties I used in doing
the active constraint. Both processes give us the same answers.
What about a reality check? We need to calculate w (wall thickness) for each material:
Strength

Thermal Maximum

W1 [m]

W2 [m]

W [m]

ULD Wood (Trans)


LD Rigid Poly Foam
VLD Rigid Poly Foam
MD Flex Poly Foam
Carbon Foam

1.31E-03
8.49E-04
4.99E-03
2.27E-03
3.33E-03

3.40E-03
2.89E-03
2.80E-03
5.47E-03
6.89E-03

3.40E-03
2.89E-03
4.99E-03
5.47E-03
6.89E-03

LD Flex Poly Foam

5.31E-03 4.55E-03 5.31E-03

Material

Again, we can see that the "active" constraint is different for different materials (it is the
one that requires the largest wall thickness to be satisfied). Do the thicknesses seem
reasonable?
End of File.

Week Six: Page 20

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Seven
Winter 2012

MATERIALS SELECTION WITH SHAPE:


PART ONE
Reading: Ashby Chapters 9 and 10.

Up until now, all our materials optimization has considered all materials on a same shape
basis. But, for many mechanical designs, the shape of the material plays a large role in its
performance. For example, I-beams have a better stiffness/weight ratio for bending loads
than solid square beams of the same cross sectional area.
Let's look at an example of how the shape of the material might affect the design. An
important example is for the case of torsion.

EXAMPLE:
Minimum mass torsion bar with square cross section, A = b X b, and limited torsional
deflection.
MODEL:

MOP:

Week Seven: Page 1

W. Warnes: Oregon State University

CONSTRAINT:

For a beam of solid square cross section, the torsional moment of area, K is (reference
Tabel 9.2, or Appendix B.2)

PLUG INTO THE MOP:

Okay, we get a reasonable materials performance index, not that different from what
we've used before. BUT, if we allowed the beam to be a hollow cylindrical one, we could
put more material further away from the neutral axis of the torque, and improve the
stiffness and strength of the torsion bar.

Week Seven: Page 2

W. Warnes: Oregon State University

Look at the torsional moment of area of a hollow cylinder, Khc:

If we write it as a factor multiplied by the K of a solid square, we can DEFINE the factor
(call it ) as the shape factor for twisting:

depends on the inside and outside diameter. If I make a plot of the variation of with
the inside diameter of a hollow cylinder for a fixed cross sectional area, I see something
like this:

As the inside diameter increases, holding the area constant, the shape factor increases.
What does this mean for the torsion bar? Lets find out

Week Seven: Page 3

W. Warnes: Oregon State University

We now can write our constraint equation as:

Plug into the MOP to get:

Our materials performance index now includes the shape factor, , and in such a way
that, in order to maximize performance, we also want to increase the value: a hollow
cylinder is a better performer in this application than a solid square cross section, even if
we use the same volume of material.
This is no surprise, but why should it affect our choice of materials? The reason is that
different materials have different limits on the shapes they can be formed into
(processing) and different mechanical limits on how thin they can be made and used
without buckling failure. In other words, the mechanical properties and available
processing (both properties of the material) set limits on the possible shapes, and
therefore the maximum value of for a given material can be thought of as a material
property as well. The ultimate limit of the value of is determined by the mechanical
properties responsible for Euler buckling of the thin wall; as increases, the wall gets
thinner and thinner until it will elastically buckle under the required load.
Empirical measurements of the shape factors for particular materials have been carried
out and are described in Ashby's book, similar to Figures 9.6 and 9.8, reproduced below:

Week Seven: Page 4

W. Warnes: Oregon State University

SHAPE FACTORS
Ashby approaches this with the idea of a shape factor which is defined as a dimensionless
number that characterizes the efficiency of a shape, regardless of size, for a given mode
of loading, referenced to a solid square section of the same cross sectional area. (Note:
any shape may be used as the reference shape.)
There are four shape factors:
Week Seven: Page 5

W. Warnes: Oregon State University

For STIFFNESS (deflection limited design)

! Be (bending)

!Te (twisting)

For STRENGTH (failure limited design other than


buckling)

! Bf (bending)

!Tf (twisting)

Derivation: TENSILE LOADING


The deflection in tensile loading is given by

Since depends only on the cross sectional area, A, it doesn't matter what the shape of
the cross section is, so no shape factor will be needed in tensile loading design ( = 1
always).
Derivation: BENDING STIFFNESS
For a cantilever beam, the deflection is

where I is the second moment of the area. Now a change in shape does affect the
deflection because I depends on the shape.
The second moment for a solid square section is:

We write the moment for the shape we are interested in as a multiple of that for the
square:

The shape factor, , is a dimensionless number that describes how much the shape affects
the deflection relative to the square cross section:

Week Seven: Page 6

W. Warnes: Oregon State University

Derivation: BENDING STRENGTH


The maximum stress in a cantilever beam is at the point farthest away from the neutral
axis:

! I $
Where Z is the section modulus # = & , so again the failure will depend not only on the
" ym %
cross sectional area, but also on the SHAPE of that area. In a similar way to what we did
with bending stiffness, let's look at Z = I/ym for a solid square cross section:

Now we can define our shape factor:

So we end up with the following four shape factors (shown in detail in Ashby's Table
9.3):

Week Seven: Page 7

W. Warnes: Oregon State University

EXAMPLE: Lightweight beam in bending with a deflection limit.


MODEL:

MOP:

CONSTRAINT:

SHAPE: (elastic design in bending)

plug into the constraint equation:

plug into the MOP:

Week Seven: Page 8

W. Warnes: Oregon State University

pull out the performance index:

If we don't consider the shape (or all the materials shapes are the same) then we can
remove the 's from the performance index and we get the same performance index as
before:

What does the shape analysis tell us? We can increase the performance of our material by
forming it into a shape with a large value of . Since the possible range of depends on
the ability of the shape to resist failure by local elastic or plastic deformation (which is
determined by the buckling equations), we can consider -MAX to be a material
property. Approximate values for these material limits can be found in Table 9.4 (or in
the CES datbase.)
Ashby lists two rules of thumb in determining the maximum shape factor for a
material, based on mechanics assumptions about local elastic buckling:

Useful for initial estimates of whats possible!

Week Seven: Page 9

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Seven
Winter 2012

MATERIALS SELECTION WITH SHAPE:


PART TWO
Reading: Ashby Chapters 9 and 10.

Perhaps another shape factor example will help clarify the idea behind the use of shape
factors. Let's look at an entirely new design:

A SIMPLY LOADED BEAM IN BENDING!


We want this beam to be centrally loaded, not fail under load, and have a limited center
deflection. The beam will be made of an I-beam of an as yet unspecified material. We
want to use the least expensive material to do the job. Start with the regular, plain-vanilla,
no-shape-factor approach:
THE MODEL is simple and straightforward:

The cross section is a little more complicated than usual, being an I-beam. We have a
breadth of b, height of h, webbing thickness of t, and length of L, which is fixed. The
others are free to vary:

Week Seven: Page 10

W. Warnes: Oregon State University

Our measure of performance is:

where A is given by the cross sectional area of an I-beam: A = 2t ( b + h ! 2t ) .


We have two constraints: limited deflection, and no failure under load:
DEFLECTION CONSTRAINT:

NO FAILURE UNDER LOAD CONSTRAINT:

Okay, where does that leave us? We have two constraint equations and three
unknowns...UNDERCONSTRAINED. We can either

Add a reasonable constraint (hmmm... maybe limited mass? vibration frequency?


heat flow?)
Fix one of the dimensions (which one?)

In any case, we are going to be stuck with an awful looking algebra problem to do to
solve these equations for the three unknowns b, h, and t. We might get some help by
making a thin wall approximation somewhere, but even so, the algebra is unpleasant
(unless you're that kind of engineer...).

Week Seven: Page 11

W. Warnes: Oregon State University

SHAPE FACTOR TO THE RESCUE?


Let's look at the problem using a shape factor. The big advantage of a shape factor is that
it massively simplifies the algebra by replacing the complicated formula for second
moments and section moduli with easy-to-use equations. The hard algebra is put off until
later.
DEFLECTION CONSTRAINT WITH SHAPE FACTOR

The second moment of the area for an I-beam is now written as:

This gets us to a materials selection index right away!

NO FAILURE UNDER LOAD CONSTRAINT WITH SHAPE FACTOR:

where the section modulus is now written as:

Week Seven: Page 12

W. Warnes: Oregon State University

which leads us to a second selection index:

Now, we can couple these constraints together to give us a coupling equation:

We have a very different situation from the first approach, because we have traded off the
THREE free parameters of b, h, and t for ONE free parameter of A. And the algebra is
nice (even if you're NOT one of those kinds of engineers...).

CHOOSING THE RIGHT I-BEAM


So, how do we find the dimensions of the I-beam so we can order parts and finish up on
schedule? The MATERIAL CHOICE (determined above from the coupling equation)
gives us values for all the materials properties:

Elastic Limit
Young's Modulus
Price
Density
Shape Factor for Elastic Bending
Shape Factor for Bending Strength

Week Seven: Page 13

W. Warnes: Oregon State University

Once we have the material properties, we can do a REALITY CHECK calculation to find
the necessary cross sectional area:

We know we want an I-beam shape, so we also know the following information:


Shape Factor for Elastic Bending:

Shape Factor for Bending Strength:

Cross Sectional Area:

It looks like we're in good shape; three equations, and three unknowns will provide us
with the necessary b, h, and t. The algebra is pretty ugly now, but we can either suffer
through it, use one of the symbolic manipulators (in math packages such as Mathematica
or Maple), or make friends with one of those kinds of engineers.
In addition, by using the shape factor approach, we don't even have to decide about the
shape until the end. This can be a particular advantage if one is incorporating aestheic
design into the problem as well.
End of File.

Week Seven: Page 14

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Seven
Winter 2012

MATERIALS SELECTION WITH SHAPE:


PART THREE
EXAMPLE: MATERIALS FOR BICYCLE FORKS
DESIGN: Minimize mass while supporting a load in bending without failure.
MODEL: Cantilever beam in bending.
CONSTRAINT: No failure under an end load F.

Use the definition of the shape factor for failure in bending:

This produces the constraint equation as:

Solve for A:

Substitute into the MOP = P =

Week Seven: Page 15

1
:
m

W. Warnes: Oregon State University

On a chart of versus (chart 2), we start by ignoring the shape effects (let = 1) ); this
lets us generate a set of viable candidate materials easily (and besides, we dont have any
charts that include the shape factor):

Top candidate materials are (ignoring ceramics): CFRP, G&K-FRP, Woods, and Al, Mg,
Ti alloys.

How do these perform without shape?


Calculate the performance index based on the properties:

Week Seven: Page 16

W. Warnes: Oregon State University

MATERIALS [MPa]
CFRP
500
G&K FRP
300
Woods
50
Al alloy
500
Mg alloy
250
Ti alloy
1000

[Mg/m3]
1.5
1.5
0.4
2.8
1.9
4.3

M [(MPa) 2/3/(Mg/m3)]
42
30
34
23
21
23

Now, how does shape affect these?

MATERIALS M [(MPa) 2/3/(Mg/m3)]


CFRP
42
G&K FRP
30
Woods
34
Al alloy
23
Mg alloy
21
Ti alloy
23
Steel

16


9
9
3
10
5.6
4.9

M x 2/3
182
130
71
107
66
66

13

88

(*All data for taken from Ashby's Table 9.4 except Mg and Ti, which came from the
rule-of-thumb rules from last lecture.)
Steel, which was out of the running before, shows up pretty well when we consider
shape.
SUMMARY- Selection With Shape
1.
2.
3.
4.

Analyze using Ashby's method to determine shape dependent performance index.


Apply to selection chart as if = 1 (no shape effects).
Generate a list of the top performing candidate materials (and near misses).
Apply the shape factor to the performance index for each material and compare.

OR
5. Analyze using CES to include maximum shape factors in the selection chart.
REMINDER: The shape factors are only useful in bending or torsion (not tension)
loading where shape can make a difference.
Week Seven: Page 17

W. Warnes: Oregon State University

SHAPE FACTORS: REALITY CHECKS

Look again at designing a low weight bicycle fork. We had one constraint, on strength,
modeled as a cantilever beam and analyzed with the shape factor appearing in our
definition of Z. The final selection criterion was

! 1$
What does the selection table tell us? First Column: M is tied to the MOP # & , so that
" m%
the CFRP fork will weigh about 2.6 times less than the steel fork (Msteel =16, versus
MCFRP= 42), and about 1.8 times less than aluminum. Since the doesn't appear in
column 1, this is the comparison on an EQUAL SHAPE basis.
Reality check time: Look back at our derivation for the performance index. We
developed an equation for the free parameter A as

If we make some assumptions about the loading, F = 100 [lbs.] = 445 [N],
and L = 0.3 [m], then

Plug these into the equation for the free parameter to find the cross sectional area:

MATERIALS
CFRP
Al alloy
Steel

f [N/m2]
1.5 x 109
6.0 x 108
2.0 x 109

A (=1) [m2]
6.59 x 10-5
1.21 x 10-4
5.44 x 10-5

The column of areas assumes a solid square cross sectional area (because we have
assumed = 1).

Week Seven: Page 18

W. Warnes: Oregon State University

Now, let's think about shape. Look at our equation for the cross sectional area (free
parameter). It depends on the shape factor! If we use a shaped material, we can
accomplish the same job with less cross sectional area, and therefore lower mass, and
therefore higher performance!
What do we gain by using a shaped steel material at its limit of shape factor (about 13
according to Ashby)?
First, we know that using a shape factor of 13 for steel makes it perform almost
twice as well as a solid square of CFRP ( = 1).
Second, we know that the shape factor for the steel will allow us to lower the
amount of material by lowering the cross sectional area needed to do the job:

This is a drop in mass by a factor of six! Shapes are good! Shapes are our friends!
What does a shape factor of 13 look like for the steel? (Using the shape factor table on
page 252-3 is really helpful here...)

OVAL CROSS SECTION:

Week Seven: Page 19

W. Warnes: Oregon State University

RECTANGULAR CROSS SECTION:

HOLLOW CYLINDRICAL CROSS SECTION:

Week Seven: Page 20

W. Warnes: Oregon State University

So, there are two points from this:


1) values will affect the value of the free parameter (shaping the material means we can
use less to do the job).
2) With a given and a given AREA, the choice of the shape to use is still up to us.

End of File.

Week Seven: Page 21

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Case Study, Week Eight
Winter 2012

MATERIALS SELECTION CASE STUDY:


GOLF CLUBHEADS
("Advanced Materials to the Fore", S. K. Liu, MRS Bulletin, Nov 1993, p.93; "Advanced Materials in Golf
Clubs: The Titanium Phenomenon", C. S. Shira, F. H. Froes, Journal of Metals, May 1997, p.35.)

There is a lot of mystique and controversy over the use of "high-tech" materials in sports
equipment. There is also a lot of hype, to the point that there are now rules in place to
protect consumers from unscrupulous manufacturers searching for the "tech-appeal"
buck.
With a little thought, we can decide if the hype is valid or not, from an engineering point
of view. One example is the material used in golf clubheads.
There are three types of clubs--woods, irons and putters. These have different use
conditions, purposes and therefore, design requirements.
One thing that all the clubheads have in common, in recent design, is the goal to increase
the moment of inertia as the head rotates around the shaft. The idea is that, if the
rotational moment of inertia is large, the head will twist less during contact with the ball,
and give you a straighter shot.
Week Eight: Page 1

W. Warnes: Oregon State University

PUTTERING AROUND
The result of the increase in rotational moment for a putter ends up something like this:

High density W addition on the ends of a lightweight Ti clubface. Why Ti?


Why not look at Al or Mg?

Other constraints on the putter are pretty limited.


Week Eight: Page 2

W. Warnes: Oregon State University

IRONING OUT THE PROBLEMS


For the irons, a similar modification to existing clubs can be made, and manufacturers are
doing so. With irons, we have a second concern: supplying repeatable backspin to the
ball. In this case the designers have gone to increasing the surface roughness in the face
plate in order to increase the coefficient of friction with the ball (there is a maximum
allowable roughness).
With many materials (steel, or brass for instance) the surface roughness decreases with
use, as the surface asperities get worn down or plastically flattened due to the contact
stresses.

Basically what one needs is a very high hardness material in the asperities so that they do
not plastically deform. Look at chart 2 again:

Week Eight: Page 3

W. Warnes: Oregon State University

By far, the winners are the engineering ceramics, followed by cermets, steels, composites
and Ti.
One approach is to embed diamond particles in a strong matrix, such as steel. No one has
looked at cermets (to my knowledge).
Step back, and remember that we also want to increase the moment of inertia, which
means using a low density matrix and adding high density materials to the periphery:
(Ti + diamond) + W.

WHAT WOOD WE DO NEXT?


The thinking here has been along similar lines: increase the moment of inertia around the
shaft to avoid off-center shots. This has led to hollow head castings, to push the mass out
as far as possible. We also have a couple of concerns having to do with distance.
The first MOP is minimum deflection (DELTA). The argument goes that you want a
small deflection so that you're not storing a lot of energy in the clubhead; you'd rather
have it in the ball.
We can model the clubhead as a flexible circular plate clamped around the edge:

We could chicken out and decide to maximize E, but we can come up with some
constraints, which makes this into an optimization problem:

no failure under bending load


mass less than a fixed value (about 200 [g])
no brittle failure
no yielding under impact with the ball

How many free parameters do we have here?

Week Eight: Page 4

W. Warnes: Oregon State University

Lets pick two constraints to make this a fully determined design:

Plug into the MOP:

Week Eight: Page 5

W. Warnes: Oregon State University

What if we pick another pair of constraints?

Week Eight: Page 6

W. Warnes: Oregon State University

For M1, we can use chart 5, with a slope of 2, upper left selection region:

Candidates are:

Lead, tin alloys


Al, Mg, Cu alloys, concrete
Steels, Zn, polymer foams
W
Ti

For M2, we need to prepare a table (or use CES software), since we don't have an
appropriate Ashby chart in the book.

The big winners for M2 are the ones with large KIc: steels, Ni-alloys, Ti, Cu.

Week Eight: Page 7

W. Warnes: Oregon State University

A second MOP is the damping capacity, which is a measure of the ANELASTICITY of


the material. If the anelastic deformation takes energy away from the ball, that's bad.

The energy loss is:

The candidate materials are:

Steels, Cu, Ti, Al, Mg all look good


Zn, concrete, composites, woods and foams all look bad.
End of File.

Week Eight: Page 8

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Eight
Winter 2012

USING COUPLING CHARTS AN EXAMPLE


Assume we have an overconstrained design that gives us two materials selection criteria,
M1, and M2:

The performance equation for this design is

so we want to make a coupling equation that looks like this:

Our coupling chart will have axes of M1 versus M2, and we know that we want to select
materials with a constant value of M2/M1 = A/B. Assume A/B = 0.004 [dimensionless for
this example]. We find the position of the coupling line as follows.

Week Eight: Page 9

W. Warnes: Oregon State University

The slope for the coupling line is 1 (as it ALWAYS is for all coupling charts). This gives
us the following position of the line on the coupling chart (I made the chart with CES):

The materials that are selected from this coupling chart (in rank order) are:

Rank
1
2
3
4
5
6

Material
High Carbon Steel
Low Alloy Steel
Medium Carbon Steel
Wood (parallel grain)
Cast Iron
Al Alloys

Alrighty...let's say we notice that the ratio of M2/M1 could be simplified by canceling out
the C* from both axes.

Week Eight: Page 10

W. Warnes: Oregon State University

Our design still gives us M2/M1 = A/B = 0.004, but we have a different set of axes for the
coupling chart. The position of the line on the new coupling chart is

First, let's look at the plots. Are they the same? NO! What about the materials that they
produce? The second chart gives the following set of rank ordered materials:

Rank
1
2
3
4
5
6

Material
Low Alloy Steel
High Carbon Steel
Ni Alloys
Medium Carbon Steel
Stainless Steel
Ti Alloys

For SELECTION charts (using a single M) we can simplify the equations and axes and
get the same materials, but for COUPLING charts we CAN'T simplify the axes. The best
rule of thumb is to NEVER SIMPLIFY THE AXES.

Week Eight: Page 11

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Eight
Winter 2012

DESIGN WITHIN LIMITS


There are some types of designs in which a constraint parameter must operate within a
range of values, rather than as either a maximum or minimum value, as we've talked
about so far. How does this change the optimization process? Let's do an
EXAMPLE: Minimum mass cantilever beam (yawn).
MODEL: Rectangular cross section (h x b), fixed L, b, free h, load F.

MOP:

CONSTRAINT:

(Oooooooh.... is this two constraints or just one?)


Let's start off assuming it's two constraints and analyze them separately:

Week Eight: Page 12

W. Warnes: Oregon State University

MAXIMUM DEFLECTION:

MINIMUM DEFLECTION:

Uh-oh... they're the SAME? How can we tell them apart? Do we really want to have the
same materials performance index for a maximum deflection limit and a minimum
deflection limit?
REALITY CHECK:

Week Eight: Page 13

W. Warnes: Oregon State University

So... the design optimization is the same, max or min, but the value of the free parameter
will be different. The difference between a maximum and a minimum limit comes in the
reality check.
NOTE: Optimal design gives you the best material for the job, and tells you the free
parameter values you need for that material to work based on EXACTLY MATCHING
your constraint; there's no overdesign and no underdesign--it's OPTIMAL. If you're faint
of heart, you can shift the free parameter away from the optimal choice (thereby
increasing your safety factor and moving away from optimal design). If you include a
safety factor in to begin with, then you have already covered yourself and can go with the
optimal choice.
Back to the design: Since we are thinking of these as two separate constraints, we should
be able to couple the performance indices together. What happens if we do?

Okay... not such a good idea. Why not? They're not really two different constraints! It's
one constraint with two limiting values.
SO for this design, with 1 free parameter and 1 constraint, we still have only one M
index, and the job of meeting the two limiting values falls on the reality check and final
choice of the free parameter.

Week Eight: Page 14

W. Warnes: Oregon State University

Let's complicate it a little (oh goody!). Let's introduce a second constraint to get a 2C1F
over-constrained design.
FAILURE CONSTRAINT:

So, we have two performance indices:

Let's couple these together. We'll get two coupling equations:

Week Eight: Page 15

W. Warnes: Oregon State University

How does this show up on the selection charts? We need to make a coupling equation
chart:

The search region is still the upper right of the plot, since this region maximizes both M
indices. We should look for materials that fall between the two coupling lines. We'll still
need to do a reality check against both limits of the deflection constraint.

AND SPEAKING OF COUPLING EQUATIONS...


What happens if you have a design that gives you coupling between similar indices even
though the constraints they come from are different?
EXAMPLE: A lightweight cylindrical column, unknown A, fixed length.
CONSTRAINT ONE: No buckling

Week Eight: Page 16

W. Warnes: Oregon State University

CONSTRAINT TWO: Natural frequency high

COUPLING EQUATION:

Since both Ms are essentially the same, all materials lie on a common line on the
selection chart. The big question then is where does your coupling constant line lie? In
this case, the materials lie along a line of slope = 2. What happens if your coupling
constant doesnt intersect the materials line at all?

If the performance indices are IDENTICAL, what happens then? Lets say we have a
design situation in which we have two selection indices such as

Week Eight: Page 17

W. Warnes: Oregon State University

The coupling equation gives us:

The values of A and B are found from the known (or assumed) fixed parameters for the
design. If the ratio of A/B does not equal 1 then the coupling equation cannot be satisfied
by any material. The coupling equation tells you whether the design is possible or not.
Nifty!
Not only that, but you now have a clear explanation of WHY the design will not work.
Writing out the equation A/B = 1 will show you what variables need to be changed (and
by how much) in order to make the design work. It may also require a change in
operational approach that will require a whole new analysis.

USING COUPLING EQUATIONS WITH LIMITS


Consider a design that gives you coupling a coupling chart, and the constraint parameter
is a limiting value, for instance ! <= ! O .
After analysis and preparation of the
coupling chart, you find your coupling
chart as schematically shown to the
right.
Now, the coupling line gives you the
optimized materials, for which ! = ! O ,
but, the RED region gives you the
materials that have a SMALLER value
of the constraint parameter, and
therefore satisfy ! < ! O .

A handy way of thinking about the selection process, contributed by one of the design
groups in the W06 class in their final report.
End of File.

Week Eight: Page 18

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Nine
Winter 2012

OPTIMAL DESIGN WITH COMPOSITES


Reading: Ashby Chapter 11 and 12

As you have probably noticed by now, there are often times when the design statement
requires two conflicting constraints or objectives and no single material can do a good job of
fulfilling the job. In many cases, the job can be better done using a combination of two or
more materials together, essentially requiring a composite approach to the materials
selection. Ashby calls these materials hybrids.
The goal in composites is to get a beneficial set of behaviors out of the components without
taking a hit on some other front. Ashby suggests that the behavior of composites follows one
of the models shown in the figure below:

A: Best of Both
B: Rule of Mixtures (based on volume
fractions, f)

P1C = f1P11 + (1! f1 ) P12


C: Weak link behavior
D: Worst of both

It doesnt happen very often that we can take advantage of the type A behavior, but we can
usually count on something between the rule of mixtures (ROM) and the weak link
behaviors. Just how the composite material behaves depends a lot on the geometry of the
composite structure as well as the properties in question.
Ashby organizes the broad world of composites into four geometrical types, as shown in the
figure below. These are the conventional composite material, the sandwich structure, the
lattice structure, and the segmented structure.
Week Nine: Page 1

W. Warnes: Oregon State University

In addition to developing a taxonomy of composite materials types, Ashby presents (in an


earlier edition) a nice summary table of the potential for functional improvement allowable
by each composite type:

The rest of chapter 11 is more focused on developing new composite materials than in
finding composites to fit a particular design situation. Ill take a slightly different tack for this
issue.
Week Nine: Page 2

W. Warnes: Oregon State University

In designing using composite materials, there are three approaches, which I'll break into two
different categories: MONOLITHIC COMPOSITES, and COMPOSITE STRUCTURES.
Ashby describes these as the simple solution and the hybrid solution.

First, let's look at MONOLITHIC COMPOSITES, or the SIMPLE SOLUTION.


What do I mean by monolithic composites? These are composites for which we treat the
material as a homogeneous piece of material with properties of the bulk composite.
One way of designing with monolithic composites (and by far the easiest) is to have a
comprehensive database of composite materials properties. With such a database, all the
materials are treated just as we'd treat a metal alloy or a polymer. Their properties would be
plotted on the materials selection charts and everything proceeds as we have been doing. We
have a (small) database of composite properties already in the Ashby selection charts,
covering the polymer matrix composites, a couple of cermet materials (WC-Co, for
example), and concrete (a rock, sand, cement composite).
BUT, what if we want to invent our own composite materials, or try to extrapolate the
behavior of a composite that is not listed on the selection charts?
Week Nine: Page 3

W. Warnes: Oregon State University

Start with the Rule of Mixtures:

We can look up the properties of the individual components from the selection charts and
using the rule of mixtures, determine the properties of the monolithic composite. But what do
we use for Vr? The advantage of the composite is that we have adjustable properties, so it
would be nice to leave Vr as something adjustable for our application. How to do this?
Using the rule of mixtures, we can work out the change in the composite properties as a
function of Vr, given the properties of the individual components:

MATRIX MATERIAL
Strength [MPa]
Density [Mg/m3]

1
1,000

REINFORCEMENT MATERIAL
Strength [MPa]
Density [Mg/m3]

100
10

VOL%
REINFORCEMENT
0
0.001

0.9
1

Composite Strength
[MPa]
1
1.099

90.1
100

Composite
Density [Mg/m3]
1000
999.01

109
10

Now we can plot these on the selection charts. How do these sets of materials appear on the
selection charts? Well, they are nicely behaved LINEAR functions of Vr, but we want to look
at them on LOG-LOG plots. Tricky, because of the conversion from a linear plot to a log-log
plot. Look at some examples for a guide: For each example below the change in the first
property (X-axis) is two decades (from 1 to 100). The second property changes by a fixed
amount either increasing (red line) or decreasing (blue line) with volume fraction of
reinforcement. On the left is the linear-linear plot, and on the right is the matching log-log
plot.

Week Nine: Page 4

W. Warnes: Oregon State University

2X-1Y

2X-2Y

2X-3Y

NOTICE that even though the change in the properties is linear with volume fraction of
reinforcement material, the data does not plot that way on a (log-log) selection chart. There is
a significant difference in the selection chart appearance depending on whether or not the
property1-property2 behavior has a positive slope or a negative slope. In addition, the
positive slope case changes its curvature depending on whether there is a bigger change in
the X-axis property or in the Y-axis property. Let's sketch this for a couple of new cuttingedge composites on a selection chart.

Week Nine: Page 5

W. Warnes: Oregon State University

1) PP matrix-Al reinforcement composite


(1/2 x 1 positive)

2) Epoxy matrix-Fir
reinforcement composite
(1/2 x 0 zero)

3) Pb matrix-SiC
reinforcement composite
(1/2 x 2 negative)

4) Pb matrix-Mo
reinforcement composite
(0 x 1 inf)

5) LDPE matrix-SiC
reinforcement
(1/2 x 3 positive)

The next point is that, by doing the optimization process for materials selection, you develop
a selection line. DEPENDING ON THE SLOPE OF THE SELECTION LINE, AND THE
SELECTION REGION you will find out whether or not a particular composite material will
be a better performer for you. With the selection line, you can determine the Vr that would be
needed to optimize the performance of the composite.

Week Nine: Page 6

W. Warnes: Oregon State University

NOTES:
1. The underlying assumption is that the composite will obey the ROM for the
properties of interest. This is not always true, especially near the limits of large and
small Vr, but is a good start for the initial design stage.
2. If the composite line on the selection chart is a straight line, this almost always means
that the composite doesn't help you at all: one of the components will almost certainly
be a better performer. However, when one looks at a different set of properties, the
value of the composite might come into play. (EXAMPLE: SiC reinforced- Al matrix
composites. Compare composite properties on chart 2 against chart 7.)
COMPOSITE STRUCTURES
This is the term I use for the second category of optimal design with composites. Often times
what you find as you start the optimization process is that no single (monolithic) material can
actually give you the performance you want. This is discovered either from doing a "reality
check" with the materials selection indices you have developed, or from the coupling
equations. Your only choices at this point are to either

change the fixed parameter and constraint levels of the design to make it fit the
materials available,
give up, or
use more than one material to do the job.

To pursue the third option, there are a variety of approaches, which range from fairly
straightforward to infinitely complex. Let's start with a straightforward one, and use an
example we've already been working with--the container design for the microbrewery from
week five. In this design we had a basic model of a cylindrical thin walled shape in which
only the wall thickness was a free parameter. We applied two constraints (no failure under a
compressive load due to stacking, and long time with small change in the temperature), found
a 2C1F overconstrained design, developed two performance indices, M1 and M2, and coupled
these together to get the best choice of a MONOLITHIC material to do both jobs. Let's
revisit this with a COMPOSITE STRUCTURE approach.
First, draw the new model:

Week Nine: Page 7

W. Warnes: Oregon State University

then write the MOP: Minimum energy content:

Since this now depends on two materials we have a more complicated design. For the
straightforward approach, let's assume that the two materials we will use are independentthat is, material 1 will ONLY help us with constraint 1, and material 2 will only help us with
constraint 2. If this is the case, then the MOP will be maximized if we can independently
minimize the energy content of each layer:

In essence, we are changing from a one MOP design to a two MOP design. What we end up
with is a materials selection criterion for the strength layer given by

and a materials selection criterion for the thermal insulation layer given by

Week Nine: Page 8

W. Warnes: Oregon State University

Applying these to the appropriate selection charts (chart 18, strength versus energy content;
and the self-generated chart from the week five lecture, thermal conductivity versus energy
content), we can find the materials that best satisfy these needs of this design.
(The answer, which you may convince yourself of outside of class, is to use wood, pottery
ceramics, or steel for the strength layer, and polymer foam or wood for the thermal layer. It is
interesting that historically these containers were made out of pottery and wood long before
stainless steel and aluminum came along!)
The third, and more complicated, approach is to allow each material to provide some benefit
to each constraint. The strength constraint becomes:

The thermal constraint can be simplified if the thermal conductivity of one of the layers is
much larger than the other, to give:

With two constraint equations and two free parameters, we can solve for the free parameters:

And plug into the MOP:

Week Nine: Page 9

W. Warnes: Oregon State University

UGLY! Why? We have crossed the line for an initial assumption (made in the first week) on
which the optimization process we've been using is predicated: the measure of performance
can be separated into the three INDEPENDENT functions of f1, f2, and f3. By doing the
composite structure analysis, we have coupled together the geometrical properties and the
material properties so that these are no longer separable!
Does that invalidate what we've got for our MOP above? NO WAY! It now means the
optimal choice of materials will require a different selection approach, which gets us into the
realm of non-linear, non-separable optimization theory, and a need for another course.
(NOTE: Rule of Mixtures modeling of composites is very rudimentary and not nearly the
while story on composite properties. To learn more about the real composite models,
consider taking the Composite Manufacturing course (ME499, Fall term), or Polymer
Composites (WSE530, spring term of odd numbered years.))
End of File.

Week Nine: Page 10

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Nine
Winter 2012

ENVIRONMENTAL MATERIALS SELECTION


Reading: Ashby Chapter 15.

In the course this term we have looked briefly at environmental concerns in materials
selection using the idea of ENERGY CONTENT as a way of measuring the impact of the
processing of materials on the environment. In Ashbys new edition book, he has further
developed the ideas involved in environmental design. In fact, he is also now offering an
add-on to the CES software called the CES Eco-selector, which provides additional
information for ecological concerns in materials selection for design.
One of the difficulties in considering environmental aspects of materials use in design is that
there is not universal agreement about how to account for green design, or even what
constitutes green design. As well, the interaction between materials and the environment is
a complex one, and really requires a broad, systems-level approach.

One view of the complex interactions surrounding materials and energy consumption is
shown in this schematic. Often seemingly straightforward changes (such as new technology)
result in complex changes in materials use and energy consumption.
Week Nine: Page 11

W. Warnes: Oregon State University

Design for the environment focuses on a time scale of 10 years, while design for
sustainability looks at the longer time scale of 10-100s of years. The later requires a
different set of tools than what we can deal with here (legislative, global, political, social), so
well look at the shorter term problems.

THE MATERIALS LIFE-CYCLE


The first step in thinking about materials and the environment involves understanding the
life-cycle of a material in an application. Ashby breaks the life-cycle of materials or products
into four phases; Material production, product manufacture, product use, and product
disposal.
The following figure schematically illustrates that, at each of these phases, there is a potential
for the consumption of energy and materials as inputs, and for the production of waste (as
either heat or physical waste).

Based on these four phases, we can develop some measures for the environmental impact of
various products and materials in each of these phases. To begin, lets think about the energy
input required to produce a raw material in the first phase.
MATERIAL PRODUCTION: PHASE ONE
Most energy consumed in this life-cycle picture comes at present from fossil fuels, and we
can come up with a figure for the number of MJ of energy consumed per kg of material
produced. This is called the production energy of the material, and can vary widely, for
instance between aluminum and copper.

Week Nine: Page 12

W. Warnes: Oregon State University

Some of the production energy is stored in the material and can be recovered in some way
later on (perhaps in burning the material at disposal). Other materials also contain another
type of stored energy. For instance, polymers are primarily made from oil feedstock which
contains chemical energy, while timber contains solar energy stored in the wood during its
growth. This energy is called intrinsic energy. The Ashby approach is to include this
intrinsic energy in the production energy when it comes from non-renewable sources (like
polymers), but not from renewable sources (like timber). Production energy values range
typically from 10-100 [MJ/kg].
Producing materials also produces unwanted waste materials (pollution), including (but not
limited to) CO2, NOx, SOx, and CH4. Sometimes these wastes can be significant (for instance,
producing 1 [kg] of aluminum from fossil fuel energy produces 12 [kg] of CO2, 40 [g] of
NOx, and 90 [g] of SOx. Other materials dont have as large an impact on the environment
from waste (timber has essentially a negative output of CO2).
In any case, the idea here is to track the energy input and pollution output associated with the
production of each materiala significant and ongoing database effort.
PRODUCT MANUFACTURE: PHASE TWO
Processing materials into usable shapes also involves an energy input, and these energy costs
can be measured and/or modeled. For example, we can estimate the energy needed to melt
materials (as needed in a casting process) from basic properties.

Assuming a 30% efficiency of turning fossil fuels into electrical energy for the melting, we
find

And for most metals and alloys, the melting energy is in the range of 0.4-4 [MJ/kg].
Similar estimates can be made for the energy of vaporization (used in physical vapor
deposition processes) or deformation work (such as rolling or extruding). For metals and
alloys, these numbers are in the range of 3-30 [MJ/kg] (vaporization) and 0.01-1 [MJ/kg]
(deformation).

Week Nine: Page 13

W. Warnes: Oregon State University

PRODUCT USE: PHASE THREE


The amount of environmental impact that occurs during product use clearly depends
significantly on the particular product. For instance, a highway bridge consumes a lot of
energy and material during the manufacturing phase, but requires very little energy in the use
phase. Automobiles consume energy while they are being used as well as while they are
being made. Ashby breaks down products into a set of product classes based on their energy
and materials input, and the loads that they must sustain in use.

Again, we can track the flow of energy, materials, and waste production through the product
use phase and well see some pretty big differences in the performance of different products,
designs and materials.
PRODUCT DISPOSAL: PHASE FOUR
Quantifying recycling is, apparently, quite a difficult operation. There are always energy
costs associated with recycling, and these also contribute to the waste stream. However,
generally recycling a material saves considerably on energy compared to the materials
production energy.
Just because a recycling process is energy efficient, doesnt mean that it is cost efficient. This
often depends on how widely distributed the materials are. For scrap that is locally generated,
during the manufacturing process for instance, the process of recycling is easy and generally
already being done at high efficiency. Widely distributed materials, from junked products,
are harder to collect and recycle, which costs more money to perform the recycling effort.
Many materials are not capable of being recycled at all. Metal matrix composites are an
example. The intimate mixture of a ceramic reinforcement (for instance a metal carbide
powder) within a metal matrix (such as aluminum) cannot readily be separated into the
component metal and ceramic again. As design pushes toward more miniaturization, the
economical separation of the component materials becomes even more difficult.

Week Nine: Page 14

W. Warnes: Oregon State University

Generally, the information about recycling materials comes down to a non-quantitative


indication of whether the material can or cannot be recycled, down-cycled, biodegraded,
incinerated, or landfilled.

ENERGY CONSUMPTION BY PHASE


Generally, it is found in pratice that one of the four phases of the product life-cycle
dominates in terms of the impact on the environment. If we simplify the data to consider just
the energy consumption, we can produce a plot of the energy used by a product during the
different phases of its life.

What the charts above show is that, quite often, one of the phases is the dominate phase,
accounting for as much as 80% of the energy consumption during the entire product life
cycle. Which phase is the dominant one clearly depends on the product, but the implication is
that, in order to achieve significant benefits from redesign or informed materials selection,
we will need to target the correct phase. Making a 20% reduction in the energy use in the
manufacturing of a bicycle doesnt really affect the environmental impact of the bicycle. A
change in materials to one that has a smaller production energy cost will be more likely to
produce results.
The other thing to notice about the figure above is that, when the differences between the
phases are so large, precision in the data isnt important. An uncertainty of a factor of two in
the energy costs wont really change the results, so quibbling about the details isnt important
(yet).

Week Nine: Page 15

W. Warnes: Oregon State University

The process of environmental design, then, starts by figuring out which of the four phases of
product life is the dominant one. What technique we use to attack environmental problems
will depend on the dominant phase.
In thinking about the things weve discussed this term, the techniques for efficient design and
working towards maximum performance are all focused on the PRODUCT USE phase.
Improvements in performance by providing better strength to weight ratio materials, or
optimizing the shape to produce the smallest mass possible for a given application, or
providing the largest thermal insulation for the weight of material used, are all examples of
positive ways to impact the environment through intelligent design.
The other phases require other techniques to make an impact, and some of these are listed in
the figure below.

To impact the material production phase, we need to examine materials that will accomplish
our design goal with a minimum of production energy content. We already looked at this a
bit with the example of the green beverage container, which used the Ashby material
property of energy content to produce materials selection indices such as:

The following plots show the large range of production energy content per kg and per m3 for
materials.

Week Nine: Page 16

W. Warnes: Oregon State University

With the CES software we can make selection charts that incorporate the production energy,
as we have already seen, and in the newest version of the CES software, this can be done
with properties such as the amount of CO2 produced per kg of material. Cool!

Week Nine: Page 17

W. Warnes: Oregon State University

CASE STUDY: DRINK CONTAINERS


You probably suspect by now that I am inordinately concerned about containers for
beverages, but Ill go ahead with this case study from Ashby anyway. At least its not a
cantilever beam!
Thinking about the energy consumption characteristics of beverage containers, we can see
that the main energy consumption happens in the materials production and the product
manufacturing phases.
To get the best material, we want to select the material that minimizes the production energy
per unit of capacity of the container. We can start by looking at the production energies for
the different materials. What materials look reasonable?

Okay, so we have a candidate materials list, with production energies for each. Just for
giggles, lets use as candidate materials the set of materials that are commonly found in drink
containers at present and see how they stack up.
Now we can look up the product manufacturing energies (assuming we have a
comprehensive database), and we get the following data:
Candiate
Material
PET
HDPE
Soda glass
5000 series Al alloy
Plain carbon steel

Production Energy
MJ/kg
84
80
15.5
210
32

Forming
Method
Molding
Molding
Molding
Deep drawing
Deep drawing

Forming Energy
MJ/kg
3.1
3.1
4.9
0.13
0.15

Interesting to note that the forming energy is a small fraction of the production energy for all
these materials (with the exception of the soda glass).
Adding the energies, and dividing by the volume of the container, we can make a list of the
performance of each of these materials in this application. Before we look at the answer,
what is your bet about most environmentally sound material for this application? Notice that
all of these materials can be fully recycled.

Week Nine: Page 18

W. Warnes: Oregon State University

The answers are:


Material

Container

PET
HDPE
Soda glass
5000 series Al
Plain steel

400 ml bottle
1 liter milk bottle
750 ml bottle
440 ml can
440 ml can

Mass
g
25
38
325
20
45

Mass/liter
g
62
38
433
45
102

Energy/liter
MJ/liter

Surprised at the results? Me too!

CASE STUDY: CRASH BARRIERS


Im thinking here about two different types of crash barriers to protect people in automobiles.
First, there are the static barriers that line the roads, particularly around sharp corners or steep
drops. The second are the mobile kind, such as the fender and crash protection on the car
itself.
In the static barriers, the dominant phases of energy consumption are the production and
manufacturing phases, as in the drink containers. Once theyve been put in place they dont
use any energy or create any pollution. We will want to maximize the energy the barrier can
absorb per unit of production energy. The energy absorbed in a crash is given by the area
under the stress-strain curve:

And the materials selection index will be:

The automobile fender is different. Since it is part of the car, it adds to the weight, and thus to
the energy consumed in using the vehicle. The principle phase of energy consumption is that
of product use. Here, we are more concerned with the weight that the fender adds to the car,
so well want to maximize energy absorption per unit weight, giving us:

Week Nine: Page 19

W. Warnes: Oregon State University

The top chart is the one to use for static crash barriers. We want large values of energy per
unit production energy, and we find that the steels are at the top. It also looks like PE and PP
would be good choices. Anyone know what they use for those big plastic containers around
construction sites?
The bottom chart is for moving crash barriers. Here Ti alloys and stainless look okay, but we
see that the polymer materials really look much better.
The point is, different designs call for different measures of performance (where have I heard
that before?), and looking at the energy flow during product life is one way to help
determine the best M-O-P for the J-O-B.

Week Nine: Page 20

W. Warnes: Oregon State University

CASE STUDY: WALL PANELS


One last example and there aint no more. Here we look at the design of environmentally
sound panels, and again, the application affects how we approach the choice of materials for
the design. Lets look at panels that emphasize each step in the product life phases.
MATERIAL PRODUCTION PHASE PREDOMINANT
We want a flat panel to provide a given strength, and be stiff enough to not buckle or deflect
(or vibrate?). The application involves a stationary panel (like the wall of a storage container)
and there are no energy concerns in its use (not a refrigerated container, I guess). From
numerous examples over the course of the last nine and a half weeks, we know that the
performance indices for these panels will be:

where we have used the minimum production energy as our MOP. The selection charts
shown here are the inverse of the M-values given above; we want materials with LOW
values on the plots.
Plywoods and particleboards look pretty good in these plots (bully for K-mart!), while steel
comes out pretty good for the metals. Polymers all show higher energy contents.
PRODUCT MANUFACTURE PHASE DOMINANT
We saw when we looked at the drink containers that the manufacturing energy inputs are
generally much smaller than the production energy content. Choice of material will dictate
the choice of processing, but this has little impact on the overall energy content of the
product. Probably more important is the impact of the manufacturing process on the LOCAL
environment in terms of effluent release in the community. Probably clean manufacturing
practices are the only solution to manufacturing phase dominant products.

Week Nine: Page 21

W. Warnes: Oregon State University

PRODUCT USE PHASE DOMINANT


Now, our panel is in an energy intensive application (a panel for a storage refrigerator for
fine beverages, or a beverage delivery truck side panel, or even a refrigerated beverage
delivery truck side panel!), so we will want a different optimization.
Lets assume a delivery truck side panel, so that stiffness and strength per unit mass become
the important selection criteria. The selection plots are ones weve seen before (again, plotted
as 1/M), and we see that the winners are CFRP, plywood and Be, with Al and Ti a lot better
off than steel.

Week Nine: Page 22

W. Warnes: Oregon State University

In strength based design, CFRP wins handily, and even polymers do better than steels.
PRODUCT DISPOSAL PHASE DOMINANT
This is an area that only recently has seen much interest from manufacturers. Recycling is
difficult because of the contamination materials collect as they are used. This whole area is
an ongoing research focus, and will be for the foreseeable future.
End of File.

Week Nine: Page 23

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Ten
Winter 2012

PROCESS SELECTION: PART ONE


Reading: Ashby Chapters 13 and 14.

Why should we worry about process selection at all? There are several good, practical
reasons to go through the effort of process selection:

Different processes may produce the same component from the same material, but
with different properties. (For example: spray forming (thermal-spray) vs. casting.)
The SIZE OF THE BUBBLE on the selection charts gives a good idea of how much
processing can vary the properties of a given material.
Early consideration of processing can save a lot of money later on.
The [$/kg] material costs are usually only a small part of the total cost of making
something, especially compared to the [$/part] costs of manufacturing processes.

Ashby has pulled together a process database, both as part of the book, and as part of CES.
This is another great tool to have in your design toolbox, as youll see. I want to talk about
the topic of process selection in four parts: Process Types, Process Attributes, Process
Selection Using the Book, and Process Selection Using CES.

PROCESS TYPES
Ashby has done a survey of all the standard (and not-so-standard) processes for materials
manufacturing, and has come up with a pretty useful way of categorizing them. As seen in
Figure 13.2, he breaks them into four categories:
PRIMARY SHAPING PROCESSES: These are the big gunsprocesses like casting,
deformation (which includes forging and rolling), powder processing (for ceramics, and
increasingly, nano-structured materials), and special methods like electro-forming and rapid
prototyping.
SECONDARY PROCESSES: There are only two of these, but they are important: machining
and heat treatment. They are secondary because they are used on something thats already
been through some primary process.
JOINING: which includes welding, adhesive bonding, and fasteners. And finally
Week Ten: Page 1

W. Warnes: Oregon State University

FINISHING or SURFACE TREATMENT: Many of these are cosmetic, but there are
important processes for performance here as well (polishing to get needed flatness or surface
roughness, or coatings to provide environmental protection, for example.)
To become familiar with the different processing types, reading through section 13.3 is
definitely a good idea.

PROCESS ATTRIBUTES
In putting together the process database, Ashby has followed the same philosophy as in the
materials database. Each process is defined by a set of attributes, and, just as with the
materials database, each attribute is filled in for each process. (Note that processes have a
different sets of attributes based on which type of process they are.)
The process attributes is probably the most important part of the process database. A quick
perusal of the database information will make you look like a total expert on processing
$$Ka-ching$$!
The attributes for each type of process are shown in figures 13.3 and 13.4, and the values of
each attribute appear over a range, just as materials properties do. Ultimately, the process
selection will involve matching the attributes of the candidate processes with the attributes of
your design and material. SOhow do we do it?

PROCESS SELECTION BY THE BOOK


Process selection is similar to materials selection in that we want to initially consider ALL
processes as candidates at the start of the design, and winnow down the options as we learn
more about the design requirements. It is an iterative process and should ideally be
considered at each stage of the design, in increasing detail as the design refines.
There are two types of process selection you may want to do. The first is to use an existing
process as a way of eliminating materials, and the second is to use the materials/design to
select a process. These two approaches differ in philosophy, requirements, and outcomes.
In the first case, your company may have an existing process or a specialization in a given
process route. It would be useful to be able to choose materials for the component you are
designing that will take advantage of the processing expertise that you already have, and the
resources that are available.
In the second case, you may not have such processing capabilities, and plan to either develop
them or contract them out. It is important to select the best, most economical process for the
materials you plan to use, chosen from all of the processes out there. Essentially the measure
Week Ten: Page 2

W. Warnes: Oregon State University

of performance for the process selection is going to be minimum cost, though it might
alternatively be minimum time or maximum quality.
The first case is easy, as it fits into our existing system of materials selection, using an
additional selection stage or two to help us choose an appropriate material.
CASE ONE: PROCESS SELECTION AS A MATERIALS CONSTRAINT
There is only one of the process selection charts from chapter 13 that we need to use in this
case, figure 13.22. We may possibly want to look at the others, but we'll save them for later.
Figure 13.22 shows a generic overview of what materials may be fabricated with what
processes. This can be used as a selection stage to eliminate materials not suitable for use
with "your" process. Since it is a generic overview, it should be used with some caution.
EXAMPLE: You are designing a component, and your company specializes in RESIN
TRANSFER MOLDING, which is essentially polymer molding operation. How do you use
this process as a filter to eliminate materials in the selection process?
FIRST: Look at figure 13.22. It
shows the set of materials that
are compatible with resin
transfer molding to be all in the
polymers area, and specifically,

Now we can go to the materials selection part. For example, perhaps the design analysis tells
us to choose materials with an

Week Ten: Page 3

W. Warnes: Oregon State University

Use chart 2, and drop our line down to hit the materials left after the process selection.

CASE ONE: PROCESS SELECTION AS A MATERIALS CONSTRAINT USING CES


Of course, this is much easier to do with the help of CES, and more accurate, since the
computer can track more detailed information about the processes and the materials than we
can present on a graph.
With CES, the first stage in selecting a material can be a process selection stage, which will
strongly limit your materials choices.
Lets assume that our company is in the business of EXTRUSION (which is a deformation
process.) Start by making a new GRAPHICAL SELECTION [NOTE: make sure that you are
selecting from the materials database for your project. I used all bulk materials for this
example.]
For the axis of your choice (the X-axis in my example), click the ADVANCED button, and
then click on the TREES tab. From the pull-down menu, select the ProcessUniverse option.
Now open the Shaping folder, and then the Deformation folder. Double-click on the
Week Ten: Page 4

W. Warnes: Oregon State University

EXTRUSION select PROCESS from the list on the left, and choose the process, or processes
you want to use from the list on the right. [NOTE: you can choose multiple processes by
selecting more than one process from the list.] This will only select materials that can be
EXTRUDED for the selection plot.
For the other axis, choose any material property that you want to--it doesn't have to be
important for your design, but it could be. I chose PRICE. In my selection stage, a lot of
materials passed. I didn't use any additional selection at this point; the plot only shows the
materials that are capable of being extruded.

Now we can go ahead with the rest of the selection process as we are familiar with it.
Perhaps our next selection stage uses chart 2 with a materials selection index of:

Week Ten: Page 5

W. Warnes: Oregon State University

WOW! We end up throwing out a lot of materials, some with better performance in this
application than the chosen materials, but we find out which materials are the best to use,
given the limitations of the process that we will be using to fabricate them.

CASE TWO: SELECTION OF OPTIMUM PROCESS


To perform OPTIMAL process selection, we again let the design requirements drive the
decision process. If we have done the materials selection analysis, we may already have a set
of candidate materials with which to work. The selection process devolves to the task of
matching PROCESS ATTRIBUTES, to the COMPONENT ATTRIBUTES. Each process
has a set of attributes, as weve already seen.
The design requirements will generate a similar list of attributes that the component will need
(for instance, a free parameter determined by the materials selection process), and we look
for a match. As in the first approach, we can either do this by hand or by CES.
There are seven charts that Ashby has put together for process selection. These are:
Figure 13.22: The materials/process matrix, which we have already seen.
Week Ten: Page 6

W. Warnes: Oregon State University

Figure 13.23: The shape/process matrix, which indicates which shapes are possible with
which processes.
Figure 13.24: The mass range chart tells the limits on the size of the part, in kg, that can be
made with any given process.
Figure 13.25: Section Thickness sets out the range of thicknesses that different processes can
achieve.
Figure 13.30: Tolerance: Very useful for knowing how close to net final shape we can expect
the process to get us. Will we need a final finishing stage?
Figure 13.31: The RMS Surface Roughness chart. How good a finish does the process leave?
Last of all is Figure 13.34, the economic batch size chart. How many parts can we expect to
make economically using a given process approach?
The component attributes can be plotted on these charts as either boxes (if the attribute is
within a range of values), or as a horizontal line (if the attribute has a maximum or minimum
allowable value).
Some of these charts will be limiting in terms of processes available, and others will not.
Let's do an example from Ashby involving a cantilever beam...
EXAMPLE: CERAMIC VALVES FOR FAUCETS
We are looking to replace the rubber washers and brass valve seats in standard water faucets
with something that won't wear out (causing leaking) and won't corrode (causing leaking).
The plan is to go with a pair of matched ceramic disks, with finely polished surfaces in
contact. The ceramics have a high hardness and a high resistance to corrosion in water. Some
far-sighted engineers (probably someone who took this course last year) has already
developed Al2O3 faucet valves. The problem is that, when the water temperature changes
rapidly, the Al2O3 cracks due to thermal stresses, either within the ceramic itself, or between
it and the metal making up the faucet. Can we do better?
MODEL: Disk of ceramic.

Week Ten: Page 7

W. Warnes: Oregon State University

MOP: Maximize temperature variation that leads to thermal shock cracking.

CONSTRAINT: Thermal strain in surface of the disk on a rapid temperature change is:

Check out chart 12: /E versus .


The materials of choice
are:

NOW, how do we
make it? Stay tuned to
this same station until
next time...

Week Ten: Page 8

W. Warnes: Oregon State University

ME480/580: Materials Selection


Lecture Notes for Week Ten
Winter 2012

OPTIMAL PROCESS SELECTION: PART TWO


Reading: Ashby Chapters 13 and 14.

Process Selection for the Ceramic Valve for the Faucet


In order to select the process, we need to determine the attributes of the design against which
we will attempt to match the attributes of the process. To do this, we need to know some of
the fixed parameters of the design, and the materials that have been selected from last lecture.
The design statement tells us some of the geometrical parameters of the valves:

Diameter =
Thickness =
Mating surfaces flat and smooth:

Using this information, we can begin to fill out the attributes that are important for process
selection.
VOLUME: easy...

SIZE (measured in units of weight):

NOTE: For many calculations, we can get by with a generic "average" density of 5000
[kg/m3] (or 1000 [kg/m3] for polymers). We don't have to be too precise with these
calculations.
Week Ten: Page 9

W. Warnes: Oregon State University

MINIMUM SECTION THICKNESS:

OKAYNow we put all this together to help us plot the correct stuff on the process
selection charts. I made the following table to help keep track of it all:

Design Attribute
Volume [m3]
Section Thickness [mm]
Tolerance [mm]
Roughness [mm]

Material
SiAlON
SiN
SiC
Zr2O3
RANGE

Density (min)
[kg/m3]
3200
2370
3000
5000
--

Value
1.60 x 10-6
5
0.02
0.1 m

Density (max)
[kg/m3]
3300
3250
3200
6150
--

Size (min)
[kg]
5.12 x 10-3
3.79 x 10-3
4.80 x 10-3
8.00 x 10-3
3.79 x 10-3

Size (max)
[kg]
5.28 x 10-3
5.20 x 10-3
5.12 x 10-3
9.84 x 10-3
9.84 x 10-3

Now we apply these to the Process Selection Charts using a multiple selection stage
approach. Since the candidate materials in this case are all of the same class (ceramics), we
can lump them together and use the range of values from all four materials to do our search.

Week Ten: Page 10

W. Warnes: Oregon State University

Process Selection Stage 1: Materials Compatibility Chart (Figure 13.22): Already down
to eight processes, only three of which are primary shaping processes.

Process Selection Stage 2: Process-Shape Chart (Figure 13.23):


This shape is slightly more complicated than a simple prismatic circular shape, but not too
much more, so we'll say its a non-circular prismatic shape.
None of the primary processes are eliminated based on shape.

Week Ten: Page 11

W. Warnes: Oregon State University

Process Selection Stage 3: Process-Mass Range Chart (Figure 13.24):


We plot our mass range on
here:

Powder methods are just on


the edge. Do we eliminate
them? Since they are so
close, and we are still early in
the design process, we better
keep them in.

Process Selection Stage 4: Process-Section Thickness Chart (Figure 13.25):


Nothing eliminated here either.
Process Selection Stage 5: Process-Tolerance Chart (Figure 13.24):
Well, there goes powder
processingits way out of
the needed range. Down to
two

Week Ten: Page 12

W. Warnes: Oregon State University

Process Selection Stage 6: Process-Surface Roughness Chart (Figure 13.31):

Whoops! Now theres nothing left in the primary processes! What do we do now?
Fortunately, all of the finishing processes can get us into the right surface roughness range,
so we just need to check them against the materials compatibility chart to make sure we are
alright there, and it looks as if grinding, lapping, and polishing are compatible with ceramics.
Because we have to introduce a secondary finishing process to meet the roughness spec,
maybe we can reconsider powder methods. We can use the same finishing step to provide the
powder methods with the needed tolerance, as well as getting the surface roughness correct.
This leaves us with three potential approaches: powder methods, electro-machining, or
conventional machining, each followed by a secondary finishing step.

Last of all is Figure 13.34, the economic batch size chart. A quick check of the batch size for
the three primary shaping processes tells us what we have to do. If we are only making a 100
to a few 1000 parts, we can use one of the two machining processes, but for any number of
parts larger than that well want to go with powder processing.

Week Ten: Page 13

W. Warnes: Oregon State University

PROCESS SELECTION USING CES


This is a case study out of the on-line literature that comes with CES. The design goal is to
build a handle for opening and closing the windows in a house or building.

The materials selection analysis has already been done, and the cheapest material for the job
turns out to be a zinc-based alloy. We want to determine the process. Putting together our
design and materials attributes gives us a list of:

Material Class: Non-ferrous Metal (Zn-based Alloy)


Process Class: Primary, discrete
Shape Class: 3D-solid
Mass: 0.1 [kg]
Minimum Section: 3 [mm]
Tolerance: 0.2 [mm]
Surface Roughness: 2 [um]
Batch Size: 1,000,000

Because we are looking to economically make a lot of these handles (on the order of
1,000,000), we want to try to get everything out of a primary process, without resorting to a
secondary operation. The handles must have a good surface finish in order to operate well.
To start the process selection in CES, we need to start with a NEW PROJECT. First thing,
we need to change the "Select From" to use "PROCESS UNIVERSE: Shaping Processes"
rather than "MATERIALS". Now we will be using the process database to plot the selection
charts. It would be a good idea to take a moment to look at a typical process data record.
There is a lot of information here about a lot of different processes--very impressive
database!
We want to build a selection chart, which happens in just the same way as with the materials
selection charts, except that the axes are going to be different. For this example, there are five
selection stages.

Week Ten: Page 14

W. Warnes: Oregon State University

STAGE ONE: Mass Range (Size) versus Primary Process = TRUE

The process property of "PRIMARY" indicates whether or not the process can be used as a
primary (or stand alone) processing route. It is of a Boolean-type, and when you initially plot
this property, it will show two boxes on the X-axis, labeled "TRUE" and "FALSE". You can
choose to display only one of these values (as I have done) by using the magnifying glass
icon, and clicking on the value you want to display. In the selection plot, a box selection has
been used to grab all the processes that are capable of producing a part in the size range of
interest (around 0.1 kg).

Week Ten: Page 15

W. Warnes: Oregon State University

STAGE TWO: Tolerance versus Shape Class = 3D-solid

The box selection chooses the processes capable of a tolerance of 0.2mm or better.
STAGE THREE: Roughness versus Process Class Discrete = TRUE

The Process Class of DISCRETE describes whether or not the process makes discrete
(individual) parts, or many at a time.
Week Ten: Page 16

W. Warnes: Oregon State University

STAGE FOUR: Minimum Section versus Materials Class = Non-Ferrous Metals

STAGE FIVE: Economic Batch Size versus Process Class Discrete = TRUE

Here we are looking for the cheapest way of producing the 1,000,000 parts we want. Only a
few processes are economically viable at that batch size.

Week Ten: Page 17

W. Warnes: Oregon State University

The final list of candidate processes that come through all the selection stages are:

Automated machining
Die pressing of powder
Powder metallurgy forging
Powder injection molding
Pressure die casting

According to this analysis, the important attributes that really narrowed the selection were
tolerance, roughness, and batch size. To determine more, we need to look more fully in detail
at the costs of the candidate processes. The CES software has a complete set of ECONOMIC
ATTRIBUTES for each process as well. This information can be used for more detailed cost
analysis, a subject that we won't get into here.
End of File.

Week Ten: Page 18

W. Warnes: Oregon State University

You might also like