You are on page 1of 145

Availability Simulation Modeling in a

chemical manufacturing company


A feasibility study by means of the development of a
simulation model for the Resin plant of the Lexan
production train of GE Plastics.

J.A. van Luijk


July 2003

Faculty of Technology, Policy & Management


section Energy & Industry

Supervisors:
Ing. J. Bruijnooge
Dr. Ir. M. den Hengst-Bruggeling
Dr. Ir. P.M. Herder
Prof. Dr. Ir. M.P.C. Weijnen

A failure a day keeps the profits away


R.H. Adler

ii

Preface
This report is the result of the project carried out for my Master of Science thesis for the
study Systems Engineering and Policy Analysis, at the faculty of Technology, Policy and
Management of Delft Technical University. The study has been performed at GE Plastics,
Bergen op Zoom, the Netherlands.
As the title indicates, this study has covered the implementation of availability simulation
modeling throughout the organization of GE Plastics. In order to be able to draw
conclusions an availability simulation model has been developed for one of the plants of
the Lexan production train. Availability simulation modeling has a technical and an
organizational aspect that both need to be taken into account in order to successfully
implement the tool. Therefore this study and this report can basically be divided into two
parts.
The first part deals with the actual development of the simulation model and results in
technical conclusions and recommendations concerning the content and the general use
of the model (chapter 5)
The second part is an identification of all organizational prerequisites to successfully
implement availability simulation modeling throughout the organization of GE Plastics
(chapter 6). For this part I have conducted many interviews with people from other
companies working in the same field, which I hereby would like to thank.
Also my thanks goes out to those who have advised and assisted me over and over during
this project to help me to successfully complete it.
Hanna van Luijk,
Bergen op Zoom, July 2003.

Summary
Availability simulation modeling has been described in literature in much detail.
However, in industry it is still hardly applied. This study identifies the reasons for this
gap and describes the prerequisites to overcome it. This has been made possible by
carrying out a feasibility study at GE Plastics where an availability simulation model has
been developed for the resin plant of the Lexan production train. This feasibility study
covers two areas: a technical part and an organizational part. Results obtained for both of
these parts are described for the situation at GE Plastics and for a general situation.
Reliability and maintenance engineering are very important for a plant availability. The
better the reliability and maintainability the better the availability of a plant is. From a
benchmark in the chemical industry it can be concluded that most companies are now in
the second stage (of three) of maturity with regard to the maintenance and reliability
engineering policy. Availability simulation modeling belongs to the third phase.
There are several types of availability simulation modeling. Two most common are
Markov modeling and Reliability Block Diagram (RBD) analysis. The latter has been
used in this study due to its transparent character. The simulation engine was a Monte
Carlo engine because the situation was to complex to analyze in another way. The
functional objectives set for the model in this study have been:
Identification of equipment logic, in which way are the various pieces of
equipment interrelated and dependent of each other?
Identification of critical equipment, either due to frequent failure or due to high
maintainability.
Influence of redundant equipment and integrated buffers on (system) reliability.
Identification of knowledge and data available for reliability engineering issues.
After a brief investigation an appropriate software tool has been selected that combined
all these factors (RBD, Monte Carlo simulation and the possibility to meet the functional
objectives).
In this study a model is built for the resin plant of the Lexan train in Bergen op Zoom.
This plant forms a polycarbonate from the monomer BPA and phosgene.
The physical setup of the model is determined by use of the EFDs and PFDs,
information from the control panels and is validated by expert knowledge.
For the determination of the failure frequency distributions EMPAC (work order
registration system) has been used. This data system does not contain information on the
repair times of equipment and therefore expert knowledge is used to estimate these times.
When the model was completely populated it has been run and validated by the dent
administration (registration system of lost production) and expert knowledge. This finally
resulted in acceptance of the model.

ii

From the model runs and the validation the following conclusions can be drawn for
simulation modeling in general:
Select a type of software that is more capable of modeling process properties
like process flows and intermediate storage.
Close the gap between the two different types of measuring maintenance work
and output dents. Or, at least, register what the differences are and
consequently apply one measure within a system.
Use the CMMS to its full extent, meaning:
Make sure the numbering of equipment is always consistent with the
existing rules. This also means correct (in)active selection in the asset
sheet.
Clearly register modifications performed on equipment in the CMMS.
State whether the asset is shut down or not during the performed
action.
Include repair times in the system
Make sure the right work is under the right asset number. Add an asset
number for the storage of work orders on general equipment.
Make the system accessible for reporting and data extractions. It
should be possible to select fields, copy them to another application
and work with them.
From the example case it has become obvious that building an availability simulation
model is a very knowledge and labor intensive occupation.
GE Plastics is part of a large divisionalized company and therefore works with a high
degree of standardization of work processes. To be successfully implemented availability
simulation modeling should fit in these processes. These are the so-called Six-Sigma
processes. Because availability simulation modeling is applicable on various fields within
a manufacturing company it is best incorporated in a general department that can do work
for various fields. Here the engineering department is chosen. In this department the
modeling is close to two other functions: maintenance optimization and warehousing.
To assure a successful implementation in GE Plastics the projects in which modeling is
implemented first should be selected carefully. The model needs to have high added
value in the projects and show results and payoff fast. The initial build will at this
moment be performed by an external expert. Also people involved within GE Plastics
(engineering and plant specific engineers) need to be trained. Lastly there needs to be
attention for evaluation of the first projects in order to maintain and increase support for
the implementation of simulation modeling.
When these results obtained at GE Plastics are translated to a generic form that is
applicable to any company within chemical industry, the following conclusions and
recommendations can be made:
Availability simulation modeling is not widespread throughout chemical
industry. It is regarded as a next generation tool that will be used after the
structural improvement programs.

iii

Availability simulation is now mainly performed by external consultants


because it is a knowledge intensive activity, which takes a lot effort to buildup.
Laws and regulations become more and more strict and require more and
more quantified insight in the plants and processes. This will require more and
more simulation modeling.
It is technically feasible to implement availability simulation modeling in a
company when it has fulfilled the following prerequisites:
o There is a CMMS used at a level that it provides adequate accurate
data to fill the model.
o There is an independent database to validate the model.
o An appropriate type of software is available.
It is organizationally feasible to implement availability simulation modeling
when a company takes the following aspects into account:
o Availability simulation modeling should fit into the business processes
within a company.
o The model should be owned by an appropriate part of the organization.
This depends on the function the model should fulfill.
o Staff should be trained.
o A process should be developed for implementation of the tool, which
at least contains guidelines for which projects to select for
implementation and how to evaluate the projects.

iv

PREFACE ...................................................................................................................................................... I
SUMMARY.................................................................................................................................................. II
1

INTRODUCTION............................................................................................................................... 7
1.1
1.2
1.3
1.3.1
1.3.2
1.4
1.5
1.6

BACKGROUND .............................................................................................................................. 7
PROBLEM STATEMENT AND RESEARCH GOAL ............................................................................. 10
THEORETICAL BACKGROUND ..................................................................................................... 11
Relevant definitions .............................................................................................................. 11
Previous research ................................................................................................................. 12
RESEARCH QUESTIONS ............................................................................................................... 14
RESEARCH APPROACH ................................................................................................................ 14
ARCHITECTURE OF THE REPORT ................................................................................................. 17

WHAT IS RELIABILITY ENGINEERING?................................................................................ 18


2.1
AVAILABILITY, RELIABILITY AND MAINTAINABILITY ............................................................... 18
2.2
MAINTENANCE POLICIES IN THE CHEMICAL INDUSTRY ............................................................... 21
2.2.1
Three stages of maturity ....................................................................................................... 21
2.2.2
Companies evaluated............................................................................................................ 22

MODELING...................................................................................................................................... 26
3.1 GENERAL MODELING ......................................................................................................................... 26
3.2 AVAILABILITY MODELING.................................................................................................................. 27
3.2.1
What can availability simulation modeling add? ................................................................ 27
3.2.2
Markov modeling.................................................................................................................. 27
3.2.3
Reliability Block Diagram .................................................................................................... 28
3.2.4
Simulation engines................................................................................................................ 30
3.2.5
Which model and simulation engine to use?......................................................................... 30

AVAILABILITY SIMULATION MODELING IN GE PLASTICS............................................ 31


4.1
EVALUATION PREVIOUS PILOT PROJECTS.................................................................................... 31
4.1.1
BPA1, Bergen op Zoom ........................................................................................................ 31
4.1.2
Chlorine plant, Burkville Alabama USA............................................................................... 31
4.1.3
Lexan Plant, Cartagena Spain........................................................................................... 32
4.1.4
General conclusion pilot projects......................................................................................... 32
4.2
OBJECTIVES OF THE MODEL........................................................................................................ 33
4.2.1
Functional Objectives........................................................................................................... 33
4.2.2
Business objectives of the model........................................................................................... 34
4.3
INVESTIGATION AVAILABLE SOFTWARE ..................................................................................... 35
4.3.1
Factors determining the choice of software.......................................................................... 36
4.3.2
Choice of software package.................................................................................................. 37

MODELING EXAMPLE CASE ..................................................................................................... 38


5.1
5.1.1
5.1.2
5.2
5.2.1
5.2.2
5.2.3
5.3
5.4
5.4.1

DESCRIPTION LEXAN PRODUCTION TRAIN AND RESIN PLANT ................................................. 38


General introduction Lexan production train.................................................................... 38
Description of the Resin plant .............................................................................................. 40
SELECTION OF DATA TO DETERMINE FAILURE FREQUENCIES ...................................................... 41
Essential Data retrieval........................................................................................................ 41
Database selection................................................................................................................ 41
Database usage..................................................................................................................... 43
SELECTION OF MODELED SECTIONS ............................................................................................ 44
GENERAL MATHEMATICS AND STATISTICS ................................................................................. 45
Failure frequency functions.................................................................................................. 45

5.4.2
5.5
5.6
5.6.1
5.6.2
5.7
5.7.1
5.7.2
5.8
5.8.1
5.8.2
5.8.3
6

INVESTIGATION ORGANIZATIONAL PREREQUISITES .................................................... 64


6.1
6.2
6.3
6.3.1
6.3.2
6.4

Repair time functions............................................................................................................ 46


BUILDING THE MODEL ................................................................................................................ 47
VALIDATION SIMULATION MODEL ............................................................................................. 48
Validation by dent administration ........................................................................................ 48
Validation by expert knowledge............................................................................................ 52
EXAMPLE OF MODEL USE............................................................................................................ 58
Extension of time between shutdowns................................................................................... 58
In-line cleaning M-316 steam jets ........................................................................................ 60
CONCLUSIONS AND RECOMMENDATIONS ................................................................................... 60
General insight ..................................................................................................................... 60
Conclusions from examples .................................................................................................. 61
General simulation modeling ............................................................................................... 61
ORGANIZATIONAL CHARACTERISTICS GE PLASTICS .................................................................. 64
ANALOGY PROCESS MODELING .................................................................................................. 65
PREREQUISITES FOR SIMULATION IN RELIABILITY ENGINEERING ............................................... 65
Business processes................................................................................................................ 66
Physical embedding in the organization............................................................................... 68
HOW TO REACH THE DESIRED STATE? ........................................................................................ 70

CONCLUSIONS AND RECOMMENDATIONS .......................................................................... 73


7.1
GE PLASTICS ............................................................................................................................. 73
7.1.1
Conclusions .......................................................................................................................... 73
7.1.2
Recommendations................................................................................................................. 74
7.2
GENERAL ................................................................................................................................... 75
7.3
FURTHER RESEARCH .................................................................................................................. 76

REFERENCES ........................................................................................................................................... 77
APPENDIX A

LIST OF ABBREVIATIONS................................................................................... 79

APPENDIX B -

SUPPLIERS AND SOFTWARE PACKAGES ...................................................... 80

APPENDIX C

IDEF0 DIAGRAM LEXAN TRAIN..................................................................... 84

APPENDIX D

RELIABILITY BLOCK DIAGRAM RESIN PLANT .......................................... 92

APPENDIX E

SELECTION MODELED SECTIONS................................................................. 106

APPENDIX F

INPUT FUNCTIONS SIMULATION MODEL................................................... 112

APPENDIX G

MODEL SIMULATION OUTPUT....................................................................... 117

YEARLY SHUTDOWN .............................................................................................................................. 117


Transportation Weighing section...................................................................................................... 117
Precipitation Section......................................................................................................................... 120
Drying Section Line 1 ....................................................................................................................... 123
Drying section Line 2........................................................................................................................ 126
4 YEARLY SHUTDOWN ............................................................................................................................ 129
Transportation weighing section ...................................................................................................... 129
Precipitation section ......................................................................................................................... 132
Drying section line 1......................................................................................................................... 135
Drying section line 2......................................................................................................................... 139
APPENDIX H

OUTPUT DENTS 2002........................................................................................... 142

vi

Introduction

1.1

Background

Management of chemical plants is looking for ways to increase their insight in the plants
and help to make business decisions supported by data from the plants. Often,
management does not have the right knowledge and experience with these technical
plants to make the right decisions without any supportive data. It cannot predict with
certainty, for example, which investments will be very profitable for the company in the
future. From a financial business perspective the availability of a manufacturing plant is
interesting because it is a critical factor for the actual total production capacity of the
plant and its financial returns. The reliability of the plant is also an aspect that determines
whether the output of the plant is as planned.
Within GE Plastics, maintenance has therefore become more and more the focus of
management as being able to contribute in a positive way to production and profitability.
From a maintenance engineering point of view reliability is of interest because it
determines the amount of maintenance to be performed. In addition, maintenance should
be properly planned and executed, supported by the appropriate date. To be able to make
decisions dealing with planning of (preventive) maintenance (what, on which pieces of
equipment), GE Plastics believes that a availability/reliability simulation model should be
built. The idea of developing an availability model has come specifically from the
maintenance department within the GE Plastics Lexan business. The model would
support an increase in the efficiency and effectiveness of planned preventive and
corrective maintenance and thus achieve a higher reliability of the plant and less output
loss.
The model is supposed to give insight into what results will come from certain
management decisions (investments). These decisions can be in various stages of plant
life cycle, from design thru stretches and modifications to maintaining existing plants.
In a graphical representation the hierarchical decomposition of the set objectives within
GE Plastics will be as in figure 1.1.

Corporate Priority

To maximize nett income

Otimize product throughput and equipment reliability


at optimal cost

Maintenance priority

Development of reliability/maintenance resources


allocation optimization model

Maintenance strategy

Maintenance tactic

Execution of tactics determined with model

Fig. 1.1: interrelating corporate priorities with maintenance tactics (Campbell, 2001).

Besides the more general business objective of profitability improvement there are some
other specific issues that can be addressed by using more thorough reliability modeling,
and thus increasing reliability.
First, the model will increase the insight in how various detailed maintenance strategies
and tactics will affect reliability.
A second deliverable of increasing the reliability of the plant is an EHS aspect. During a
certain time period, there will probably be fewer people in the plant working and thus the
probability for personal or environmental accidents will decrease. The decrease of
incidents is achieved by the decrease of risk of breakdowns, which is an important factor
of the license to operate.
Third, the rules and regulations set by government are also becoming stricter. The Seveso
directory of 1982 states that every operating company should produce a safety report
(Seveso II directive [96/82/EC]). The Dutch implementation of this directory states that
for every plant there should be quantified scenarios of how an accident might occur and
what the consequences are (CPR 20). This can also be determined and analyzed by
simulation. Hence, an increase of the use of simulation can be expected.
Fourth, throughout the organization of GE all quality assurance and improvement
projects are performed by a Six-Sigma process. In this process the goal is to achieve a
probability of 3,4 failures in a million actions. Two Six-Sigma processes exist. One for
the development of new processes, and one for the improvement of existing processes.
The process for the design of new situations is the Design For Six Sigma (DFSS) and
exists of 6 tollgates that need to be fulfilled going thru the project. These tollgates are:
project opportunity identification, Scope definition, Conceptual design, basic
engineering, EPC and commissioning and startup, and finally Operation and project close

out. These tollgates all can be subdivided in smaller steps, which will not be further
explained here.
The improvement process is called Define Measure Analyze Implement and Control
(DMAIC). From the name the five steps included in this process can be easily identified.

S ix S igm a

D FS S

D M A IC

Tollgate 1:
P roject oppertunity
identification

Define

Tollgate 2:
Scope definition

M easure

Tollgate 3:
C onceptual design

A nalyze

Tollgate 4:
Basic engineering

Im prove

Tollgate 5:
E P C , com m issioning,
startup

C ontrol

Tollgate 6:
O peration, project close out

Fig. 1.2: Six-Sigma processes decomposed

As the name suggests Six-Sigma processes are very much based on statistics. Since there
is very little possibility to measure reliability and experiment with it in another way than
simulation, this tool can play an important role in these processes.
Besides the GE Plastics business rationale of developing and using a reliability
simulation model, literature confirms all supportive arguments: a proper use of the model
will increase the knowledge of the plants and their processes, the equipment involved,
part and system reliability and eventually it will enable to increase the profitability of the
company. RAM (Reliability, Availability and Maintainability) is considered one of the
two most significant areas for profitability improvement (Williams, 2001). Moreover,
simulation modeling will contribute to the increase of safety and environmental
performance, which is an important factor in maintaining the permit to operate. The
model can achieve this by supporting several (management) decisions with real data.
Very important for the actual application in GE Plastics are studies on the
implementation in industry. These have only been made up to a certain stage. The
implementation and recognition of reliability as an important (business) issue has been
made by industry and described in various publications (E.g. Jones et al., Blann). These
mainly focus on the change towards structural programs to improve reliability and

availability and not yet on a advanced tool as simulation modeling. Main research that
has been performed on the application of availability simulation in an industrial company
has focused on only one project and the results of that particular project. Some have also
tried to give a general guideline for which type of modeling to use (Cochran, 2001,
Dekker, 1995). However, no studies are available that describe the practical problems
encountered when first implementing simulation modeling in an industrial organization.
This study has been performed in order to find these practical problems and give some
guidelines how to deal with them.

1.2

Problem statement and research goal

Extensive, theoretical research has been performed by many authors on the technical
possibilities of and achievements that can be made by availability simulation. This will be
further detailed in the next section. However, it is not yet used in the chemical industry as
a standard tool. Therefore GE Plastics cannot copy the situation from another company.
A large gap exists between the theoretical state of the art of modeling and the practical
implementation in chemical industry. Aim of this research, therefore, is to identify the
potential technical and organizational problems and the prerequisites to resolve this, by
actually building and applying such a model in industrial practice.
This report will deal with the opportunities for the implementation of availability
simulation modeling in GE Plastics. These opportunities will be identified by evaluation
of a simulation model to be built for the resin plant of the Lexan production train1. On
the basis of this study the decision can be made to implement availability simulation
modeling on a regular basis.
By performing the feasibility study advise will be given whether to implement
availability simulation modeling at a large scale within GE Plastics, or not. This study has
two aspects: a technical feasibility study and a study how implement this tool in the
business processes and in the organizational structure. After the study at the company
level of GE an evaluation will be made whether these results can be copied to any other
company in chemical industry.
The different types and levels of objectives and their relation can be illustrated by the
following graph:

Lexan is a polycarbonate plastic that is produced by operation of several chemical plants and a
compounding plant.

10

Decision support in
maintenance strategies

RE/maintenance
background

Modeling
in RE

Organizational prerequisites
for implementation

Modeling types
and software

Benchmark in
Chemical industry

Simulation model
For the resin plant

Evaluation other
pilot projects (GEP)

Technical insight and


recommendations

Organizational insight
and recommendations

Fig. 1.3: Levels and objectives of the study

1.3

Theoretical background

1.3.1 Relevant definitions


In this study there are several important terms that need to be further defined.
First word that requires a detailed definition is availability itself. What is exactly meant
by this word?
Availability is defined as the ability of an item to perform its required function at a
stated instant of time or over a stated period of time (BS4778, 2000).
Availability is determined by the reliability and maintainability of the equipment. The
army is the authority in the field of reliability engineering, hence the U.S. military
definition is given:
Reliability is the probability that an item will perform a required function without
failure under stated conditions for a stated period of time.
From this definition certain things can be concluded:
Reliability is determined by a statistical function that states the probability of an event
occurring.
Reliability is only applicable on a certain function that has been stated in advance. Thus it
is very important that all items will be given a certain function.
The determined reliability is only valid for a certain period of time and for a specified
condition. If these specifications change, as a result the reliability will also change.
Obviously, before working with the reliability these specifications should be set.
The word failure needs to be explained. A common definition for failure is:
A failure is the termination of the ability of an item to perform a required function.
There are two types of reliability that should not be confused with each other. The first
type is part reliability and is only applicable on a single piece of equipment. Part
reliability can be determined by using statistical methods on the failure rate data.
However, if several pieces of equipment interact and form a system (like in a chemical
plant), the term system reliability is applicable. This is the reliability of the overall
system, which is determined by the part reliability of all pieces of equipment involved
and the way they interact. System reliability is very important because it appears that

11

when a large number of reliable components are combined into a large structure, the
result is not necessary a reliable structure (Avontuur, 2000).
Part reliability
reliability estimation
(part reliability is estimated
by statistical methods from
failure rate data)

System reliability
reliability prediction
(reliability is predicted by
system architecture and
part reliability)

Fig. 1.4: relation between part- and system reliability

Besides these theoretical definitions the practical use of the term reliability is important.
Reliability in GE Plastics is practically defined as the unexpected downtime caused by a
unit or plant not performing as planned.
Because this research project has been started from a maintenance point of view and the
method to increase reliability is considered to be maintenance, it is also useful to take a
closer look at the definition of maintenance. Because the noun maintenance is derived
from the verb to maintain, and they are closely related, this definition is also given.
To maintain: cause to continue (Oxford)
To keep in an existing state (Webster)
Maintenance: Ensuring that physical assets continue to fulfill their intended
functions.
From both definitions it can be concluded that it is very important to know the specified
functions which the equipment should fulfill, only then can be determined whether
actions should be taken (maintenance) to maintain the piece of equipment in the desired
(specified) state.

1.3.2 Previous research


Reliability and availability issues have been the subjects of a large amount of studies in
recent years. The fields reliability and availability and narrowly related issues can be
applied on are very diversified. Until now theoretical research has mainly gone deeply
into one of these fields in order to optimize it by building large and complex
mathematical models. Examples of these fields and research performed are: optimization
of maintenance in life cycle process design (Pistikopoulos, 1998), the role of materials
selection in the optimization of plant costs (Sommariva, 2002), the minimization of
environmental risk by maintenance strategies (Pistikopoulos, 2000), reliability
optimization in process synthesis (Goel et al, 2002), and many more.
In the optimization studies of these various fields several times simulation modeling is
introduced as a useful tool. This tool has been subjected to research studies itself because
several types of simulation modeling exist. This research tries to answer the question
which type of modeling and simulation is suitable for which simulation, mostly by
performing an example case. More extensive background on this topic can be found in
work from Chisman (1997), Korver (1994), Fritz & Bertsche, van Rijn (1998) and more.

12

In chapter 3 of this report two common types will be described and the most appropriate
one for this study will be selected.
Because there are so many studies performed on the various fields of application and the
various types of (simulation) modeling and supportive software, it is very hard to oversee
the total field of reliability and availability simulation modeling. This can raise the
question whether there could be integration and classification of some of these types, for
example like has been described by Ramesh and others (1998). The variety demands
structuring before the choice for an actual type can be made. In this study also various
types of software for a certain type of modeling will be compared.
In a company that operates chemical plants many (business)activities are required.
Several management systems will run various parts of the organization. The management
system used for running a chemical process is called process management. This process
management can be divided into activities of several levels that together will assure an
optimal performance of the chemical process. The various activities are:
Planning
Scheduling
Supervisory control
Monitoring & diagnosis
Optimization & regulatory control
This hierarchical decomposition that has been made for process management can actually
be applied on all activities within a process plant, thus also maintenance activities can be
planned and executed with the same steps.
The lower the activities are in this hierarchical chain, the less they are related to the
activity maintenance management. However, to do a good job on the maintenance
management a thorough insight in the activities at the bottom of the hierarchy and the
data and information at this level is needed. Availability modeling can make the link
between these lower and upper levels in the hierarchy. These lower levels will provide
the input of the model that in its turn will contribute to the development of optimal
maintenance management.
Availability modeling will be the tool that can guide this maintenance management, for
example when it concerns capacity planning and investment decisions on equipment.
This can be considered one of the reasons why it is useful to develop a reliability model
for the Lexan production train.
The reason an availability simulation model is needed and no theoretical method can be
used (like analytical analyses) is that the studied system is too complex to achieve the
insight in another way. It is not possible to obtain quantified expectations for up- and
downtime or to even do thoroughly supported qualitative statements in another way than
statistical analysis of plant data. This is very extensive work for complex systems. Only
subjective judgment can be given by operating and maintenance personnel.
Besides, a model provides the opportunity to experiment with several equipment layouts,
and their influence on reliability, which would be too expensive to do for real in the plant.

13

There are several types of availability simulation models. Two most commonly used will
be discussed in chapter 4. The choice of which to use in this study will be made
afterwards.
The number of equipment and their interdependence determines the complexity of the
studied system. There are several possibilities for their interrelationships (series, parallel).
It is not difficult to identify all the equipment involved, however the difficulty will be to
reduce them to an amount where the model will give enough insight in the plant and it is
still easy to work with.

1.4

Research questions

From the problem statement and the theoretical background it can be concluded that the
main question that needs to answered in this study is:
What are the general technical and organizational prerequisites and guidelines for the
useful application of availability simulation modeling in a chemical manufacturing
company like GE Plastics?
This general question of course can be supported by several lower level questions. These
are:
What is the state of the art of reliability and availability modeling in literature?
How widespread is the implementation of reliability and availability modeling in
industry?
Is it possible to build a useful model in GE plastics?
o What are the technical pitfalls and prerequisites?
o What are the organizational pitfalls and prerequisites?
How should the process of implementation of simulation modeling be?
Can this example case be generalized for any company in chemical industry?

1.5

Research approach

In the following chapters there will first be a more thorough description of the terms
reliability, maintainability and availability and how and why they are interesting for a
manufacturing company. Their interrelation is described and how certain management
strategies will influence them. After reading this chapter it should be clear what is meant
by these terms, why they are interesting for a company, how they interact, and what the
objective is for this project.
In order to place the development of the model in a more general scope an industry
benchmark is performed to evaluate how far other chemical companies are with their
development of reliability engineering and reliability simulation. Can GE Plastics learn
from the experiences of other companies? To answer this a thorough knowledge of
common maintenance strategies and perception of reliability engineering within industry
is necessary. Also the development of those strategies throughout the years will be
explained. A benchmark will be made of how other companies deal with related topics in
reliability engineering and how widely the reliability simulation modeling technique is
spread throughout chemical industry.
14

When it is clear in which type of maintenance and reliability engineering strategy


availability simulation modeling fits there will be a more extensive description of
modeling. Attention will be paid to modeling in general and two specific availability
simulation modeling types.
In order to be able to identify the prerequisites for implementation a feasibility study for
is performed within GE Plastics. It will be feasible to implement when it is technically
possible to build the model and when this way of working fits in the organization.
Important here is also the evaluation whether the criteria and prerequisites in this study
are applicable in general.
The technical feasibility study is performed first by actually building a simulation model
for a plant of GE Plastics.
This study should provide information on data, model types and software. With regard to
data it should be clear why certain data is used, whether the data are accurate enough to
be used and where the data can be retrieved. With regard to the model types the choice
for a certain type must be clear. There must be an investigation of which software can be
used to build the simulation model and the choice for a certain package should be clear.
An evaluation needs to state whether the right choice has been made for model type,
software type and data retrieval and usage. Does the built model actually provide the
expected insight?
After the development of the model, the main research focus in this project will be the
implementation of the use of this model in the organization of GE Plastics. To achieve
this goal there needs to be a thorough insight in the organization as it is now. Also all
possible fields of application for the model should be well known. By combining these
two facts there can be an initial plan for how the model should be used, who should be
responsible for it, operate it, modify it etc.
These investigations will result in a business view on simulation modeling. This should
be very much related to the business processes that are in existence in the organization.
Thus a conclusion and recommendation will be made on the place of the model in the
organization and a conclusion or recommendation will be made on the processes in which
the model fits in the organization. Finally there will be a part that deals with how the
desired state (organization and process) may be reached.
The answer for these two objectives (technical and organizational) will be given at two
levels. The first level will be at company-level. Here recommendations and conclusions
will be drawn for the actual study at GE Plastics. The second level is a more generic one:
the recommendations made for the specific case will be adjusted to a more generic form
that they may apply on any company in a similar situation.
The topics on which research has been performed in this study are colored blue in the
following graph:

15

are there genaral prerequisites and guidelines for


implementation of avialability simulation
modeling?

Implement availability
simulation?

organizationally
feasible?

technically
feasible?

model type?

industry
benchmark

data?

previous
projects?

software?

organizational
characteristics

business
processes

build model

evaluate
oppertunities

validate: OK?

OK

NO

No implementation

NO

OK

organizationally
feasible!

technically
feasible!
Implement
Availability
Simulation

General prerequisites for implementation

Fig. 1.5: Research objectives (orange), GEP objectives (green) and research in the study
(blue).

16

1.6

Architecture of the report

The next chapter (chapter 2) will deal with the question what reliability engineering
actually is and what the situation is in industry with regard to reliability engineering and
simulation modeling.
Chapter 3 will provide some more insight in the various aspects of modeling, what does it
add, which types re known and which types will be used in this project?
Chapter 4 will describe the development GE Plastics has already been thru with regard to
availability simulation modeling, by evaluating the previous pilot projects. From there the
objectives are set for the model that will be developed in this study. Some software
packages are evaluated and one is chosen for this project.
Chapter 5 will describe the model built in this study. All aspects of modeling are
described: selection of what exactly needs to be modeled, data retrieval and analysis,
underlying mathematics and statistics, how the model is actually built and eventually the
validation of the model output.
Chapter 6 will describe the organization of GE Plastics and the main business processes
within the organization. Special attention will be paid to the way simulation modeling can
be integrated in these organizational characteristics and typical business processes.
In the last chapter the conclusions and recommendations will be given of how to
implement availability simulation modeling. These will be divided in some that are
specific for the situation in GE Plastics and some that are applicable on any company
within the chemical industry.

17

What is Reliability Engineering?

This chapter is included to give more insight in the theoretical background of availability
and reliability simulation modeling and other related topics.

2.1

Availability, Reliability and Maintainability

When a project is done concerning reliability this usually has the objective to improve or
stabilize the output of the plant involved. Actually one should speak of improvement of
availability in this case.
Availability is commonly used as the term for the combination of reliability and
maintainability.
Definitions of availability and reliability have been given in chapter 1.3. Maintainability
is a measure for the reparability of a piece of equipment. When it has failed, what will be
the expected time to repair the equipment? The reparability of equipment depends on the
function, composition and physical placement of the asset.
The steady state or long term availability is the easiest to obtain and can be obtained by
using the MTBF as a measure for reliability and the MTTR as a measure far
maintainability. The often used expression for mean availability is then obtained (Rotab
Khan, 1995):

A() =

MTBF
MTBF + MTTR

From this formula it can be concluded that increase of the reliability (MTBF) will
increase the availability since the influence of maintainability will decrease. The larger
time between failures will be in comparison with the repair time, the more the availability
will approach one. This would mean the plant would always be available. This is of
course desirable but unfortunately non realistic where it relates to cost. The challenge is
to assign the maintenance resources in the plant in such a way that reliability is
maximized while availability is approaching to one as close as possible against optimized
costs.
Because the whole study is done with the aim of increasing reliability and thus
availability it is interesting to know which factors influence the availability. At a highly
aggregated level there can be three factors identified, they are plant design, operation and
maintenance. Their interaction is shown in the following graph:

18

Fig. 2.1: relations between availability, reliability and maintainability (Lamb, 1995 p.10)

Plant design influences reliability and maintainability (and thus availability) in that way
that it includes redundancy (reliability), workspace (maintainability), equipment type
(both reliability and maintainability), spare parts etc. The operation is also of influence on
the reliability of a plant, if it is always operated at a 100%, the reliability can very well be
less than when it always operates at 80%, if it is assumed that it is continuously operated
with stable throughput. Also maintenance is of influence on the reliability of a plant.
There is a huge difference in a plant where the maintenance strategy decided upon is
operation until breakdown (fully corrective) or there has been a strategy developed to do
preventive maintenance to prevent the (critical) equipment units from breaking down at
any random moment.
In this case plant design is already fixed, this can be used as a tool to improve availability
but is very costly. Operation is also not the way to improve availability, since it is desired
to operate at a high rate from business perspective, and operating conditions are fixed
because of the process specification. Obviously development of a suitable maintenance
strategy can increase availability. Question on which needs to be focused is how to
determine which maintenance strategy achieves most increase of reliability, increase of
maintainability and thus increase of availability. Also small modifications, additions and
changes in the design can be made to increase reliability.
The explanation of these relations does not immediately fully explain why it is favorable
to maximize availability. When there is focus on the maximization of availability this is
often a result of more focus on Life Cycle Costing, or of increased production targets.
When production targets are increased in the operational phase it is obvious that the plant
must be available enough time to meet those targets. When the focus on availability is a
result of Life Cycle Costing in the design phase the study will take all the investments
and returns into account that will be made during the total life of the plant and will try
and maximize income and profit from that point of view. Availability fits into this scope
because a plant will be more profitable when it is available and runs a larger percentage
of its lifetime e.g. Utilization of Total Invested Capital. This can be called the
maximization of life-cycle income. A simple presentation of a plant life cycle cash flow
profile is given in the graph on the next page:

19

Fig. 2.2: Simple presentation of a plants life cycle cash flow profile (Lamb, 1995 p.19)

When the reliability performance is improved the revenue profile in the figure above will
move upward as a result of the greater physical availability of the plant, which means a
greater production time. This is only desirable when the increase of production can be
expected to be sold.
Very important is to find the optimum in allocation of maintenance resources. This means
that the availability will increase as much as possible at as low as possible maintenance
costs. In this graph this would mean that the revenue profile above the x-axis would
increase as well as the operating and maintenance cost beneath the x-axis would decrease.

Income
+

Costs

Availability

Maintainability

+
Maintenance
actions

Reliability

+
Redindant
equipment

Fig. 2.3: causal diagram of relevant quantities in availability optimization

20

Also in the figure above it can be seen that the optimization of availability in order to
increase income is a trade-off between the costs and maintenance actions and redundant
equipment. From this figure can also be seen that the relation between maintenance
actions and reliability is not always the same. If an equipment has warm-up time it will
decrease overall reliability if there is too often maintenance performed (because of
increased failure during warm-up period). This diagram does assume that the increase of
availability will be translated in an increase of production, sales and eventually income.
This is of course only true if there enough demand.
How can reliability, maintainability and thus availability be improved?
When the goal to improve availability is stated, this means that reliability must be
increased (increase of MTBF) and maintainability must be increased (decrease of
MTTR). To know how to achieve this the factors influencing them must be known. One
of the most important factor to increase reliability is the maintenance performed on the
equipment units. Effective maintenance will increase reliability because it will increase
the performance of the piece of equipment. Maintainability will increase with increasing
maintenance actions because the maintenance activities will be divided into smaller
activities, which will take shorter to perform. However, there is an upper limit to these
maintenance activities. Too much will ask to too much downtime of the equipment and
possibly of the system. Also, it is not said that all the maintenance activities are effective
and will actually result in an increase of reliability. Availability modeling can contribute
to find the optimum in the frequency of performing planned maintenance actions.

2.2

Maintenance policies in the chemical industry

When studying a reliability engineering subject this is of course very closely related to
the maintenance policy that is followed in the company. Even stronger, reliability
engineering strategies will only be executed in companies that have a certain maintenance
policy. This section will give a description of the various stages a company can be in with
regard to their maintenance policy and will try to give insight in which phase some
important chemical companies are, to eventually identify the prerequisites to enter the
stage in which reliability and maintenance engineering are in its most mature and evolved
form.

2.2.1 Three stages of maturity


In the development of maintenance policies there can be three stages identified (Smit,
1997). In the following section these three stages will be described, then the companies
interviewed in the benchmark are positioned within these stages.
Maintenance management phase
The first phase the maintenance policy of a company can be in is the maintenance
management phase. In this phase the policy towards maintenance is mostly focused on
breakdown repairs and the efficiency of these activities. There may be some preventive
actions or inspection; however the main focus is on performing maintenance actions as
efficient as possible. There is attention for planning and scheduling of maintenance
activities. The most advanced tool used in this phase will be a maintenance management

21

system that registers all the work done and can be used in the planning and scheduling of
activities.
Maintenance technology phase
In the second phase, the maintenance technology phase, the focus will shift from the
efficiency (how and when do we perform?) to effectiveness (why do we perform certain
tasks?). The consequences of maintenance actions are taken into account in the sense that
maintenance is regarded as a way to keep the equipment operational. This means that
here the availability of equipment will become a point of focus.
In this phase structural improvement programs such as RCM and RBI will be of great
interest because they help determine the optimal way of performing maintenance to
maximize uptime of the separate equipment units. There will be much attention for
production lost due to the unavailability of equipment. This will decide on which
equipment to perform the improvement programs.
Maintenance engineering phase
In the third phase of development in maintenance policies is the maintenance engineering
phase. As the name indicates (engineering) there are more theoretical bases on which the
maintenance policies are funded. As the former phase transition the scope of maintenance
in this phase will be even higher in the next phase. With this transition the scope changes
from the existing equipment units at that moment to the life cycle of the system.
Also, this is the first time that there will be attention for maintainability of the equipment
and the system they are part of. With the extension of the time horizon there will be life
extension plans and programs developed.
Main interest here is the optimization of maintenance policies at trade off with acceptable
costs. Because the scope of time and studied system is more complex in this phase
simulation and modeling tools are essential for and add value to decision making.
The phases described above briefly form the maturity process the maintenance policy of a
company can go through. Main reason a company slowly makes the transition from one
phase to another is the continuous drive for improvement. Only when there is large
improvement at reasonable costs the transition will be made. In the next phase a small list
of well-known chemical companies will be positioned in maintenance maturity model.

2.2.2 Companies evaluated


During this project interviews have been conducted with seven large chemical companies
concerning their approach on reliability engineering. In this chapter these companies are
evaluated on several aspects of their reliability engineering strategy and execution. The
overall score of the company on these aspects will determine the position of the company
in the development phases. There will be a graphical representation of the results at the
end of this chapter.
How to determine the position of a company in the development phases? As stated above
the companies strategy will be evaluated on several aspects. First these aspects are listed
and explained, and then the actual evaluation will be made. This evaluation has been
reviewed by the interviewed companies.

22

The aspects used here have been derived from the characteristic of each development
phase, the efficiency, effectivity and optimization. The aspects are organizational and
operational result6s from these characteristics. Aspects chosen here to position a
company in the development phases are: organization of reliability engineering, structural
effectiveness improvement programs, life cycle costing evaluations and the digital tools
used for the programs in the various phases.
From the previous explanation of the different phases it may be obvious that the more of
these aspects the company has realized and embedded in the daily work processes, the
more mature the company is in its maintenance strategy development.
Since all the companies evaluated have already passed the phase of maintenance
management they all have a CMMS implemented to keep track of all maintenance
actions performed in the plant. This is not used to distinguish the differences between the
companies and thus the use of a CMMS is not evaluated in the following table. The
companies will only be evaluated on the tools that facilitate the policies and support the
decisions in the maintenance technology and the maintenance engineering phase.
In general it can be stated that most chemical companies have just made the transition
from the maintenance management phase to the maintenance technology phase. There are
a lot of companies focusing on the implementation of improvement processes such as
RCM and RBI. In several companies there have been some pilot projects with reliability
simulation; however, often it is found that much improvement can still be made on the
low hanging fruit (easily identified improvement opportunities without simulation).
However, there are big differences in the way companies have organized their vision and
strategy on reliability engineering. Some have different policies even within one site
(every plant its own) where others have the same policy implemented globally. The
companies that have their strategy organized globally are in a further stage than the ones
that do not.

23

Organization
RE
Global
plant
strategy
1
2

+
+

+
+

GEP
6
7

+
+

+
+

Structural programs Tools


Simulation
RCM/RBI LCC/
Pilots
Structural
etc.
LC
mgt.
+
+
Titan
+
+
+
+
Own/RCM
+
turbo
+
RCM
Turbo
+
RAT
++
+
+

BlockSim
Consultant
SPARC
RRM

+
+
+

Table 2.1: Evaluation of the maintenance and reliability strategies of several chemical companies

It should be noted that the tools can be subdivided in the categories simulation tools and
tools that facilitate the structural improvement processes. The tools belonging to the first
category (simulation) are: Titan, BlockSim, SPARC etc. The rest belongs to the second
category (facilitating for improvement process).
Two conclusions can be drawn from this evaluation:
The first is the difference in implementation of the structural improvement processes.
This can be done with some generic studies that can be applied on a lot of equipment or
thorough studies but at low speed (both receive a +) as opposed to a high speed
implementation of thorough and equipment specific studies (which receive ++).
Secondly, the statement made above that most companies are in the second, maintenance
technology phase, can be proven right from this table. It can also be concluded that they
are aware of the methods and technologies in the next phase and will now have to
develop a strategy to make this step. For the case of GE Plastics a recommendation will
be made for the strategy and execution plan to make this transition in a further stage of
this report.
When the companies are plotted in a graph that shows the maturity of the maintenance
function in their organization the graph in figure 2.4 is obtained. Since no measurement
exists for maturity the y-ax does not have a scale. The length of a maturity phase is
estimated to be between 1-2 decades industry wide. If a single company assigns a large
amount of resources to improvement of maintenance management a stage transition is
estimated to be possible within a couple of years.

24

Maturity of maintenance function

Maintenancee
1
engineering

2
3

GEP

Maintenance
technology

Maintenance
management

Time
Fig. 2.4: Evaluated companies within the trend of maturity of maintenance functions

From this investigation it can be concluded that all evaluated companies in the chemical
industry are at almost the same point in their development. There are hardly companies
that can be taken as an example of best practice on how to manage a higher stage of
reliability and maintenance engineering. Thus, it can be recommended to search for other
industries (air force, navy, energy plants, etc.) and learn from the experiences they made
during the implementation of more advanced tools that brought the maintenance and
reliability function to a higher level.

25

Modeling

In general modeling is used in situations that are too complex or costly to give any other
judgment that can be validated by reality. Mostly, in these situations there is only a
personal, emotional judgment based on experience
In the following chapter first some general aspects of modeling will be explained, and
secondly some specific aspects of availability modeling.

3.1 General Modeling


In every day life there are a lot of ways in which modeling is used. The most easy way in
which modeling is used is the gut feeling that people have about the way aspects of the
world around them work. Many people will not regard this modeling, however it is an
interpretation of the way systems in the surrounding world really work. This may request
a definition of a system first. In modeling practices a system can be viewed at that part of
reality that is regarded relevant considering the given problem, that can be seen as a
whole and consists of connected parts and that shows its behavior as a whole.
Modeling is basically a systematical representation of these systems in the real world. Of
course there are several ways in which the real world can be formalized. Depending on
the aspects that are considered relevant and more or less important the model will have
certain properties.
The major benefit of using formalized models is that the behavior of certain systems can
be studied and their behavior can be described in detail without actually impacting the
real situation.
In this project a simulation model is developed. What exactly is meant by the word
simulation? According to Shannon (1975) simulation is:
The process of designing a model of a real system and conducting
experiments with this model for the purpose of either understanding the
behavior of the system or evaluating various strategies (within the limits
imposed by a criterion or set of criteria) for the operation of the system
Thus the difference between modeling and simulating is that a model describes reality
and simulation also involves the active experimenting, interpretation or evaluation steps.
How these general ideas on modeling and simulation apply on this particular case of
availability modeling is described in the following section.

26

3.2 Availability Modeling


3.2.1 What can availability simulation modeling add?
When improvement of the reliability and availability of a plant is pursued, insight is
needed in which equipment units have great influence on reliability and availability and
preferably also in how certain maintenance strategies or tactics can influence this. The
influence of certain pieces of equipment on reliability can be two fold. Either the
equipment fails so often that there is loss of output due to often downtime, or the
equipment fails only on exception but is of such complexity or criticality that it takes
very long to replace or repair. The equipment that fails very frequently can be said to
have very low reliability and the equipment that takes very long time to be replaced or
repaired can be said to have very low maintainability. Both result in low availability and
are thus considered critical in this project. Availability simulation modeling can provide
insight in these cases by identifying exactly those pieces of equipment that are critical for
availability. With a simulation model experiments can be made with different
maintenance strategies and their influence on reliability and maintainability. Also the
influence of various adjustments in process layout (e.g. extra redundant equipment) can
be studied. Major deliverable of availability simulation modeling is that it quantifies the
results of analysis and the analyses are based on real data, derived from the plant.
Summarizing, the benefit of availability simulation modeling will be:
Availability simulation modeling will use real plant data to predict the availability of the
plant and its assets. Other factors (maintenance strategy, plant layout etc.) may be
changed; availability simulation will show the influence of changing these factors on
availability, reliability, resources and costs.
In reliability modeling two types of models are commonly known. One is based on the
Markov theory that analyses the transition of possible states of the studied system. The
second is based on the Reliability Block Diagram Analysis which is based on the system
logic and the reliability of the models incorporated in the system.

3.2.2 Markov modeling


Markov modeling is based on the assumption that a system and its components can be in
different states. A component, at lowest level, can be either up or down, while a system
can be in any possible state identified depending on the components of which it is made
up and the state they are in. A Markov model is a so-called state-space model and
describes the transitions of one state to another. The transition probabilities only depend
on the present state of the system (Wolstenholme, 1999). The model should include all
components, the states they can be in and the frequency at which they change state. A
component has a failure frequency with which it changes from its up state to its down
state and a repair frequency with which it changes from its down state to its upstate
(OConner, 1995). For a system with a great amount of components this can easily
become an extensive calculation. In the following graph a representation is given of a
Markov model for a fully repairable item.
27

FAIL
2

OK
1

Fig. 3.1 Markov model for a single fully repairable item

For a complex system the calculation can easily become very extensive, since all the
assets in the system can be either functioning or not, and the system has N assets the
different states the system can be in will be 2N with 300 elements this will be
approximately 2 * 1090! Also the graphical representation is somewhat difficult to
understand if one is not completely familiar with the system and the possible states it can
be in. All states have a certain block shape and the possible transitions from one stage to
another are represented by the lines in the diagram. This results in a diagram that shows
all possible failure states and derivates identified.

3.2.3 Reliability Block Diagram


A Reliability or System Block Diagram (RBD) shows all equipment involved in the
process and how they are interrelated. In this case interrelation only applies on the
influence the equipment will have on the final output of the system when it fails. So the
aim of making an RBD is to achieve insight in the physical assets involved in the
performance and the way they are dependent and interrelated. In other words, an RBD is
a schematic representation of a system in which the connections symbolize the
interdependency and functioning of this system (Cochran, 2001).
An RBD only provides insight in the consequences for the system functioning when an
asset fails, but does not reveal fault characterization information.
There are two ways in which these pieces of equipment can be related. They are either set
parallel or they are set in series. If equipment is set parallel they can never stop the full
chain that follows in the process, only decrease capacity of the following chain. This is
because the node of the parallel equipment with the failure then becomes the bottleneck
that determines capacity. As can be seen in the following picture:

28

Fig. 3.2: RBD of a 4-component system, with one redundant item.

If B and C are equally important for the ongoing of the process and perform the same
function, the process will fail when B and C both fail, the process will run at half capacity
when either B or C fails and the process will run at full capacity if neither B nor C fails.
When equipment is set in series failure of any node will cause failure of the full chain,
which can be easily seen in the next picture:

Fig. 3.3: RBD of a 4-component series system

If either one of the blocks in the chain fails, there will be no throughput to the rest of the
chain.
If a reader of an RBD relates this diagram too much to the process flow it can cause some
confusion. An RBD actually does not take the real function of a piece of equipment into
account, only what the influence is of malfunction of a piece of equipment for the rest of
the process. It deals with the interrelation of the pieces of equipment and not the actual
real function in the process. In this way a heat-exchanger that is very critical for the
process in the reactor, but which is physically placed on top of a reactor aside of the main
process flow stream, be in series in the RBD because its failure can cause a total
production stop.
In this way an RBD provides a thorough insight in the interdependence of the various
pieces of equipment. However, like almost all models, an RBD can be made at several
levels. The process can be divided into various sections, and these can be subdivided into
equipment or asset level. Depending on the deliverable that is aimed at, a choice should
be made which level of detail the RBD must have.
Advantage of making an RBD before starting to build a simulation model is that the
architecture of the RBD can directly be copied into the simulation model. The simulation
model also has several layers and divisions between sections etc. with which also
experiments can be done at a higher level or with parts of the plant. The only difference
29

between an RBD and the simulation model is that an RBD is just architecture and that the
simulation model also has content: the blocks can either function or fail.

3.2.4 Simulation engines


The way in which these Markov and Reliability Block Diagrams are analyzed can be of
two kinds.
The first is an analytical way, in which the engine analyzes the logical outcome by
analyzing the trends that the separate parts should show in their behavior. Therefore the
system should be fully represented by analytical functions, which can become very
complex in large systems. However, analytical simulation is very fast compared to Monte
Carlo simulation since it only analyses one simulation. Analytical modeling is typically
suited for calculation of steady-state availability.
The second is a Monte Carlo simulation engine that works in a probabilistic way. It runs
various simulations in one run and subjects the system to typical events when simulating.
It is an event driven engine that generates random numbers that correlate with a certain
state of the system. E.g. when a piece of equipment has a reliability of 0.8 then all the
numbers generated between 0 and 0.79 will make the item work in the model, while
numbers from 0.8 on will state the item failed in the model (Armstadter, 1971 p.176).
Monte Carlo Simulation can calculate time dependent availability.

3.2.5 Which model and simulation engine to use?


When studying a system with a lot of components that can all have various states, it will
be obvious that the use of an analytical Markov model will require extensive labor and
provide little insight for people not fully familiar with the model and studied system. A
Reliability Block Diagram however, provides huge insights also for people not
acquainted with the model, because it does not show the difficult system states but the
physical assets in the studied system and only needs easy understandable input to produce
its easy understandable output.
With regard to the simulation engines the main benefit from Monte Carlo simulation is
that it models the studied system in a probabilistic way by randomly drawing times to
failure and repair from the stochastic distributions. Also system complexity is irrelevant
to the method (Korver, 1994). Analytical modeling, although it is faster, does not take the
probabilistic character of reliability and availability into account. Advantage though is
that what if analyses are easy to make fast when performed by computer.
In a system calculation each component represents a dimension and thus a system with k
components requires a k-dimensional calculation. This effect limits the ability of
analytical methods to be a solution for advanced calculations of modern complex system
predictions (Dubi).
In this project an RBD physical setup of the simulation model and a Monte Carlo
simulation engine are chosen.

30

Availability Simulation Modeling in GE Plastics

4.1

Evaluation previous pilot projects

There have already been previous pilot projects on availability simulation modeling
within the Lexan business of GE Plastics. Most of them have been performed in order
to solve a specific question on reliability, availability and maintenance management.
There has actually never been an overall review of these studies and their successes.
Hereto the following section is included in this report. It should give insight in the
process GE Plastics has already been through with regard to simulation and thus say
something about the opportunity of implementing of availability simulation modeling in a
more structural way.

4.1.1 BPA1, Bergen op Zoom


The study in the BPA1 plant has been performed by Delta Consult in 1997 using the
product SPAR. This study has dealt with simulation of the (pre)reaction section of the
BPA1 plant as well as the demonstration of the ability of modeling to show the effect of
certain chosen maintenance strategies on the plant.
Also the time between shutdowns has been of interest in the simulation study. However,
main conclusion that could be made is that the output of a model is only just as good as
the input data. The possible improvement in this project had approximately the same
percentage as the uncertainty of the complete model, which made it hard to use. At this
time GE Plastics was still working with another CMMS that was less accurate than
EMPAC and a lot of failure rate data has been estimated in this project.
The conclusion on the content was that most of the preventive maintenance was
completely out of place and not satisfying the goal it was meant for. However, major
improvement has been made on this topic since then with the implementation and
execution of RCM2 in the total BPA1 plant. This has been a major contribution of the
modeling project and has been positively evaluated.

4.1.2 Chlorine plant, Burkville Alabama USA


The reliability simulation study in Burkville in the beginning of 2000 was performed by
the Clockwork Group in combination with Jacobs engineering and concerned the effect
of the extension between rebuild/overhaul of three chlorine compressors from one to
three years. No economic evaluation has been made.
Most of the data used in the SPAR simulation model has been based heavily on
interviews with plant maintenance experts. Industrial databases have been used as a
benchmark.
Conclusions from this study have been that the extension of time between overhaul
would be very profitable, but that it could unmask hidden failure modes that have not
been predicted by the model. In this case they should be included in the model. Another
recommendation that has been made is to use the model as a continuing tool in the
management of plant reliability, maintenance and operation. For this purpose the model

31

should be updated with new data as there is more learned about the system. However, no
action has been taken on this recommendation. The advise to extend the time between
rebuild/overhaul has been regarded very valuable.

4.1.3 Lexan Plant, Cartagena Spain


The Lexan production plant in Cartagena, Spain is the youngest of all and had its startup only four years ago. When starting up the factory there have been a lot of problems
concerning equipment reliability. Equipment failed very often. This led to the decision to
make a simulation model of the complete train.
In Cartagena the model has also been built with the help of outside experts, in this case
Delta Consult and students of the University of Cartagena. The software use for this
model is BlockSim. This is the only site in which the model is still actively in use.
Initially the model has been filled with industry and plant data. As time goes by the
model is more and more filled with plant specific data.
Because the model is developed from a maintenance point of view it is still stored within
the maintenance department. Within this department there are two main uses of the
model. The first is the comparison of the model output, which forecasts downtime due to
mechanical failures, with the actual downtime and failures. It is assumed that the
difference is caused by other sources such as human failure and process upsets, and these
causes can be regarded as opportunities for improvement. The second main use of the
model is in plant & equipment investment discussions.
New development is that the model will be linked to the data resource systems, EMPAC
and the dent administration, which makes it a lot easier to update the model.
Although the model is still in use in Cartagena it very much depends on the person who is
local maintenance leader to which extent it is used because it is not yet an
institutionalized method in the normal business processes and decision making.

4.1.4 General conclusion pilot projects


From the evaluation of three previous pilot projects it can be concluded that the outcome
of reliability modeling has been received with great interest. The same recommendation
has been made over and over concerning data acquisition: the data available is sufficient
to make an initial model, but it is preferable to monitor the desired data with more care
and update the model over time. The potential improvements have still not been made up
to the desired stage, as can be concluded from the model built for this study.
Another observation that has been similar in most of the cases is that the model has been
built by outside experts and it has been used for a single improvement project and not for
continuous improvement. The model built in this study has proven that outside experts
are not necessary for building a useful model.
Cartagena might be regarded as the exception in the ongoing use of the model. However,
if the function of maintenance leader would be fulfilled by someone who would not
appreciate simulation modeling as a valuable tool, the model would probably disappear.
For instance no GE Plastics people have been professionally trained in using the software
and building models. In this sense Cartagena looks better continuity wise, but the
situation is still not at all one of a fixed, continuous and widespread use of the model.

32

In general the conclusion is that there have been enough pilot projects in various sites to
bring availability modeling to the attention of upper management and make a company
decision when and where to use availability modeling in existing business processes.
Now it is time to proceed to the next step that may consist of a more general
implementation of the model and parallel with this an adjustment of the organizational
structure that will make the use of the simulation model and the resulting improvement
last.

4.2

Objectives of the model

4.2.1 Functional Objectives


To be able to decide upon the most appropriate maintenance strategy there are several
sub-questions that the model can also answer in order to increase insight in the interaction
between maintenance tactics and reliability and maintainability in the plant.
The functional objectives can be subdivided into initial functional objectives that the
model really should be capable of answering when it is first built. These are the so-called
need-to-haves. Then there is a section of secondary functional objectives that are
desirable that the model will eventual answer, but are somewhat more sophisticated or
need a more extensive model than the first draft built. These can be regarded the so-called
nice-to-haves.
Initial functional objectives
These are the objectives that are related to the very basics of reliability engineering,
which is why they have been selected here as initial objectives. First a brief overview is
given of these objectives, and then they will be briefly further explained.
Initial functional objectives for the availability simulation model within GE Plastics are:
Identification of equipment logic, in which way are the various pieces of
equipment interrelated and dependent of each other?
Identification of critical equipment, either due to frequent failure or due to high
maintainability.
Influence of redundant equipment and integrated buffers on (system) reliability.
Identification of knowledge and data available for reliability engineering issues.
These objectives have been set by combination of the desires of GE Plastics and the
characteristics of simulation modeling. What objectives area good starting point to get
acquainted with simulation modeling and achieve enough insight to state whether
availability simulation modeling is technically feasible.
To be able to actually make a model and to understand how equipment can influence
reliability and how they are interrelated the functional dependency needs to be identified.
This is very useful to increase insight in the functional dependence of the various pieces
of equipment.

33

The model should also be able to state the influence of redundancy of all the various
pieces of equipment units in a change of reliability and availability of the plant. This is
actually already a second step in simulation modeling: experimenting (as opposed to the
first step building).
By building this model the exact identification should be made of the accuracy of data
needed for reliability engineering in the plant.
Secondary functional objectives
These are the objectives that the model should meet when it is developed somewhat
further than the initial model. These will include allocation of resources, spares, costs etc.
These are parts of the model and reliability engineering that are of interest, but when
starting on a new initiative as modeling, not immediately pursued. This would make the
project extensive and complicated. This is a dimension of the model that can be added in
a later stage. Moreover, this also requires more data acquisition, which is a difficult part
of modeling.
However, when these aspects are added to the model it should be possible to make the
following analyses:
Spare and warehouse optimization
Integration and optimization of resources allocation
Cost optimization.
Because these analyses are a level beyond the model as is developed for this project they
are not included further in this report. It should be noticed however that the model needs
a lot more input data for these analyses and that they have to be available and analyzed
first.

4.2.2 Business objectives of the model


When the objective is stated to build an availability simulation model of a plant there
must be business reasons on which this objective is based. Eventually every business
objective can be reduced to one final goal and that is always an economical one. How
does the model developed in this project fit into this overall business objective?
To answer this question the role of the model needs to be traced back to where it fits in
the high-level business process. This business process always starts with the development
of a certain vision that can be translated to a company mission. In this case the mission is
an economical one: maximize profits. This can be achieved by using risk-based decision
making to optimize the balance between asset performance and reliability on the one
hand and on the other hand the costs that are needed to achieve this performance and
reliability.
A certain amount of knowledge is needed to change this mindset into a tactic can achieve
profit maximization. The knowledge is a vital part of the organization, and can be used in
the application of several tools.

34

In the field of reliability engineering these tools are amongst others, availability
modeling, RCM (reliability centered maintenance), RBI (risk based inspection), etc.
The final function of the model is to find the optimal allocation of recourses for the plant.
The model can therefore be viewed as a tool that enables the transition of the business
process to the execution tactics.
When the outcome of the model (the functional objectives) is used in the correct way (the
outcome is interpreted correctly and the correct decision is made based upon the data), it
can reduce maintenance costs and improve uptime of the plant, which may be equivalent
to higher production.

4.3

Investigation available software

When starting a reliability-modeling project one of the first decisions to make is the
software in which the model will be built. This decision should be based on the matching
of the objectives stated for the model (which output should it be able to generate?) and
the capability of the software to provide this output. In general, it is best to choose a
program that exactly meets the requirements. One should choose a program as simple as
possible that makes the model easy to build and simple to use.
In the next section the results of a brief research are given. It can be concluded that there
are a lot of possibilities, certainly there are even more than given here. However, this
section is included to give an impression of the different types of available software and
of which factors influence the decision of which software package to use.
There are several companies that supply a variety of software on reliability,
maintainability and availability subjects. Below, the largest suppliers are listed with their
main packages and the possibilities of these packages. Very often packages are
complementary and easy to run together. Most of the suppliers provide more packages
than listed below. However, a selection has been made for the packages that would most
probably fit in a project similar to the one described in this report.

35

Table with overview of characteristics of various types of software

BlockSim
SPAR
Titan
MAROS
TARO
Toolkit
AvSim+
SPARC
Optagon

Model Sim.
type engine
RBD
MC/
AN
RBD
MC
RBD
MC
RBD
MC
RBD
MC
MKV
MC
RBD
FT
RBD
MC
FT
RBD
AN
RBD
MC

Flow/
storage?
+

Spares

Costs

Resources

Programming?

+
++
+
++
-

+
+
+
+
+

+
+
+
+
+

+
+
+
+
-

Y
Y
N
N
N

++

+
+

N
N

Table 4.1: software packages and their characteristics

From this chapter it can be concluded that there are a lot of different possible software
packages to use when performing a reliability evaluation and optimization project by
means of simulation. Interesting is how to determine which package will serve the project
best. Naturally this will be that package whose output will be most similar to the goals
stated for the modeling project, and capability of the package for the representation of the
modeled system. However, as there is a wide variety of possible packages, there will be
several packages suitable for the project and the decision will be made on how easy it is
to use the package, the price of the package.
Also in this project there has been a decision made to use a certain package for the
simulation project, this decision is described in the following chapter.

4.3.1 Factors determining the choice of software


When determining which software to use it is extremely important to have clearly stated
which modeling questions should be answered by use of the model. In this project, the
development of a reliability model for the Resin plant, the goals can be divided into
practical modeling questions and higher, more abstract business and organizational goals.
In the choice of which software to use the practical modeling questions will be most
important. Only when the model answers the right questions it will be able to provide the
output with which the business and organizational goals can be met. These modeling
questions, or the functional objectives for the model have been described in a previous
chapter. Here a brief recapitulation is given.

36

The main modeling questions that should be answered by the reliability model of the
resin plant are:
Determination of critical equipment for reliability.
Determination of system reliability
Influence of redundant equipment on reliability
Influence of time and preventive maintenance on part and system reliability, and
start on optimization of preventive maintenance
Because the main modeling questions are known, the actual output that that model is
supposed to generate is known, and a software package that meets these requirements can
be chosen.
However, if it can be assumed that the model will be used for more objectives in the
future and it will be expanded, this should also be taken into account. Use of the model
should not be limited by the capabilities of the software used. Also future utilization
should also be taken into account when choosing the software package.
When choosing the software type the following aspects need to be taken into account:
1. Meets primary needs
2. User friendly
3. Fits company IT guidelines
4. Meets future needs
5. Nice to have features

4.3.2 Choice of software package


As is clear from the goals stated above this project does not ask for a highly sophisticated
software package and all studies packages would be possible to use. Main goals can be
easily modeled with hardly any time delays, integrated spares, cost analysis or any other
advanced, integrated system. There are only questions related to practical system
properties. Thus a software system does not need to have excessive possibilities to model
more features. The package only needs to apply on the physical assets in the studied
system, their properties and their interrelation.
However, this can still be achieved with several software packages. For practical reasons
the Reliasoft package of Weibull++ and BlockSim is chosen, since GE has a worldwide
license for these packages.
If this modeling technique will be more and more integrated in a company it will
probably gain in subjects on which it will be applied. It is likely that BlockSim cannot
provide the desired output due to its restrictions. It will be very well possible that the now
chosen package will someday be replaced by another package that is more sophisticated
and will have more modeling possibilities.
There will be more attention for which package this might be at the end of the modeling
section, when the experience is gained which features that are essential in modeling these
types of processes.

37

Modeling example case

5.1

Description Lexan production train and Resin plant

Before the model is built and the desired experiments can be done and evaluated there
needs to be a thorough understanding of the processes performed in the simulated system.
A simulation model can only succeed and provide the desired output if the system with
all its complexities, interrelations and dependencies is represented well in the physical
setup. Therefore, investigating the system is a very important step in the modeling cycle.
The processes that are performed within the Lexan train can be described from several
points of view. In the scope of this project some are more appropriate than others. There
are two that will be described in this chapter and a third one will be integrated in the
actual modeling.
The first description will be the chemical one: which chemical processes are performed in
which plants? In the following section the main chemical reactions will be described.
The second approach that is used to describe what happens in the plants is the process
approach. All chemical reactions are performed in several process steps. These steps are
the main building blocks of a so-called idef0 or SADT (Structured Analysis Design
Technique) diagram that can be found in appendix B. Reason to use this technique to
describe the process is that it also includes the equipment units that perform the
individual process steps. This will increase the insight in the equipment involved in the
process and their criticality. Moreover it also provides a link between the commonly used
process oriented descriptions and the reliability model that will be developed which is
more equipment (object) oriented.

5.1.1 General introduction Lexan production train


The Lexan production train consists of several plants that convert several raw materials
into polycarbonate. Lexan is a GE Plastics brand name for polycarbonate that is used in
the construction and automotive industry.
The borders of the studied system (Lexan production chain) are quite obvious: there is a
physical set of plants that contribute to the eventual production of Lexan. The only
plant that could cause discussion is the utilities plant. This plant does not provide real
feedstock for the following plants in the Lexan production train, but without its output
the Lexan production train is not able to run. In this study it has not been considered a
part of the Lexan production train because it does not provide an input material that will
be processed into polycarbonate.
Two plants are operated to produce raw materials for the actual production process, two
plants perform the actual transitions of raw materials into the polycarbonate and finally
there is one compounding plant that transforms the produced polycarbonate powder into
the form in which it can be sold, pellits.

38

Thus, there are five plants that can be connected almost in single series. They are:
Chlorine, BPAI&II, Phosgene, Resin and Lexan Finishing. The logical setup of these
plants can be seen in the following figure:

Phosgene

Chlorine

Lexan finishing

Resin

BPA II

BPA I

Fig. 5.1: Logical setup of the plants in the Lexan production train.

The two plants that produce the raw materials for polycarbonate are the BPA and the
Phosgene plant. The output of the Chlorine plant is used in the Phosgene plant, Phosgene
and BPA are transformed into polycarbonate in the Resin plant. This will be further
explained when the chemical reactions in the plants are explained.
In the last plant, which is a compounding plant no chemical reaction takes place. Only
pigments and additives are introduced while melting the powder. This will determine the
color and grade (the final type) of plastic that is produced. The grade of the plastic
determines its final strength, flame retardence, UV resistance etc. This plant will finally
deliver the Lexan pellits, which are used by the customers of GE Plastics to make final
physical products.
Chemistry
As stated previously there are only two plants in which the process contributes to the
chemical composition of the final product.
To provide some insight in these plants the chemical process performed within these
plants is described in this section. The BPA made in the Lexan train in Bergen op
Zoom is actually produced in two different plants. They have some difference in the way
they operate; however, the final product is the same.
As the name (Bis Phenol Acetone) suggests, BPA is produced from two parts phenol and
one part acetone. The reaction looks like this:
CH2

CH3
C

O + 2

OH

HO

CH3

C
CH2

39

OH + H2O

In the following plant, the Resin plant, the BPA is combined with Phosgene in order to
produce the desired polycarbonate, which looks like this:

CH2
HO

C
CH2

O
OH +
Cl

CH2
O

C
Cl

O
O

+ 2n HCl

CH2
n

The other plants part of the Lexan production train are to support these two plants, and
to provide them with the necessary feedstock and utilities.
From the discription of the chemical reactions in the Lexan production train it can be
concluded that the actual production of polycarbonate takes place in only one plant in
the train, the Resin plant.
Also, from the logic of the interrelation of the plants (figure 6.1) it can be concluded that
the Resin plant is a bottleneck in the production train. Chlorine and Phosgene are also on
the critical path but are more stable plants due to the fact that they run gaseous processes
and the resin plant runs a powder process. The BPA plants are parallel and the probability
that they both fail is small. They have little overproduction and there is intermediate
storage. Moreover it can be sold as a product when it is not further processed to
polycarbonate.
This led to the choice develop the availability model for this resin plant. Because an
availability model requires insight in the simulated plant the resin plant is described in
more detail in the next section.

5.1.2 Description of the Resin plant


The Resin plant can be divided into several sections. Distinction can be made between
the sections that perform the main production activities and the auxiliary systems. Main
process performed in the Resin plant is the combination of BPA monomers into a
polymer resin. Type and chain length is determined by addition of certain chain stoppers.
The main sections in the Resin plant that can be identified are: formulation, reaction,
extraction, concentration, precipitation and drying. In this section only a brief description
of what happens in the various steps is given. The more extensive description can be
found in the appendices B and C where the Idef0 diagram and the RBD from the Lexan
street and the Resin plant are given.
Formulation
The BPA is transported from the BPA plant and here it is desolved in water and a solvent.
The solvent is added to desolve the Resin that will be formed in the next reaction step.

40

Reaction
In the reactors phosgene is added that will cause the reaction to take place according to
specifications.
Extraction
After the reaction has been performed the polycarbonate that has been formed during
reaction needs to be extracted from all the other substances present in the mixture. This is
been done by centrifuging in six steps. After this step there will be no more water in the
mixture and a thick polymer solution is transported to the next step.
Concentration
The polymer solution is concentrated by heating it up in heat exchangers. This causes the
solvent to evaporate and the polymer to thicken. The polymer and the solvent vapors are
separated in a cyclone, after which the polymer solution will continue in the process.
Precipitation
In the precipitation section the solvent that is left in the solution is removed, so a wet
powder will be the output of this section. The solvent is removed by first a common heat
exchanger, then the remaining solvent is evaporated by the use of steam jets and led
through two very long pipes that are steam jacketed. Eventually powder and damp are
separated in a cyclone.
Drying
Drying is performed in several drying steps, first is drying in a column, then a spiral dryer
then the damp and polycarbonate powder are separated in a cyclone. Finally post drying
will be performed and then the dry polycarbonate powder will be ready to be transported
to Lexan Finishing.

5.2

Selection of data to determine failure frequencies

5.2.1 Essential Data retrieval


Essential factors in reliability and availability engineering are failure rate data of
equipment. These data should state the frequency of equipment failure and the result of
equipment failure: does the equipment go down and how long does it take to repair? The
availability and accuracy of these data will form the success or failure of the model. Only
reliable and accurate data will give a reliable and accurate representation of reality.

5.2.2 Database selection


In order to determine failure frequencies a lot of data is needed. In the best situation these
data would exactly state when there is work performed on a piece of equipment, in which
cases the asset really fails, and cannot perform its intended function, and how long it
takes in that case to repair it. Unfortunately these data on repair times is not available in
this format.

41

There are two possible data sources for the uptime and downtime of the various pieces of
equipment in the plant. The first is a special database in use in the company to capture the
data on work performed on equipment. There can be several types of computerized
databases in which this data is stored. In literature and industry these databases are called
a computerized maintenance management system (CMMS). This system is developed to
collect all data concerning maintenance and related topics (spares, costs etc.) and to
connect these several parts of the organization. For example, when a piece of equipment
has broken down and needs to be replaced this system also includes the warehouse
inventory, so immediately it will evaluate whether the spare part is in stock or needs to be
ordered. There can be cost and finance systems included in a CMMS, so then
immediately the financial consequences of certain maintenance actions can be determined
by the CMMS. Also there can be a resource allocation system included that for example
assigns available people to open work orders.
However, if a CMMS is implemented in the organization it is not always used at full
extent. The success of the CMMS is a result of the accuracy of the people who are using
it, who are the people who fill the database with information.

Fig.5.2: place of a CMMS within an organization (McKenna, 1997 p.22)

As stated before the most essential data needed to make the model functional is accurate
data on failure frequencies. This data will be collected from this system, which is called
EMPAC (enterprise maintenance planning and control) at GE Plastics.
The data from EMPAC that will be used to calculate the failure frequencies are the work
orders. Work orders are the requests that actually make sure and register that there is
work performed on the equipment. Thus all actions performed on a certain machine are
included in the system, and it should be possible to derive the frequency with which work
is performed on the equipment from this database. It should be noted however that not all
work on an asset causes to actual downtime of that asset. One should be fully aware of
this difference when starting to work with this data. How this problem has been dealt
with in this specific case will be described in a further part of this report

42

The second system is the DAS (Dent Administration System), which registers the lost
output (planned output real output). Every day the lost output is assigned a cause which
is registered. It is very well possible that output loss is caused by other factors such as
human errors or process aspects (deviation of stated operating specifications etc.).
The database that is used in this project is EMPAC due to two reasons. The first is that
EMPAC contains the data at the desired level, which is the level of technical failures, at
which the model is actually built. DAS contains data at a higher level (also human errors
and process dent etc). It is possible to filter the technical dents from the dent
administration and use this to verify the output of the model based on EMPAC data.
The second reason to use EMPAC data is more important. This can be referred to as the
so-called iceberg theory. EMPAC shows all work preformed on the assets and thus also
the near misses that did not cause downtime. This will provide more insight in the
failure rate and mechanism of the asset and the impact it has on total production. An
analysis can be made of which percentage of the failures (work orders) actually causes
real downtime. Also when the maintenance strategy is evaluated, assets that take a lot of
time to maintain (many work orders), but hardly cause downtime can be taken into
account, which cannot be detected when the dent administration system is used.
The DAS will be used to validate the model. The downtime generated by the work orders
in EMPAC should be approximately equal to the dents registered in DAS under
mechanical dents.

5.2.3 Database usage


In EMPAC the work orders can be found by selecting the tag number of the equipment
and then query the work order link in the database. Using the work orders is also a
problem: there will be one for every action performed on the equipment unit. This also
includes inspection and a lot of minor maintenance actions in which the equipment is still
up. Thus, if these are all used in the failure frequency (when the equipment is really
down) calculation, the determined failure frequency will be far too high, and the model
will show far more production loss and decrease of reliability and availability when used
in the model than can be observed in reality. To determine the failure frequencies there
should be a list of all work orders that cause downtime, however, this is not available.
Thus this selection has to be made manually while reviewing the data. Problem here is
that there is no real information on whether a work order really has been made as a result
of equipment failure. The only way to solve this is to select the work orders, with help of
plant people, of which can be assumed that they are a result of equipment failure. So the
model will give the impression that it is very accurate and only based on hard data but
there is still a human factor involved.
The model and determined failure frequencies can be validated by use of the dent
administration. The data in EMPAC and DAS are truly independent and maintained by
different functions in the organization. In this system every time there is less output
produced than planned, this output loss (dent) is assigned to a cause, like equipment
failure. The equipment failure from this administration should be the same as the failure

43

rate calculated from the selection of work orders. This validation assures that a human
error cannot cause a major difference between calculated failure frequencies and failure
frequencies as they are observed in reality and felt in the plant and output.
The way to determine the actual up- and downtime of the equipment in the plant is the
frequency of failure (work orders) plus the use of an appropriate Time To Repair (TTR).
The TTR is the time, which it takes to solve the problem in failed equipment. This TTR
can be a fixed value (MTTR) or a distribution, both determined from the repair times
estimated by plant personnel. Most favorable situation is that which the TTR is a
distribution from which repair times are drawn. In most cases a repair time is drawn that
will cause no output loss, and only some cases generate a TTR large enough to cause
output loss and generate model output that correlates with the dent administration.
Hardly any data is available on the time that a piece of equipment has been down as a
result of a failure and the corrective maintenance time related to that failure. The only
way to deal with this is the use of interviews with plant personnel that can state which
work order resulted in an action that actually took the equipment down and the time this
took to bring back up.
The resulting input functions that have been calculated for the model can be found in
appendix F. It should be noted that these are the distributions that are inserted in the
model after the first brief reality check in which weird distributions, such as normal
distributions as failure rate have been discarded and replaced for more acceptable
functions. When the initial data analysis did not result in an acceptable input distribution,
there has been a distribution copied from the same asset type in another processing line,
or all the data from the same asset types in the different lines were gathered in one set and
then there has been a distribution estimated for all assets. Depending on the variety
between the functions for the different lines a function is copied (little variance) or an
average function is determined (large variety). Reason to choose for this way and not the
use of general data bases is that the same equipment in the other processing lines operate
at the same conditions and are thus expected to reflect the behavior of the asset in
question better than a general distribution.
It can be concluded that modeling suggests the use of data and objective outcome but
there is still reasonable input required of human experience and estimation. This may
result in a variation of the model output compared to real observations. The influence of
estimation will vary with the function and the physical place of the equipment for which
estimation has been made. It is recommended to start collecting failure data (frequency
and impact) more accurately as of now and update the model as an iterative improvement
process.

5.3

Selection of modeled sections

Due to limited time in this project it has been decided not to include a full data analysis
for all assets in the model of the total Resin plant, but to concentrate on a couple of
sections and to try to increase insight and obtain some in-depth recommendations rather
than superficially evaluate the total plant.

44

In order to be able to select the appropriate sections there are two aspects of all assets
evaluated. The first is the availability of work orders for corrective maintenance activities
and the second is positioning of the assets in the section in the Pareto of the output dents
(appendix H). Assets that score high on both aspects can be expected to be interesting
from a reliability point of view since they need frequent maintenance activities and are
responsible for a large part of the output loss. It can be expected that these are the big
fish where a lot of improvement can be made with little analyzing effort and all aspects
of modeling can be evaluated.
In appendix E a table is included in which these aspects are stated for all assets. From
these results four sections are selected to be investigated further in the model. These
sections are: transportation/weighing, precipitation, drying and the air blower system. In
further analysis the air blower system is integrated in the drying section.
Striking in this evaluation is that the main reaction chain is a lot more critical to
reliability that the auxiliary systems. Thus in the selected sections there is only one
auxiliary system. Also the sections where the processed material is powder or thick slurry
(transportation/weighing, precipitation and drying) are more unreliable than others, and
therefore these three sections are included in the selection.

5.4

General mathematics and statistics

The concept of reliability as a probability means that any attempt to quantify it must
involve the use of statistical methods (OConner, 1995). The input of the model will be
failure frequency distributions and time to repair distributions for all the assets studied in
the model. These distributions will be determined by making several statistical
calculations on historical data. In the following section the general statistics that are
applied in this project are explained. This explanation will include the actual application
and the interpretation of the outcome.

5.4.1 Failure frequency functions


From the data retrieved in the CMMS it is tried to determine failure rate distribution
functions that represent reality and are as accurate as possible. First the main
mathematical functions that show general aspects of the occurrence of failures are
described, than the specific failure frequency distributions are described.
The main probability of a failures occurring is described by the following mathematical
functions (Fritz & Bertsche):
probability density function (pdf), which is represented as f(x).
The cumulative density function is given by F(x), with f(t)=dF(t)/dt.
The reliability function R(t)=1-F(t)
The hazard function (failure rate) (t)=f(t)/R(t)
The most commonly used function to describe failure distributions in Reliability
engineering is the Weibull-distribution, either with 2 or 3 parameters. The generic form
of the probability density function is (OConner, 1995):
f

( x )

45

With corresponding reliability function:


R (t) = e

And hazard rate:


1
t

In this function is the shape parameter and determines the actual shape of the function
and is the scale parameter, or characteristic life.
The Weibull-distribution is so popular because one common function can describe all
sorts of failure behavior. By using different values for the shape parameter , the failure
rate can either increase, stay constant or decrease over time. In fact this function
approximates hyper exponential, negative exponential and (log)normal functions for
various values of . Thus, depending on the value of the there will be a time dependant
factor in the function. However, in BlockSim it is very well possible to use other
functions as input for the failure frequency. The selection can be between Exponential,
Normal, Lognormal or Weibull (2 or 3) distributions. These distributions are determined
by data analysis in the software Weibull++.

5.4.2 Repair time functions


Assets can fail in several ways. This can be stated as that an asset has various failure
modes. For every failure mode there is a typical repair time. For example, for a pump the
replacement of a bearing has a typical time and the greasing of a bearing as another
typical time. Mostly an asset has a dominant failure mode, meaning that there are only
few failure modes determining most of the failures. This also means that there are only
several repair times that occur often. As a result typically repair time functions are
normal or lognormal. This means that when the Normal distribution function is used in a
simulation model the repair time will stay within a certain boundary. When a Lognormal
function is used it also sometimes has a high outlier. This is also the difficulty of using
lognormal functions in simulation modeling. It occasionally generates a very high (close
to infinity) number. However, this problem is solved when a lot of simulations are run.
In its general form the p.d.f. of the Normal distribution function is given by:
f ( x ) =

(2 )

1 / 2

In which is the location parameter, equal to the mean and is the scale parameter equal
to the standard deviation.
In its general form the p.d.f. of the Lognormal distribution function is given by:
f ( x ) =

( 2

1 / 2

ln

46

The mean and the standard deviation are given by:


= e

[e (

2 + 2

) e (2

1/ 2

In which and are the mean and standard deviation of the log lifetime distribution.

5.5

Building the model

Modeling begins with the identification of the physical system that will be modeled. The
equipment that forms the system together and the interdependence between this
equipment needs to be identified. Initially the modeler can build the RBD from PFDs
and EFDs and other detailed descriptions of the process and the plant. However,
interdependence of equipment cannot always be easily identified form hard data. Thus,
when the initial RDB is built an interactive session with operators, process engineers and
technicians must be the final check whether the RBD is built according to reality. The
RBD built for the simulation model of the resin plant is attached in appendix C.
Next step in the modeling process will be the data analysis of failure frequency and repair
times in order to be able to evaluate the system on these characteristics in a later stage of
the process. The first step will be data retrieval from the available data sources (EMPAC
and possible industrial databases), then interactive discussion with professionals
(preferably an operator, a mechanical and an electrical engineer).
The distributions for the TTFs can be calculated from the data available in EMPAC.
During the analysis of the work orders in EMPAC in this study, it became obvious that
there were many work orders in the system that should not be taken into account in the
calculation of the distribution for the TTF. These mainly concerned planning activities,
meetings, and repairs on assets other than the studied one. Conclusion to be drawn is that
the work orders cannot be directly copied from the system and analyzed, but need to be
thoroughly screened.
The repair times are hardly filled in EMPAC and are therefore most accurately described
by estimates from professionals in the interactive sessions. For every identified work
order estimation has been made for the downtime of the equipment and these values have
been fitted in a distribution.
The distributions determined and used in the simulation model of the resin plant can be
found in appendix E.
With appropriate preparation the model of a whole plant can be filled quite easily. The
appropriate preparation, scrubbing of data and calculation of failure and repair time
frequencies, is the hard work of modeling. This first model will be a preliminary model.
When the preliminary model is validated with data on output loss there can be a final
interactive session to discuss the differences and odd findings in the model, and how they
may be resolved. It could be concluded that these unexpected values are reality, which
would be a great learning experience for the professionals in the plant.
So, eventually the model is built with a lot of input knowledge from professionals with
extensive experience in the plant. In three rounds of interactive sessions the modeler
47

should be able to acquire all the data needed on plant layout and equipment interrelations,
time to failures and repair times and eventually the validation of the model.

5.6

Validation Simulation Model

Validation of the model is the check whether the outcome of the model is representative
for reality. Hereto two sources are identified in the case of the Resin plant simulation
model. The first is the dent administration. In this case only the dent administration of
2002 has been used. However, the registration in the dent administration is in lost
production tons, whereas the simulation model produces downtime. Thus an extra
calculation needs to be made. The second validation method is the use of experts in the
plant. In an interactive session there can be discussion whether the output provided by the
model is valid to represent reality.

5.6.1 Validation by dent administration


In the following section the outcome of the model is compared with the data derived from
the dent administration for 2002. The downtime provided by the model has to be
translated to lost production tons. In order to do this, the following assumptions have
been made:
The plant produces ideally 650 tons of polycarbonate a day
Line 1 ( A tagnumbers) produces 40% (=260 tons)
Line 2 (B and C tagnumbers) produces 60% (=390 tons)
The model has been run for 1643 days (= 4,5 years), thus the downtime divided
by 4,5 times the production per day gives the lost production a year, and can than
be compared with the dent administration of 2002.
In the following section the comparison is made and where possible there will be a
suggestion/explanation for the difference between these two.
Transportation weighing section
Asset
F-3107
V-352B
V-354
F-306/1
F-306B
F-306 total

CM downtime
(days/4.5 years)
0.28
4.52
1.87
5.35
5.89

Throughput
(tons/day)
650
390
260
260
390

PL Model
(tons/year)
40
392
108
309
510
819

Dent 2002
(tons)
261
461
83

163

Table 5.1: validation equipment transportation weighing section

BPA filters, F-3107:


Almost all lost production in 2002 was made in 2 weeks. This suggests an unexpected
filter contamination, which can be explained by a process upset in the BPA plant. This
should have been classified in the DAS system as process failure. Because this is
exceptional it is very hard to model, and can thus not be easily simulated by the model.

48

Another reason can be the physical setup of the filters. In reality there is one filter
dedicated to each line and a third can be switched between the two lines. However, the
model cannot represent this physical setup and thus there is a standby container in which
two out of three filters need to be available in order to keep the container in function.
This results in possible switching of the (in reality fixed filters) to the other line. 41 tons
per year due to technical failures is regarded reasonable to represent reality.
V-352B:
Explanation: The O2 probes in the bin have caused a lot of troubles throughout the years,
thus they have been replaced in 2002 with a different type of probes that cause a lot less
troubles. Thus in the data analyses all the work orders dealing with the O2 probes have
been discarded, and are thus also not generated by the model.
The outcome of the model can be regarded a good representation of reality.

V-354:
Difference within 25% range.
F-306 total:
This difference can also be explained by the process upset in the BPA plant, which
produced sticky BPA that caused fouling which should have been recorded in DAS as a
process upset.
Precipitation section
Asset
M-316 A
M-316 B
M-316 C
F-323 A
F-323 B
F-323 C
M-321 A
M-321 B
M-321 C
N-336 L1
N-336 L2
N-336 totaal

CM Downtime
(days/4.5 years)
7.34
12.76
12.75
5.05
4.65
5.11
12.29
14.04
7.13
0.55
0.29

Throughput
(tons/day)
260
195
195
260
195
195
260
195
195
390
260

PL Model
(tons/year)
424
553
553
292
202
221
710
608
309
48
17
65

Dent 2002
(tons)
565
332
343
477
36
273
223
190

165

Table 5.2: validation equipment precipitation section

M-316s
The results for the M-316s are in the same order of magnitude as the dents registered in
DAS and are thus acceptable.

49

F-323s
The cyclones show a large difference between the model output and the dent
administration that cannot be explained at this stage, and will require further
investigation.
M-321s
The Fitzmills (M-321s) are predicted to cause major output loss, however, the dent
administration does not show this. Possibly, the dents caused by breakdown of the
Fitzmills are allocated to another asset in the DAS system where the symptoms of this
breakdown are felt. This will be tested in the high level evaluation in the following
section.
N-336
L1: reason of the misfit of the model and the DAS system can be the physical setup of the
pumps. In reality there is one pump dedicated to one line (one for B line and one for C
line) and a third can be switched between lines. However, the model cannot represent this
physical setup. Therefore there is a standby container in which two out of three pumps
need to be available in order to keep the container in function. This results in possible
switching of the (in reality fixed pumps) to the other line.
General: Possibly there are problems switching from the active pump to the standby
pump. In the model this does not have delay time (can be included) and never fails
(probability can also be included).
However, what happens often is that the pumps have lower throughput due to high pump
pressure. This is registered as dent but there is no work order since there is nothing wrong
with the pump itself.

Drying section
Line 1

Asset
M-384 B
M-384 C
M-322 B
M-322 C
K-308

CM Downtime
(days/4.5 years)
26.03
19.93
11.25
4.53
0.11

Throughput
(tons/day)
195
195
195
195
390

Table 5.3: validation equipment drying section, line 1

50

PL Model
(tons/year)
1128
864
488
196
10

Dent 2002
(tons)
404
564
115

Line 2:

Asset
M-384 A
M-322 A
V-342 A
K-313

CM Downtime
(days/4.5 years)
20.16
7.76
2.76
0.01

Throughput
(tons/day)
260
260
260
390

PL Model
(tons/year)
1165
448
159
<1

Dent 2002
(tons)
667
113
681

Table 5.4: validation equipment drying section, line 2

In the drying section all the output from the model for equipment deviates from the dent
administration system, further investigation is necessary.
General drying

Asset
K-311 C
K-311 DE
K-311 AB
K-311
total

CM Downtime
(days/4.5 years)
2.43
0.40
0.25

Throughput
(tons/day)
195
195
260

PL Model
Dent 2002
(tons/year)
(tons)
105
17
14
137
1931

Table 5.5: validation general equipment drying section

K-311s
Wrong data analysis has been made. Often the assumption has been made when
determining the repair time distributions that all spares etc. were available.
Delay time when switching? Possibly when the switch needs to be made this cannot be
made, as opposed to the model where the switch can always be made and no delay time is
included.
In the model there is no quiescent failure distribution included.
Maybe, when these blowers fail this result in the failure of other equipment that takes
longer to repair, and thus more production is lost than these blowers suggest.
It is also possible that the blowers slow down due to high pressure. This is not registered
as downtime or failure, thus there is no work order but there are dents assigned to this
piece of equipment.

General evaluation
In the situation where there is redundant equipment the output generated by the model is
very different than that retrieved in the dent administration. This can have several causes
that are all already described separately above. Summarized they are:
In this first version the model does not take delay times into account when
switching from the active to the standby piece of equipment.
The model does not take into account that the switch from the active to the
standby equipment may fail.
In some cases the model cannot represent the physical setup as it is in reality.

51

The dents in the DAS are allocated to the wrong equipment.


A more basic reason for the difference between the two systems is that the registration
system used for calculation of the dents in the model is different than the system used in
the dent administration system. The model generates downtime on work orders, while the
dent administration system initially allocates dents to the symptoms felt in the production
line. In this way dents can be allocated to different assets than those who actually provide
the root cause for the dent. This problem can be overcome by making a comparison on a
higher level where all the assets that can be involved in the generation of a certain dent
are compared as a system.
How significant the difference actually is can be measured by taking a 25% range and
then evaluate whether the two outcomes are within this range of each other. A range of
25% has been chosen because the quality of data used does not allow very accurate
evaluation. 25% is still a very acceptable range to determine whether the model output is
in the same order of magnitude as the dents.
Within 25% range?
Transport & Weigh
Precipitation

Yes
V-352B
V-354
M-316A

Drying L1

Drying L2

Drying General

No
F-3107s
F-306
M-316 B/C
F-323s
M-321s
N-336
M-384A
M-322A
V-342A
K-313
M-384B/C
M-322B/C
K-308
K-311

Table 5.6: are model results and data from the dent
administration within a 25% range?

Although the output may seem to differ too much this does not necessarily have to be
true. As stated above the allocation of dents may be different in the two systems, thus
first an evaluation and validation with expert knowledge is performed, before making any
further conclusions.

5.6.2 Validation by expert knowledge


Next step in the validation has been an interactive session with some experts that are very
familiar with and knowledgeable about the plant. Main insight gained in this session is
that the dent administration does not always allocate dents to the right equipment,
meaning that often the dent is allocated to pieces of equipment that show the symptoms
of the failure. E.g. blowers which need to slow down due to root causes in other
52

equipment or systems (fouling further down the line). This happens in the allocation of
dents for the system of M-322 (rotary valve), K-311 (air blower) and M-384 (spiral
dryer). Plugging of the M-384 spiral dryer causes most problems. However, first the
pressure will rise in this system and operations will produce as long as possible with
lower throughput due to slowdown of the K-311 blowers. Likewise the dents caused by
the M-322s can be very well assigned to the blowers or the spiral dryers. Thus, these
assets need to be seen as a black box system and the overall dents caused by the complete
system need to be compared. This can be seen in the following table:
model
2002
A
B
C
total
A
B
C
total
M-322 448 487 196 1131 M-322
K-311
15 105
17 137 K-311
772 579 579 1931
M-384 1165 1128 864 3157 M-384 667 404 564 1635
1628 1720 1077 4425
1439 983 1143 3566
Table 5.7: first system level validation spiral dryer system in lost tons per year

From this table it can be concluded that the model actually does generate too much
downtime for the assets in the system, this is especially true for the B line. There can be
two explanations for this outcome. The first is that the model can be actually right and the
B line does cause far more trouble than the other two lines. The second is that the B line
just accidentally had fewer problems in 2002 that resulted in a production dent. However,
the difference is so large, about twice as much, that this requires deeper investigation.
From the failure rate and repair time distributions of the M-322 B it can be concluded that
it will generate more downtime for the same number of failures than the other two M322s. Thus the decision can be made to assign another repair time distribution to the M322B in order to generate a more realistic downtime, and thus production loss for this
asset. In order to get the most realistic distribution all repair times of the tree M-322s are
taken and a distribution is fitted.
The new distribution for the M-322 B repair time will be a normal distribution with mean
0.1966 days and std. 0.0733 days. This change causes the downtime of the M-322B to
reduce from 11.2479 days to 8.7891 days and as a result the production loss predicted by
the model is 381 tons/year instead of 511 tons/year.
Also the behavior of the M-384B is investigated more deeply. When the failure and
repair time rates are compared with the other M-384s (A&C) it can be concluded that
the M-384 B has a very large , which means a fast increase of failures over time. Also
from the results from the simulation run it can be concluded that the M-384 B is predicted
to fail more than the other two M-384s. However, this does not seem to fit the perception
of experts in the plant. This problem is resolved by copying the failure rate distribution of
the M-384A onto the M-384B. This results in a new failure prediction of 21.8765 days in
4,5 years, which equals 948 tons/year as opposed to the previous 1128 tons/year.
If the same table as above is constructed, but now with the new results on the B-line, the
following result can be seen:

53

Model
2002
A
B
C
total
A
B
C
total
M-322 448 381 196 1025 M-322
K-311
15 105
17 137 K-311
772 579 579 1931
M-384 1165 948 864 2977 M-384 667 404 564 1635
1628 1434 1077 4139
1439 983 1143 3566
Table 5.8: second system level validation spiral dryer system in lost tons per year

As can be seen the results do get better, however the B-line still has a somewhat high
score. This should lead to the conclusion that the B-line actually indeed causes more
trouble than the other two lines and further investigation is necessary here.
This system level approach leads to the conclusion to still consider the model valid on a
system level however the division of production loss at a higher level is not always
correct due to cascading effects.
A same situation can be identified in the interaction between the fitzmills and the F-323
cyclones. The model shows too much dents on the Fitzmills. However, if these plug there
will be symptoms felt in the cyclones in front of the Fitzmills. To make the comparison
more real the dents of these two assets need to be added:
Asset
M-321A
F-323A
Total

Model
2002 Asset
710
- M-321B
292
477 F-323B
1002
477 Total

Model 2002
608
223
202
36
810
259

Asset
M-321C
F-323C
Total

Model 2002
309
190
222
270
531
460

Table 5.9: System level validation Fitzmill system

From this table it can be concluded that the systems approach for the Fitzmils and the
cyclones reduces the difference between the results generated by the model and the
results obtained from the dent administration. However the approach does not eliminate
the difference. This could mean that there has to be more equipment involved in the
system. Most likely this will be the precipitation line. Because, if the Fitzmill plugs the
result has to be felt either earlier in the system (the precipitation lines), or later and after
the Fitzmills are the drying lines. If the involvement of more equipment still does not
eliminate the difference the conclusion can be made that the reason is more fundamental,
like incorrect modeling or wrong data analysis.
To evaluate if the statement of involving more equipment makes any sense the previous
system of M-322, K-311 and M-384 is added, then the whole system of precipitation to
drying is covered.

54

Asset
Model 2002
F-323A
292 477
M-231A
710
M-322A
448
K-311AB
15 772
M-384A
1165 667
Total
2630 1916

Asset
Model 2002 Asset
Model 2002
F-323B
202
36 F-323C
222 270
M-231B
608 223 M-231C
309 190
M-322B
381
- M-322C
196
K-311C
105 579 K-311DE
17 579
M-384B
948 404 M-384C
864 564
2244 1242
1608 1603

Table 5.10: System level validation spiral dryer system and fitzmill system combined

The results generated by the model do seem to get closer to the observed dents in 2002
and at least two lines are in the same order of magnitude. However, the B line remains an
outlier. The large difference observed between the results of the two systems for the
spiral dryer (M-384) can be further explained. When there is a problem with the spiral
dryer so that it needs to be cleaned or repaired (out of function), the line is transferred to
the bypass, which takes 8 hours. In the model every time the bypass is used the line will
switch back to the spiral dryer as soon as it is repaired. However, in reality this may be
postponed for quite some time because this takes a lot of production time and the plant
can be run through the bypass. Although, when it runs over the bypass the maximum
capacity of 650 tons a day cannot be reached. If the system runs over the bypass for a
longer period of time, the production target will be adjusted and the dents (measured by
650 tons a day) will not be allocated anymore. Estimation by plant experts is that the
dents allocated to the M-384s can be increased by 20% to include these dents. Then the
last bar of the table above would look like:
A-line Model 2002 B-line Model 2002 C-line Model 2002
Total
2630 2049
2244 1322
1608 1715
Table 5.11: System level validation of spiral dryer and fitzmill system, including 20%
increase of the spiral dryers.

Striking is that also in this evaluation the B line has very high outcomes in the model
compared to the dent administration in 2002. This can be due to two reasons: either the
B-line has not been up all the time in 2002, due to which there are less dents allocated to
this line. However, this is not very likely because in this case all the assets in the B-line
should have smaller dents than the assets in the C-line, which is not true.
The second possible explanation is that the B-line has had more trouble during the past
couple of years which have been resolved in or before 2002. Then the failure frequency
distribution is fitted on the period with these troubles and the dents are a result of the
period without these troubles. However, because this difference occurs on several assets
this is also not a very likely explanation.
Even in the discussion with several experts there has been no possible explanation or
assumption found that could explain the difference of the behavior of the B-line in the
model as opposed to the behavior in the dent administration. This is therefore an
interesting finding to do further research on.
The same comparison can be made for the overall sections that have been modeled. The
results of these comparisons are shown in the tables below.

55

TransWeigh Model 2002


F-3107
41 261
V-352B
392 461
V-354
108
83
F-306
819 163
Total
1360 968
Table 5.12

Precipitation Model 2002


M-316B
553 332
M-316C
553 343
M-316A
424 565
M-321B
608 223
M-321C
309 190
M-321A
710
N-336
65 165
F-323A
292 477
F-323B
202
36
F-323C
222 273
Total
3938 2694
Table 5.13

Drying
M-384B
M-384C
M-322B
M-322C
K-308
M-384A
M-322A
V-342A
K-313
K-311
Total

Model 2002
948 484
864 676
381
196
10 115
1165 800
448
160 113
- 681
137 1931
4309 4800

Table 5.14

Model
TransWeigh
1360
Precipitation
3938
Drying
4309
Total
9607

2002
968
2694
4800
8480

Table 5.15
Table 5.12-5.15: Validation of model results on section level (lost tons a year)

As can be seen, an explanation needs to be found for the fact that the model predicts more
production loss and downtime than the dent administration in reality shows. Reason can

56

be that the lost production predicted by the model is calculated with the assumption that
there is a constant throughput of 650 tons a day. However, the plant does not always
operate stable and there are certainly occasions that the throughput is less than the
assumed 650 due to process upsets, grade change, external factors, etc. Also the grade
changes are not included in the model, naturally these do take time, and result for some
grades in lower output however this is something that cannot be captured by the model. If
adjustment is made of the production targets (planning) this is not regarded a dent in the
dent administration. The model does show it as lost production.
From the dent administration it can be calculated that the Resin plant had an overall
throughput of 85% of its maximum capacity. This means that all results found by using
the model should be multiplied by 0.85, which brings the results of the model closer to
the results of the dent administration.
If the 25% boundary is used again than the next tables can be built:
Model
1360
3938
4309
9607

TransWeigh
Precipitation
Drying
Total

85%
1156
3347
3663
8166

2002
968
2694
4800
8480

25%?
Y
Y
Y
Y

Table 5.16: Validation of model taking the utilization of 2002 in account

As an extra verification method a graph can be used that is made of the input from the
dent administration, however the allocation of the dents has been adjusted by hand to the
right assets.

5338

Human Failure, Process & M aintenance Dents 2002


(per unit operations)
Loss

2000

1220

1236

3000

636

4000

3218

3543

tons/counts

5000

509

6000

1000

em

ne

st
sy

os

th

er

Ph

sy
st
A
BP

ge

em

ge
t ri
fu
en
C

Fo

rm

n
ti o
ac
re
&

&
c
on
C

ry

in

&

si

lo

tra

pr

ns

ec

po

ip

rt

Graph 5.3: Registered human failure, process and maintenance dents in 2002.

57

The reason that here the graph is used and not the graph with solely maintenance dents is
that during the data analysis to determine the failure and repair frequencies also the work
order that concerned cleaning etc. due to fouling (process cause) also have been included.
These work orders have been included because these actions are performed by the
maintenance department, and in that case are considered maintenance although the root
cause may be due to the operating process/conditions. EMPAC is used as a planning
system for the maintenance personnel and work and thus also contains these work orders.
If the systems drying & silo transport, concentration& precipitation and BPA systems are
taken into account, these results can also very well be compared with the outcome of the
model.
Since both ways of validation eventually give results that are in the range of 25% with
reality it can be concluded that the model that has been built in this project can be used as
a starting point for reliability simulation. However, it is not true that the model is fully
representative for the situation as it is in reality. This is mainly due to four reasons.
The first is the capability of the software used to represent the behavior of the
system as it is in reality, mainly due to the inability of the software to model
reduction of throughput instead of modeling the asset fully up or down. Closely
related is the lack of intermediate storage and delay time modeling in the
software.
The second is the difference in the nature of the data used. As explained earlier
the failure frequencies determined from EMPAC that generate the downtime of
assets in the model really concern the root causes, as in the dent administration
the dents are allocated to symptoms.
Third may be the unreliability of data used in the determination of failure
frequencies and repair times.
Fourth can be the input of fouled feedstock due to process upsets in upstream
plants like BPA and chlorine. This can cause problems in the resin plant like
fouling, which can hardly be modeled.
A more extensive review of these explanations will be given in the conclusions and
recommendations.

5.7

Example of model use

To illustrate the actual benefit of the model in this chapter quantified numbers gained by
running the model will support a decision recently taken in the Resin plant. Two cases
will be illustrated by use of the output from the model.

5.7.1 Extension of time between shutdowns


The first case will deal with the transition from a yearly shutdown to a four-yearly
shutdown, which is currently under investigation. The major interest when making the
decision to change from a yearly shutdown to a shutdown every four years is whether the
delay of shutdown will increase unreliability and unplanned failures and downtime to an
extent that it will destroy the uptime gained by postponing the shutdown. In other words,
58

the time won by postponing the shutdowns should be significantly more than the increase
of downtime due to unplanned failures.
To make this calculation initially the model sections are run for a period of 4,5 years in
which one shutdown of 28 days is included. Only if there are significantly different
outcomes than expected, the review of the sections will be drilled down into asset level.
Availability
TransWeigh
Precipitation
Drying L1
Drying L2
Drying overall

Yearly 4-yearly Y without PM 4Y without PM


0.9657 0.9828
0.9988
0.9988
0.9490 0.9665
0.9832
0.9835
0.9426 0.9596
0.9766
0.9767
0.9592 0.9763
0.9933
0.9933
0.9655 0.9828
0.9996
0.9998

Table 5.17: simulation results for a time between shutdowns of a year and
of 4 year.

From these numbers it can be concluded that the availability (without preventive
maintenance) of the various sections is hardly influenced by an increase of failures over
time, when the time between shutdowns is increased. Thus the influence of the
availability won by not performing a shutdown every year is of far greater importance,
and a lot more time is won by increasing the time period between shutdowns than there is
lost by increasing failures.
In all cases the difference between the overall availability of a yearly shutdown and a
four-yearly shutdown is approximately 1,7 %. This number seems very accurate since the
standard deviation of the model runs is very small (Appendix G). However, the input
functions have been estimated and may thus be assumed in the right order of magnitude
but not very accurate. An increase of 1,7% means that, assuming the plant produces
225.000 tons a year, this will lead to a possible increase of production of 3825 tons a
year.
If no simulation model would have been used the following calculation could have been
made:
Total shutdown time yearly vs. 4-yearly shutdown:
Yearly:
2 weeks * 4 times = 8 weeks in 4 years
4-yearly:
4 weeks * 1 time = 4 weeks in 4 years
Thus: gain of 4 weeks of production with extension of time between shutdown. Average
over 4 years the gain is a week a year.
Production in one week is: 7 days * 650 tons/day = 4550 tons.
The conclusion can be made that there will be less increase of production than is
calculated by the increase of possible uptime. The fact that this would be so (less increase
of production due to downtime in the won time period) certainly has been expected.
However, without a simulation model there is no possible way in which this difference
can be quantified.

59

In the third and fourth column the availability of the sections without the preventive
maintenance activities is listed. By comparing the third column (availability without
preventive maintenance in the situation of a yearly shutdown) with the fourth column
(availability without preventive maintenance in the situation of a 4 yearly shutdown) the
main conclusion that can be made that the unplanned downtime of the resin plant is
hardly affected by increasing the time between shutdown.

5.7.2 In-line cleaning M-316 steam jets


In the next shutdown there will be a modification of the M-316 jets. The aim of this
modification is that the jets no longer have to be removed from the line to be cleaned.
However, the fouling rate will not change so they will still need to be cleaned regularly.
For this action the lines need to be shutdown and emptied before cleaning can take place.
The estimation is that this will result in a repair time of approximately 4 hours instead of
the previous 6 hours for plugged jets.
If this is adjusted in the data analysis the new repair time distribution function for the M316s will be a normal distribution with a mean of 0.1818 and standard deviation of
0.0454.
This generates the following results when entered in the model, and ran for a time period
of 4.5 year and a time between shutdowns of 4 years. In this table the downtime is given
in days for 4.5 years and the production loss is in tons a year.

M-316 A
M-316 B
M-316 C
Total

Old situation
New situation
Downtime Production Downtime Production Difference
(days)
loss (tons)
(days)
loss (tons)
(tons)
6.4646
374
5.304
306
68
12.3621
536
8.908
386
150
12.3145
534
9.08
394
140
1444
1086
358

Table 5.18: downtime and output loss in current and proposed situation.

5.8

Conclusions and recommendations

Because the model is proven to be valid, the conclusions and recommendations listed
below can be made based on the outcome of the model. The conclusions can be divided
in conclusions concerning general insight provided by the model, and conclusions based
on the examples performed with the model.

5.8.1 General insight


Based on the results from the model, it can be concluded that the B line causes much
more downtime and production loss than the A or C line. Thus, there needs to be extra
focus on the B line. This includes measuring of the troubles that occur and extra focus
when maintenance or structural improvement programs are performed. In this way either

60

the cause of this difference can be determined and eliminated or the effects will be
minimized.
The second general conclusion that can be made is that the Fitzmills generate far more
(few hundred tons) production loss than until now has been assumed. This is a good
reason to give the Fitzmills extra attention or perform a structural improvement study on
them.
The third general conclusion that can be made is that the spiral dryers (M-384s) are very
troublesome. They need to be subjected to a thorough improvement program, which is
already running. In the shutdown of 2003 the project once through drying line 1 project
is implemented, in which the M-384A is designed out. This means that the A spiral dryer
will be removed from the drying line.

5.8.2 Conclusions from examples


Extension of the time between shutdowns will increase availability of the studied sections
by approximately 1.7%. Even when the shutdown is not taken into account availability
stays equal, or even slightly increases. This may seem odd. However, considering the fact
that most failure distributions are Weibull with a smaller than 1, this is exactly what is
expected. A smaller that 1 means that there will be decrease of failure over time. This
indicates that the equipment in the plant never reaches steady state due to maintenance
tasks. This can be supported by the dent administration that shows a decrease of dents
when the plant runs for a longer period of time.
The planned adjustment of the M-316 steam jets in the next shutdown will decrease
output loss with approximately 25%, which equals 358 tons a year.
Recommendations that can be made from these both sections are:
Closely monitor the B line, perform improvement programs.
Perform improvement programs on the Fitzmills.
Extend time between shutdowns.
Perform adjustment M-316s.
Design-out of the M-384 B & C in SD 2004.

5.8.3 General simulation modeling


From the development of the simulation model for the Resin plant several conclusions
and recommendations can be made with regard to the implementation of simulation
modeling in GE Plastics.
The first conclusion that can be made is that the output generated by this model was not
completely in line with real production loss and expectations. Although it will be useful
to apply this model, the observed deviations need to be corrected. There are three
possible explanations that already have been stated previously. In this section a
conclusion and recommendation to improve will be made.

61

The first is the incapability of the model to simulate intermediate storage and partly
reduction of throughput as opposed to fully up or down as the software used in this
project does. Although this software has been specifically selected for this model, this
leads to the recommendation to use another type of software when implementing
availability modeling in larger scale. From the evaluation of available software
previously in this report the packages Titan, TARO, and Optagon can be selected having
the best performance in this field.
The second recommendation concerns the different ways in which dents are allocated in
the two database systems. The data in EMPAC are work orders and only reflect the work
performed by the technical services on the root causes of the problems. The data in the
DAS are allocated to the asset where the symptom is observed and thus these two are
hard to compare. Moreover, if there are mechanical or process problems that can be dealt
with without the help of the mechanical or electrical service (by operators or contractors),
there is no work order placed in the CMMS, thus this information has been discarded
when determining the failure and repair time frequencies. The other way around there are
work orders that physically use the technical service, but in fact the problem is caused by
a process upset/problem. In this case the information is included in the data analysis and
the model ends up being a representation of failures due to more causes than just the
mechanical ones.
Thus when implementing modeling in a larger scale it needs to be considered what we
exactly want to model (only mechanical failures or also failures due to other reasons)?
Data accuracy and completeness must be improved at all times to enable higher quality
models. For the allocation of dents in DAS it can be concluded that they are now
allocated to the asset that shows the symptoms, however, somewhere needs to be
recorded what has been the root cause of this symptom and what type (human error,
process, mechanical, etc.) it is.
Last recommendation concerns the use of the CMMS. When using the data from the
CMMS to determine failure and repair time distributions several observations were made.
Firstly, when an asset is replaced by a new asset, the old one stays in the CMMS. There is
a guideline for the numbering of the new one, however, this is not always followed as it
should. This causes confusion and time-consuming search work in the data analysis phase
of the modeling.
The second point is somewhat related, and deals with the modification of equipment in
the plant. Not always is it obvious from the work order that a modification has taken
place, although this is a specific category that can be selected in the WO sheet. It is
extremely important to know if and when modifications have been performed on a certain
asset because these influence the failure frequency. In the determination of failure and
repair time frequencies only the data from the modification on should be taken into
account to achieve realistic distributions.
Third concerns the work order field in the CMMS in which it is possible to state whether
the action performed on the asset brought the system down or not. This should be
populated in order to be able to fit a more reliable distribution.
Fourth is that there is the possibility to fill repair times in the system, however, this is
actually never done in reality. This is a very important improvement to make in order to

62

fit more accurate repair time distributions. In this project repair times have been
estimated, it is more favorable to fit distributions on actual data than on estimates.
Fifth adjustment in the use of the CMMS is that a lot of work orders have been found
under tagnumbers where they actually should not have been stored. There frequently are
jobs like holes in the floor, check of stairs etc. placed under an asset, which they are
physically close to. This is because they need to be planned in the time schedule and need
a cost allocation. However, this does not justify the implementation of them into a wrong
section in EMPAC. A separate asset number for general equipment like stairs and floors
can be made to solve this problem.
Lastly, to be able to make the appropriate selection and data analysis, EMPAC needs to
be accessible. It should be possible to select and copy data to another application in
which further analysis or modification can be made.
Final conclusion that can be made is that simulation modeling can be of a great added
value to the way in which GE Plastics deals with the decisions made in maintenance
engineering. However, to really be able to perform availability simulation modeling in a
structural way, the problems that have been explained above need to be resolved.
Summarizing, this leads to the following recommendations:
Select another type of software that is more capable of modeling process
properties like process flows and intermediate storage.
Close the gap between the two different types of measuring maintenance work
and output dents. Or, at least, register what the differences are and
consequently apply one measure within a system.
Use the CMMS to its full extent, meaning:
Make sure the numbering of equipment is always consistent with the
existing rules. This also means correct (in)active selection in the asset
sheet.
Clearly register modifications performed on equipment in the CMMS.
State whether the asset is shut down or not during the performed
action.
Include repair times in the system
Make sure the right work is under the right asset number. Add an asset
number for the storage of work orders on general equipment.
Make the system accessible for reporting and data extractions. It
should be possible to select fields, copy them to another application
and work with them.

63

Investigation organizational prerequisites

6.1

Organizational characteristics GE Plastics

Before proceeding with the identification of the prerequisites to implement availability


simulation modeling in a structural way some attention needs to be paid to the
organizational characteristics of the company it will be implemented in. Implementation
will only succeed if availability simulation modeling is fully integrated and accepted in
the organization, something that is not possible if it conflicts with the organizational
characteristics. Three organizational characteristics can be identified: the organizational
structure, the business processes and the people within the organization.
GE is a very large organization with a wide variety of businesses and processes which is
financial driven. This can be seen from the measure on which each business is evaluated:
the profitability. GE can be said to have a divisionalized structure (Mintzberg, 1983).
Characteristic for a divisionalized form is that it works best with machine bureaucracy
structures in its divisions. This can be explained by the standardization of outputs that is
the prime coordinating mechanism in the divisionalized structure. Each division must be
treated as a single integrated system with a single set of operational goals. By operational
is meant that the goals must lend themselves to quantitative measures of performance
control. Within GE the quantitative measure for performance control is a financial one.
GE plastics is obviously one of the divisions of GE and therefore driven to a machine
bureaucracy structure. In a machine bureaucracy there is a large focus on standardization
of work processes.
Due to the fact that GE Plastics is very financial driven there is a strong drive to quickly
make or save money. This sometimes results in more appreciation for short term projects
(since they start paying off fast) than long term projects.
Knowledge intensive projects have the characteristic that they are long term and it is hard
to always see the immediate gain of them. This illustrates that there is a tension between
a financial driven approach of running a business and a knowledge driven approach.
Striking example of the situation described above in GE is the outsourcing of the majority
of the engineering department in Bergen op Zoom, which is financially motivated but
with a risk of losing knowledge.
The first prerequisite for successful implementation in GE Plastics is thus to show the
financial gain of simulation modeling.
Another result of the fact that GE is a divisionalized organization is that there are several
standardized processes that have been implemented in a very thorough way throughout
the complete organization. One of the best examples of processes implemented like this is
the Six-Sigma quality program. This can either be an opportunity or a threat for new
processes. If they can be positioned as complying with these standard processes they will
have a greater chance for success. However, if the new processes do not fit in the
standardized processes it will be very hard to incorporate them in the organization. This
will be the second prerequisite when implementing the tool: to show it is Six-Sigma
compliant.
64

Third prerequisite is one that is valid throughout industry and does not need further
explanation: the acceptance of the tool by the people within the organization that are
expected to work with it.
Naturally these prerequisites derived from the organizational characteristics should be
taken into account when deciding on how to incorporate availability simulation modeling
in the business processes and the organizational structure of the company. The process of
implementation of a new tool as simulation modeling and the tool itself should fulfill
these.

6.2

Analogy process modeling

Since there has been a discrepancy identified between the characteristic of GEP to be
non-knowledge driven and the incorporation of a tool that is very labor and knowledge
intensive it may be useful to investigate the status of another knowledge and labor
intensive tool in GEP. This tool is process modeling. This is the most commonly known
application of simulation modeling in chemical industry. Although this is a complete
different type of modeling it can add to the insight of how such knowledge intensive
activities are dealt with in the organization of GE Plastics.
Process modeling is done for new projects as well as for existing plants to evaluate
possible process improvements and support decisions whether or not to implement them.
Because this is one of the few widespread implemented simulation methods in chemical
industry this is used as an example of how simulation modeling can possibly be
embedded in the organizational structure and business processes.
Process modeling within GE Plastics BoZ is performed by the technology department and
process engineers - on request of the plant. Modeling in Aspen is not always possible
due to the conditions of some processes. Powder processes for instance are hard to
simulate due to fouling and therefore process modeling is not always used within GE
Plastics. However, it is interesting that process modeling is performed by persons not
always physically present in the plant. Also the fact that this is not a standard day-to-day
procedure, but only is performed in special projects can be of interest when used as an
example for availability simulation.
Good thing is that process modeling as a tool is mentioned and considered in the standard
format of project management in GE Plastics. This shows that although it is a very
knowledge intensive tool it can be incorporated in a standard process and be used on a
regular (in some cases day-to-day) bases. This leads to the conclusion that it is very
important for a similar tool to be incorporated in a standardized process.

6.3

Prerequisites for simulation in Reliability Engineering

In the ranking of the companies interviewed in the benchmark there have been several
criteria used to evaluate the companies. These have been chosen because they can be
regarded as the prerequisites for a company to make the step from the maintenance
65

technology phase to the maintenance engineering phase. An approach needs to be


described how GE Plastics can integrate all criteria in its physical organization and its
business processes in a way that complies with the previously described organizational
characteristics. In this section first the processes that will benefit from modeling will be
described. Then the physical location of the model within the organization will be
described and which people should own and work with it. After the specific scenario for
GE Plastics it will be evaluated whether there can be some general conclusions made for
any company that wishes to make this transition.

6.3.1 Business processes


To identify the business processes in which availability simulation and modeling will be
of added value the previously identified functions need to be reviewed.
The first function that has been identified is the one that has actually started this
project: review of the availability of an existing plant.
The second that has been mentioned in the introduction is an EHS function: it can
identify critical equipment and actually quantify the probabilities of failure of the
equipment.
Third that has been identified in the introduction is the process layout for
reliability optimization in an engineering stage, thus in new, greenfield projects,
life-extensions, upgrades and modifications.
Fourth is the optimization of warehouse and spare part policies.
In words this may seem quite an unstructured representation. The next graph will show
the structure and connections of the various functions in which availability simulation can
be used.
Availability Simulation modeling

Maintenance

Develop new
projects

Upgrades,
life-extension
projects

Maintain existing
plants

EHS
W arehousing
Design/Eng.

DFSS

DMAIC

Six Sigma

TIME

Fig. 6.1: The application of availability simulation modeling on various fields and projects

Other business units within GE have already embedded availability simulation modeling
in the standardized quality assuring and improving six-sigma processes. E.g. GE
Industrial Systems has incorporated availability simulation as the statistical analysis the
standardized DFSS process. Since these can be divided in two parts (completely new
66

process design (DFSS) and improvement of existing processes (DMAIC)) they are shown
in the graph at the edges of the time scale. They are mentioned in this section because
availability simulation modeling can be part of and embedded in these processes, like in
Industrial Systems. In a machine bureaucracy as GE Plastics this is the opportunity for a
new tool to be successfully implemented: embedding in the standardized work processes.
Within GE Plastics availability simulation modeling can be incorporated in tollgate 3
(conceptual design) of the DFSS. This step contains the step process tolerancing in
which process simulations are made and process controls are evaluated. In this stage a
step should be included in which the same evaluations are made for the hardware
(equipment) of the plant.
In the DMAIC process availability simulation modeling can be incorporated in the
analyze step where it can be used to define the current performance of the system. In
the same analyze step it can also be used to set the targets for improvement. In the
control step it should be used to measure whether the implemented changes have the
intended effect.
At the same time the staff that are intended to execute this new DMAIC and DFSS
processes need to be trained. For the DFSS process this would be engineers from the
engineering department and for the DMAIC this would probably be maintenance
(reliability) personnel. These are both part of the larger, general department responsible
for the technical side of manufacturing. Only if they are familiar with the tool and regard
it an added value it will really be incorporated into these business processes.
It can be concluded that the application of availability simulation modeling can be used in
a wide variety of applications, and it fits in the corporate quality improvement process.
There is one thing that all these processes have in common: they are executed because
they add to the corporate mission. Like in most companies the prime corporate mission
can be regarded to be an economical one and is easily stated as profit maximization.
However, the mission of profit maximization does not immediately lead to the
application of availability simulation modeling. There is a step in between. The goal of
profit maximization needs to be translated to a certain approach on decision making
throughout the business. In what way do the processes applied in the business need to
perform in order to maximize profitability?
The answer to this question can be stated as risk mitigating decision making (personal
conversation). Availability modeling will only fully fit in the decision making within an
organization when this organization is fully focused on the utilization of the outcome of
the model. The simulation model will only achieve a certain state and authority within
business processes if the decision making is actually depending on the outcome of the
model. By using the model there can be evaluations made of mitigating risks at a certain
cost. These evaluations will lead to the development of certain tactics that can be
executed by the plant personnel from the department on which the tactic is applicable
(EHS, maintenance, operators, etc.) The decomposition from corporate mission into an
improvement strategy by help of availability modeling can be viewed in the following
graph:

67

corporate vision

Mission: $
max. throughput
optimize cost

mindset

processes

Risk mitigating
decision making

Translation by
knowledge

weigh cost vs. risk

Tool: availability
simulation, six sigma

experts,
engineers

tactics

measurement

Execution of
tactics

data collection
and storage

Operations,
technical
services

Oper. TS
automatically
CMMS

Fig. 6.2: The decomposition of the corporate mission in an improvement strategy

The major possibilities for success or failure of the model use are the connections
between the lowest sections of the graph (personal conversation). It requires a lot of soft
skills and awareness programs to actually incorporate the tactics evaluated by the model
and the evaluation of these and the data collection into the day to day work of the
operations, mechanical or electrical services. This is also applicable on the data
collection. This aspect will receive more attention in the description of the process of
change towards the desired situation.

6.3.2 Physical embedding in the organization


When the business processes are identified in which availability simulation modeling
should be integrated a more fundamental statement can be made on the physical place the
model(ing) should get within the organization.
From the description of business processes it can be seen in what way availability
simulation can contribute to the organization. It can be concluded that it is a very
knowledge and labor-intensive activity, which is not easy to learn. Modeling should thus
be a dedicated function.
Because availability simulation modeling is applicable to various fields within the
manufacturing organization, it seems logical to assign the modeling skills to a general
department that can perform studies for several disciplines and plants. In this regard the
most logical place to store availability simulation modeling is the engineering
department. Moreover, aim of applying availability simulation in this phase is the
prevention of equipment layout changes shortly after startup due to unreliability.
At this moment the knowledge to do this is not present in the engineering department.
Therefore a special modeler can be hired or the current engineers can be trained, or both.
GE Plastics is in the transition of outsourcing its engineering department, and thus this
could be an opportunity to acquire modeling knowledge from the market.
However, there is one difficulty with modeling in the engineering department. A
specialist in the engineering department can easily build the model, however, when the
model is filled and ready to use in an ongoing process it will be difficult to keep it in this
place. As soon as the model is filled it becomes plant and project specific. It will never be
68

used in day to day decisions in the plant if the people who are making these decisions are
not familiar with the model.
This leads to the conclusion that there can only be one possible way to deal with the
development and ongoing use of the model in the organization. The development ought
to be done by simulation modeling experts in the engineering department, however as
soon as the population of the model starts a dedicated engineer from the specific plant
should be involved. When the model is fully completed and ready to use in day to day
decision making it should be transferred to a plant specific (reliability) engineer. The rate
of involvement is related to the color in figure 6.3, the darker the color, the more the
involvement.
If figure 6.1, that shows which processes can profit from simulation modeling, is updated
with the people that should perform it, figure 6.3 is obtained.
Availability Simulation modeling

Maintenance

Develop new
projects

Upgrades,
life-extension
projects

Maintain existing
plants

EHS
Warehousing
Design/Eng.

simulation expert (eng. dept.)

plant specific engineer

Fig. 6.3: Involvement of experts in the several applications of availability simualtion modeling.

In the most recent reorganization in GEP all the technological aspects of plant operation
have been placed in the same organization. This organization contains maintenance,
engineering, mechanical integrity and warehousing. These are three of the four previous
identified fields on which availability simulation modeling is applicable. This supports
the decision to develop the model in the engineering department and during the
development hand it over to a plant specific (reliability) engineer.

69

6.4

How to reach the desired state?

In this report there has been attention for the technological possibility of implementing
availability simulation and there has been attention for the organizational context of the
model the only question that remains is how to reach this desired state?
Within GE there are two processes that are used to describe and manage changes within
the organization: CAP (change acceleration process) and MOC (management of change).
The CAP process can be applied on any change. The MOC process however is a
functional process for the implementation of major changes or a change that could have
negative impacts to EHS aspects. For this situation the MOC process states various
functional steps that need to be fulfilled.
For the explanation of the change that needs to be made, implementation of availability
simulation and modeling, the CAP process will be used in this chapter because of its
general character. A graphical representation is given below:

Leading change
creating a shared need
shaping a vision
mobilizing commitment
current
state

transition
state
making change last

improved
state

monitoring progress

Changing systems & structures

Fig. 6.4: CAP process decomposed

As can be seen in the graph the CAP process can be divided into 5 steps: creating a
shared need, shaping a vision, mobilizing commitment, making change last and
monitoring progress. These processes can be summarized in two ongoing processes:
leading change and changing systems & structures.
In the project of introducing availability simulation in GE plastics this report should
provide the answers for the processes creating a shared need and shaping a vision. The
shared need is a result from the fact that this type of simulation is recognized in all fields
of application as a next goal in order to improve performance. Vision presented here is
the embedding of availability simulation as a standard tool in the engineering department.
The processes that need to be started after this study are mobilizing commitment, making
change last and monitoring progress.
From the decomposition of business processes it is known that there are various fields in
the manufacturing process that can benefit from availability simulation modeling. Within
these various fields there will be general staff that will execute the tactics decided upon

70

with help of the model. Commitment is needed from all these people. How can this be
achieved?
First of all there needs to be started with changing systems and structures by the design
and implementation of simulation modeling in carefully chosen projects. It must be
completely obvious that these projects will benefit a great deal from the use of the
simulation model. This involves either high-risk or high investment projects. In addition
to the importance of the output of the model also the time in which results are obtained is
important. The projects should provide quick-wins.
After identification of such a project, the team involved should be identified. Apart from
the outsourced modeler this team should at least include the plant specific engineer, an
operator, someone from the mechanical and someone from the electrical services
department to be sure all areas involved are covered. The team leader is responsible for
the fact that all team members should have the feeling that they are responsible for the
success of the project. The benefit of having a cross functional team is two fold. On the
one hand there is a large scope of information these people will contribute. On the other,
they will all influence their own surroundings.
By implementing modeling in projects there budget can be found. If timely known in
advance the costs of the model can be incorporated in the budget request for the project.
If the budget would have to come from a fixed budget for one of the fields it is applied in
funding would be much more difficult.
When the team is working on the selected project there needs to be attention for how to
make the change last. In what way must the project be executed in order to achieve this?
The most important thing for the ongoing use of the model is the involvement of the plant
specific engineer. This engineer should be completely familiar with the model and its
functions by the time it is transferred to him. The next important thing is the
measurement and evaluation of the tactics implemented by the model. Only if these are
evaluated and registered well the model will prove its functionality and used more
frequently.
To achieve a lasting change it is also necessary to incorporate the simulation modeling in
continuous processes performed in the organization. In this case this would be in the SixSigma DMAIC process as discussed previously. However, it needs to be taken into
account that people need to be made familiar with this technique before they will start
using it.
Last step in making the change last is that modeling must be made part of a standard
function in the engineering department. This will make sure that there will always be
sufficient people trained and educated to work with this type of simulation and modeling
tools.
As stated above the measuring and the evaluation of the results obtained by help of the
simulation model is extremely important in the acceptance of the tool. The amount of
models in use and the successfully finished projects with help of the model. To
demonstrate this there should be a global evaluation of the tool after finishing the first

71

couple of projects that have been strictly selected on the criteria mentioned above. When
these have positive results this will encourage others to also start using the new
simulation technique.
All these steps together should result in a structured way to change systems and structures
whilst not overseeing important threats or opportunities of this change. However, just
following these steps will not necessarily lead to success. Just as important here are the
way soft skills are carried out.

72

Conclusions and recommendations

This study has been performed in an industrial organization. This has resulted in a high
focus on the practical application of the tool and the opportunities for implementation.
There has already been extensive research done on the theoretical background of
availability simulation modeling and the possible benefits for chemical industry. Up till
today there are not many companies that actual benefit from the oppertunities that
availability simulation modeling provides. This project has been an attempt to overcome
the gap that there is between theoretical knowledge and related tools and the actual
situation in an industrial company. This has resulted in a project that deals with many
aspects related to the implementation of availability simulation modeling. These aspects
have not been investigated on a fundamental level. A more high-level approach was
followed to make sure all aspects were taken into account on a functional level.
There are two levels at which conclusions can be drawn after this project. The first are
the conclusions and recommendations specifically for the GE Plastics case. The second
are general conclusions and recommendations with regard to simulation modeling for any
company in the chemical industry. These are treated separately in this section.

7.1

GE Plastics

This study has two aspects (the technical content of modeling and the organizational side
of implementing the model) that have been treated separately until now. In this section
these separate parts are brought together and an overall conclusion with regard to the
potential of simulation modeling within GE Plastics shall be given. The conclusions
concerning the content of the model that has been made of the resin plant will not be
repeated here and can be found in the conclusion after the modeling section, in chapter
6.8.

7.1.1 Conclusions

Availability simulation modeling has been proven technical feasible and


organizational feasible within GE Plastics.

Availability simulation modeling will be of added value in various fields of


application within the manufacturing organization of GE Plastics. The model built in
this study provided valid output. To optimize maintenance strategies and actions the
available data need to be improved.

This study has been started from a maintenance perspective but during the project
several other fields (EHS, warehousing, engineering) have been identified that can
benefit from availability simulation modeling.

The data available from the CMMS (EMPAC) have been sufficient in quantity and
quality to prove validity, however, when starting simulation in large scale it is

73

necessary to improve the quality and quantity of these data further by adding more
detail.

Availability simulation and modeling fits in the business processes of the GE Plastics.
It fits very well in the standardized Six-Sigma quality improvement processes DFSS
and DMAIC.

Availability simulation and modeling is a knowledge and labor-intensive activity. But


it is quite feasible if modern data systems are organized and utilized well in the
organization. This requires a high dedication of the modeler. Because of the
application of simulation modeling on several fields of the manufacturing
organization this activity should be a general function and is best placed in the
engineering department. Due to the organizational characteristics of the company this
will initially result in outsourcing of the build of the model.

Although there exist standardized change processes (MOC, CAP) the failure or
success of the implementation of modeling depends on the way soft skills are
incorporated within these standardized processes. These soft skills mainly include
evaluation and communication in order to prove and convince people of the added
value of simulation modeling.

7.1.2 Recommendations

When simulation modeling is started in a more structural way GE Plastics should


initiate improvement of the utilization of the CMMS. This will increase the ease and
quality of modeling in the future. In what way the use of the CMMS should be
increased and improved can be found in the conclusions and recommendations after
the modeling part, chapter 6.8.

When implementing on a large scale it should be considered to use another software


type than has been used for this project. The software should improve on possibilities
to handle process flows, intermediate storage, programming of special situations and
possible translation to fault trees.

Implement the simulation modeling technique in a logical way. Identify projects


where the model will have a large added value like high-risk projects or projects with
large investments within a short period of time and start with these. This will show
the financial gain (amongst others) of the model and increase the chance for success a
great deal.

Develop a global database in which all equipment types, failure and repair time
distributions and already developed models are stored. This will decrease the problem
of data availability and thus increase the ease of modeling with a great deal. It should
be investigated whether this can be a module within EMPAC or whether a separate
system needs to be developed. The ownership depends on the outcome of this
investigation.
74

Start with the implementation by selecting a suitable project and assigning an design
engineer and a plant specific engineer to the project. These need to be narrowly
involved by an outside expert and trained in the use of the model.

7.2

General

From this project several conclusions can also be made with regard to the status of
availability simulation and modeling in the chemical industry.

Availability modeling is not widely spread throughout industry because most


companies are still in the maintenance technology phase with their maintenance
policy and are thus focusing on the implementation of structural improvement
processes. However, it is regarded as a valuable tool for availability optimization in
design and operating stages and will be the next step after the structural improvement
programs. At this stage, however there are still more simple benefits to be gained.

Simulation modeling is now mainly performed by consultants because it requires


specific modeling knowledge and is a labor intensive activity and therefore it takes
effort to implement this in a manufacturing organization.

Due to more strict laws and regulations that require more and more quantified
scenarios etc., there will be a trend that will increase the use of simulation and risk
quantification. A rapid increase of simulation modeling can be expected.

It is technically feasible to implement availability simulation modeling in a company


when it has fulfilled the following prerequisites:
o There is a CMMS used at a level that it provides enough accurate data to fill
the model.
o There is an independent database to validate the model that uses the same
measurement system as the CMMS.
o An appropriate type of software is available.

It is organizationally feasible to implement availability simulation modeling when a


company takes the following aspects into account:
o Availability simulation modeling should fit into the business processes within
a company.
o The model should be owned an appropriate part of the organization. This
depends on the function the model should fulfill.
o People should be trained.
o A process should be developed for implementation of the tool, which at least
contains guidelines for which projects to select for implementation and how to
evaluate the projects.

75

7.3

Further research

There are hardly any companies in the chemical industry that perform availability
simulation and modeling in a large scale, there is only one exception in this respect, and
thus it is hard to find an example to compare with. There are other industries though, in
which simulation modeling belongs to the standard procedures, like the refining industry
but mainly (nuclear)power plants and the army (navy and air force) It will be interesting
to perform a benchmark study between these industries and try to apply this methodology
to the chemical industry.
It was found that availability simulation modeling has much the same interests as fault
tree analysis and risk quantification used in EHS to create scenarios. However, it is not
clear to what extent these two can be combined and benefit from each others existence.
Does extension of a modeling application with these EHS aspects improve the tool or
does it become too extensive and thus user unfriendly?
Availability is now regarded a fixed result from the designed process and plant. First the
process is designed in the conceptual stage and only after the conceptual stage reliability
is taken into account, if it is taken into account at all before construction and start-up. It
can be very interesting to start reliability studies and modeling even before the conceptual
design stage.
Although this study is an attempt to overcome the differences between the knowledge of
availability simulation modeling in literature and the application in industry there will
naturally be some pitfalls when actually implementing the tool. It will be very interesting
and useful to follow an implementation process in industry and combine the newly
identified prerequisites and lessons learnt with the ones already identified in this study.

76

References

Armstadter, B.L., Reliability mathematics: Fundamentals; practices, procedures,


McGraw-Hill book company, New York, 1971.
Avontuur, G.C., Reliability analysis in Mechanical Engineering Design, 2000.
Blann, D.R., Reliability as a strategic initiative to improve Manufacturing
Capacity, Throughput and Profitability, Marshall Institute.
BS4778 (2000), Glossary of terms used in quality assurance including reliability
and maintainability terms, London: British Standards Institution.
Campbell, J.D., Jardine, A.K.S., Maintanance Excellence optimizing equipment
life cycle decisions, Marcel Dekker inc., New York, 2001
Cochran, J.K., Murugan, A., Krishnamurthy, V., Generic Markov models for
availability estimation and failure characterization in petroleum refineries,
Computers & Operations research, Elsevier Science Ltd., February 1999.
CPR 20, Regeling Informatie-eisen BRZO 99.
Dekker, R., Groenendijk, W., Availability assessment methods and their
application in practice, Elsevier Science Ltd., 1995.
Dubi, A., Analytical methods versus the monte carlo method overcoming the
limitations of analytical methods in systems engineering, Clockwork Solutions.
Fritz, A., Bertsche, B., Algorithms for the Monte-Carlo simulation of the
Reliability & Availability of Mechanical systems, University of Stuttgard,
Germany.
Goel, H.D., Grievink, J, Herder, P.M. and Weijnen M.P.C. Integrating
Reliability Optimization into Chemical Process Synthesis, Reliability
Engineering and System Safety, 2002, Vol 78, Issue 3, pp 247-258.
Jones, W., Kubenka, J., Ledet, W., Business driven reliability at Mobils
Beaumont refinery, Marshall Institute.
Korver, B., The Monte Carlo Method and Software Reliability Theory, February
1994.
Lamb, R.G., Availability engineering & management for manufacturing plant
performance, Prentice-Hall, New Jersey, 1995.
McKenna, T., Oliverson, R., Glossary of reliability and maintenance terms, Gulf
publishing company, 1997.
Mintzberg, H.., Structure in fives: Designing effective organizations, Prentice Hall
Inc., 1985.
Moubray, J., Reliability Centered Maintenance, Butterworth-Heinemen Ltc.,
Oxford, 1991.
OConner, P.D.T., Practical Reliability Engineering, John Wiley & Sons Ltd.,
Chichester, 1995.
Pistikopoulos, E.N., Vassiliadis, C.G., Reliability and maintenance considerations
in process design under uncertainty, Elsevier Science Ltd., 1998.
Pistikopoulos, E.N., Vassiliadis, C.G., Maintenance-based strategies for
environmental risk minimization in the process industries, Journal of hazardous
materials 71, Elsvier Science Ltd., 2000.
77

Rijn, F.H. van, SWOT analysis SPARC versus Monte Carlo Simulation, July
1998.
Rotab Khan, M.R., Zohrul Kabir, A.B.M., Availability simulation of an ammonia
plant, Elsevier Science Ltd., 1995.
Seveso II Directive [96/82/EC]
Smit K. , Reference information model for technology and maintenance
departments, Cap Gemini publishing, Rijswijk 1992.
Sommariva, C., Hogg, H., Callister, K., Maximum economic design life for
desalination plant: the role of auxiliary equipment materials selection and
specification in plant reliability, Elsevier Science Ltd., April 2002
Wiliams, J.P., Predicting process systems, Hydrocarbon engineering, July 2001.
Wolstenholme, L.C., Reliability Modeling A Statistical approach,
Chapman&Hall/CRC, 1999.
Woodhouse, J., Education & Training in Asset management, ERA Engineering
Asset Management conference, October 2001.

Websites:
www.jardine.co.uk
www.reliasoft.com
www.iespds.com
www.isograph.co.uk
www.clockwork-group.com
www.optagon.info
www.resnapshot.com
http://web.utk.edu.com
Internal GE publishing:
Change Acceleration Process: CAP kit
Delta consult, Clockwork group, SPAR analysis and report: Maintenance and
inspection practices at GE Plastics BPA1 plant, Netherlands, 1997.
Jacobs engineering, Dry chlorine compression unit reliability study Lexan Plant
Burkville, Alabama, 2000.
Interviews:
Dauwe, R. , DOW Chemicals
Ginniken, F. van, DuPont Nemours
Heus, R. de, DSM
Loos, M., Koster, P.F., Huntsman polyurethanes
Muileboom, A., Exxon Mobil
Ruiter, J. de, Shell Global Solutions
Weijst, A. van der, Shell Chemicals
White, T., Shell Global Solutions

78

Appendix A list of abbreviations


BPA
CMMS
DAS
DFSS
DMAIC
EFD
EHS
EMPAC
EPC
FMEA
FMECA
FTA
GE
GEP
LCC
MTBF
MTTR
pdf
PFD
RAM
RBD
RBI
RCM
RE
SD
TTF
TTR

Bis Phenol Acetone


Computerized Maintenance Management System
Dent Administration System
Design for Six-Sigma
Define Measure Analyze Implement Control
Engineering Flow Diagram
Environment, Health and Safety
Enterprise Maintenance Planning and Control
Engineering Procurement Construction
Failure Mode and Effect Analysis
Failure Mode Effect and Criticality Analysis
Fault Tree Analysis
General Electric
General Electric Plastics
Life Cycle Costing
Mean time between failures
Mean time to repair
Probability Density Function
Process Flow Diagram
Reliability, Availability, Maintainability
Reliability Block Diagram
Risk based inspection
Reliability centered maintenance
Reliability Engineering
Shutdown
Time To Failure
Time To Repair

79

Appendix B - Suppliers and software packages


Reliasoft
BlockSim
BlockSim is based on the technique of Reliability Block Diagrams. The architecture of
the system needs to be constructed by the user, as well as the data used to calculate
failure distributions for the various blocks (Weibull++ is linked to BlockSim). When this
is done correctly the program can calculate the overall reliability for a mission time and
the probability of failure for a mission time. The package can perform conditional failure
and conditional probability of failure calculations, it can calculate the failure probability
at any given time (age) and the system MTTF. Another possibility is to determine the
important measures for each component relative to the system at any time (age). This is,
which component has the greatest impact on the system reliability.
Shell Global Solutions
SPARC
SPARC (System for Production Availability and Resources Consumption) is an
analytical tool developed by Shell Global Solutions in order to do quick calculations on
reliability issues. As an input the program requires unit capacities, network configuration,
component reliability characteristics and maintenance strategies. Because the system has
an analytical engine it can perform very fast what-if analyses. SPARC is a very easy to
use tool that fits very well in an engineering perspective. It can be used in design and
operation and as a living tool that is updated with field experience where it directly
supports engineers in their daily decisions.
However, due to the practicality of the tool it is not possible to model very complex
systems, changing configurations, buffering, throughput, maintenance resource
allocation, operating rules and multi-product processes.
Jardine
MAROS
MAROS (Maintainability, Availability, Reliability, Operability Simulation Simulator)
predicts system effectiveness in terms of reliability, availability, and productivity. It can
be used to quantify the life-cycle cost effectiveness of a system, and conduct optimization
to improve profitability. MAROS adds another dimension by integrating operability,
logistics, and economics to RAM. MAROS can be utilized throughout an assets lifecycle, from early concept screening through to business re-engineering of an existing
plant. MAROS provides the ability to alter a model dynamically as a simulation proceeds
because systems seldom remain constant over their lifetime. This feature us used for
long-term planning and adds to the benefit of life-cycle cost-benefit assessment.

80

TARO
The TARO (Total Asset Review and Optimization) simulator is specifically developed
for the modeling of refining and petrochemical plants in which the production efficiency
is a complex interaction between reliability, blending and yield rules, flow routing, and
intermediate storage options.
A TARO model comprises a sequence of individual networks (known as SubNets) that
are related to each other by a set of blending rules and utility requirements. This approach
allows modeling of complex flow routes, which includes alternative routes, mixing rules,
re-cycle loops and disposal options for each of the product streams in the overall complex
(group of SubNets). A SubNet can represent any of the following entities: process unit;
utility supply; storage or batching tank; supplier (feedstock); customer (sales pool); berth;
bulk transport.
A nodal mass balance within each SubNet is maintained along with flow balance across
the entire complex. Disposal rules can be set-up to manage the flow mass balance in the
event of abnormal circumstances occurring, e.g. turndown, re-cycle, re-routing, flaring,
dumping to slop tank. Minimum flow rates can be set for all processes.
Thus the conclusion can be made that TARO distinguishes itself from other packages by
the way it deals with flow rates and intermediate storage.
Isograph direct
AvSim+
This program analyses the availability of systems by using fault trees or reliability block
diagrams. The Monte Carlo simulation engine enables AvSim+ to model complex
redundancies, common failures, aging and component dependencies which cannot be
modeled using standard analytical techniques. The fault trees and RBDs used in this
module need to be constructed by the user, as also the failure frequency data.
In AvSim+ the logical interaction of failures, and how they affect system performance,
are modeled using a network diagram or fault tree. These diagrams may be used to model
failure and success or levels of throughput in the system. Consequences are then assigned
to any level of the logical diagram to indicate the effects of failures (financial,
operational, safety and environmental). Labor, spares and failure data may be imported or
directly entered into the program together with any operational phase information and
task group assignments. Nice feature of this program is that fault trees can be easily
transformed/imported to RBDs and vice versa.
Item software
Toolkit
Item Toolkit contains various modules to perform a total analysis at system or project
level. These modules are: Reliability prediction, FMECA, RBD, Fault tree, Markov,
spare cost, maintain (MTTR).
This is a very extensive package to analyze a full system for a lot of properties and
possible improvements. However, it is possible to only perform an analysis with only one
module from the toolkit.

81

One of the facilities of the package is that it easily imports and exports data to and from
several Microsoft applications and various CAD packages. Also, fault trees, Markov
models and RBDs are linked within this package.
Clockwork
SPAR
SPAR is a software system developed for the predictive modeling of the performance and
life cycle costs of complex expensive assets and systems. The models can quantify the
effects of design changes, spare part availability and preventive maintenance schedules
on system performance and costs. The technique is based on a real-time Monte Carlo
modeling approach. This includes the ability to evaluate time-dependent processes such
as equipment aging, the dynamic effect of maintenance and operating policies and system
wide equipment interactions. However, the program is not as good on flows and
intermediate storage as several others. This has been improved in the new package
developed by clockwork solutions especially for the process industries called STORM.
The setup of SPAR is such that the user needs a great deal of mathematical insight in
algorithms etc. to fill the model, which makes the package somewhat more user
unfriendly than several other packages. For special situations that cannot be modeled by
the standard blocks and connections there is a programming link to the fortran77
programming language.
Fidelis group
Titan
Titan is an advanced, multi-state discrete event simulation modeling tool for system and
reliability modeling that provides accurate predictions of the future performance and cost
of complex systems allowing for more profitable equipment-related decisions. The
simulation is made by a Monte Carlo modeling engine. Titan has been developed
especially for applications in the (petro)chemical industry and is therefore more suitable
for modeling flows and storages than some other modeling software. It also can model
multi state systems instead of the traditional two state Reliability Block Diagram.
Main analyses that Titan can do are: spares analysis, probability to exceed production
target, top 10/20/etc. unreliable equipment, preventive maintenance schedule
optimization, intermediate storage optimization, shipping strategy optimization, etc.
Advantage of using Titan is that it is fully compatible with Microsoft technologies for
interfacing with other applications. This means that it can immediately be linked to
existing databases and common applications like Visio for the graphical representation of
the model and Visual Basic for programming of uncommon setups.
Advantica
Optagon
Optagon is a software package that has been developed by Advantica originally to
evaluate the performance of complex oil and gas production assets. They have used this
software in both the design and operation of such facilities. Optagon represents the

82

studied system by block diagrams that are analyzed by a Monte Carlo simulation engine.
Instead of the traditional two-state evaluation of the system Optagon also allows partial
states of operation. Optagon allows each block to have a specification on failure rate
repair times, maintenance strategies, and related costs. Also the resources and costs for
the performance of maintenance can be entered and evaluated.

83

NODE:
0

T IT L E :
Raw materials

Utilities

product

process

physical/chemical

84
employees

L e x a n P ro d u c tie T ra in - O v e ra ll
process plants

A0

Produce Lexan
Lexan

waste products

specification specification characteristics

Appendix C Idef0 diagram Lexan train

N O .:
1

NODE:

TITLE:

Lexan Productie Straat

85

NO.:

MeCl
TEA
Phenol
PCP
TMTC

Acetone

Phenol

NaCl

H2SO4

CO

Chlorine
plant

A1

produce Chlorine

HCl

H2

BPAI
plant

A2

produce BPA I

BPAII
plant

A3

produce BPA II

Cl2

BPA

NaOH

phosgene
plant

produce
phosgene
A4

COCl2

resin
plant

A5

produce Resin

lexan
plant

A6

compound Lexan Lexan

FCP

NODE:

TITLE:

Lexan
Lexan
Production
Productie
Train
Straat
- Resin

86

NO.:

6-2
28

Steam

process water
Caustic
HCl

Chain
stopper

Phosgene
rec. water

MeCl

Catalyser

BPA

A51

2x formulation
tank

formulate

pumps 3x reactor
+ mixer

storage
tank

A52

perform reaction

reaction
mixture

A53

2x centrifuge line

extract

3x concentration
line

A54

concentrate

rec. water

precipitate
A55

3x precipitation
line

solvent

dry

drying
line

solvent +
steam

A56
Resin

water
to
atm.

to tank park

A57

recover solvent

to tank park

NODE:

TITLE:

Lexan Train - Resin - Reaction

87

NO.:

29

phosgene

chainstopper

formulation

pumps
N-316 A/D

A521

fill reactor

reactors 6x condensor 3x sample pumps


V-313 A/B, V-370
N-317
mixers
M313 A/B, M370

A523

perform reaction

A524

centr. feedtank
V-314 A/B, V371 A/
B

store

water+BPA,
solvent+resin

NODE:

TITLE:

Lexan Train - Resin - Exctraction

88

NO.:

30

NaOH

rec. water

interfacial

pumps
N-318 A/B, N-358 A/
B

A531

transport

5x filter
F-317

filter
A532

2x 6 centrifuges

A533

centrifuge (6x)

conc. feedtanks
V-325 A/B

waterpumps
N-3110

A534

heat exch.
H-3067

store

polymeric
solution

NODE:

TITLE:

Lexan Train - Resin - Concentration

89

NO.:

31

polymeric
solution

pumps
N-327 A/E
N-3110

A541

transport

3x heat
exchanger
H-310

4x heat
3x pumps
exchanger
N-335
H-312

A543

Concentrate

3x condensor 3x cyclone pumps Me stor.tank


H-313 V-332
N-334 V-335

A544

separate

A545

tank
V-333 A/B

store
thickened
polymeric
solution

NODE:

TITLE:

Lexan Train - Resin - Precipitation

90

NO.:

32

resinsol.

5x pumps
N-336

A551

transport

A552

3x pipe
heat exchanger
H-3063

heat

7x steamjets
M-316

3x precipitation
pipe
M-318

A553

evaporate

3x cyclone
F-323

pumps
4x condensor tank
N-338, NH-316 A/B, H-317 V345 A/B
339

A554

separate Damp/
powder

A556

recover solvent

fitzmill
M-321

A555

grind powder

evenly distr.
resin
powder

to tank park

NODE:

TITLE:

Lexan Train - Resin - Drying

91

NO.:

33

resin
powder

3x rotary valve
M-320

A561

transport

column dryer
V-342

A562

pre-dry

3x heat
exch.
3x H-3062
blowers
3x filters
rot.valve
K-311
F-342
M-322 A/
5x
C

A563

transport

A564

spiral dryer
M-384A/B/C

dry

filters
F-326 A, F-327
B

A565

store

postdryers postdryer cyclones


M-331A/B H-319 B F-328

A566

post dry

storage tank
V-341/385

A567

store

4x blower
2x
K-305,
scrubber
F-329 2x ww. 4x pomp 306
H-318 N-340

A568

separate

dry
resin
powder

coolant water
system

solvent
recovery
system
vent gas
system

transport/weeg
sectie

formulation
section

reaction
section

extraction
section

concentration
section

precipitation
section

overall

drying section

Appendix D Reliability Block Diagram Resin plant

92

93

M-3107A

M-3107B

M-3107C

F-3107A

F-3107C

F-3107B

F-306B

F-306A

M-307B

M-309A

V-352B

V-354

M-309B

transportation/weighing

V-354B

94

M-312

M-369

V-312

V-369

Formulation

95

N-316D

N-316C

N-316B

N-316A

V-370

V-313B

V-313A

M-370

M-313B

M-313A

N-317C

N-317B

N-317A

H-335

H-302B

H-302A

reactor system

H-336

H-303B

H-303A

Reaction

M-371B

M-371A

V-371A

V-371B

M-314B/1

M-314A/1

V-314B/1

V-314A/2

96

N-358B

N-358A

N-318B

N-318A

F-317F

F-317D

F-317B

F-317C

F-317A

F-317E

H-307B

H-307A

F-315M

F-315K

F-315L

F-315J

F-315H

F-315G

F-315F

F-315E

F-315D

F-315C

F-315B

F-315A

V-325B

V-325A

M-325B

N-327B

N-327A

N-327E

N-327D1

N-327C1

extraction

97

H-310B2

H-310C2

H-310A2

H-310B1

H-310C1

H-310A1

N-335A

H-312

N-335DE

N-335BC

heating system

H-312D

H-312C

H-312B

V-332/1

V-332C

V-332B

V-333A

V-333B

Concentration

98

N-336A

N-336C

N-336B

H3063A

H-3063C

N-336D

N-336E

H-3063B

N-336F/1

M-316
A1/3

M-316
C2

M-316
C1

M-316
A2/3

M-316
B1

M-316
B2

M-316
A3/3

M-318A1

M-318C1

M-318B1

M-318A2

M-318C2

M-318B2

F-323A/1

F-323C

F-323B

M321-A/1

M-321-C/1

M-321-B/1

Precipitation

solvent/
powder sep.

solvent/
powder sep.

solvent/
powder sep.

99

M-320-A/1

V-342A

M-322-A/1

air blower
systeem
H-3062A

H-3062C

M-384A/1

M384C

M-322-C

V-342C

M-320-C

air blower
systeem

M-384B

M-322-B

V-342B

H-3062B

M-320-B

air blower
systeem

F-326-A/1

F-327B/1

dryingair
system

dryingair
system

M331/1

V-340B/1

V-341/1

M-331B/1

M-341/2

V-385/1

K-308B/1

K-308A/1

K-313B/1

K-313A/1

V-338

drying

F-333

F-373/1

M-374/1

M-380/2

100

F-328D

F-328C

F-328B

F-328A

K-305B/1

K-305A/1

F-329B

F-329

N-340D

N-340C

N-340B

N-340A

H-318B

H-318

K-306B/1

K-306A/1

H-319B

H-319A

drying air system

101

K-311E

F-3103E

K-311B

K-311A

K-311D

F-3103D

F-3103B

K-311C

F-3103C

F-342A

F-342C

F-342B

H-3062A

H-3062C

H-3062B

air blower system

102

N-349 B

N-349 A

V-324

N-357B

N-357A

Z-307A/B

K-314A/B

V-368

V-381A

V-315/1

V-317

H-304/1

V-320/1

K-303B

K-303A

N-346

V-321/1

H-305/1

vent gas system

V-319B/1

V-319A/2

103

V-349

N-343B

N-343A
N-3107B

N-3107A

Z308B/1

Z-308A/1

coolant water system

104

F-341B/1

F-341C/1

N-337B

N-337D

N-337E

F-341A/1

N-337A

N-337C

H-340

H-316

H-316B

H-341/1

H-317

V-345B

V-345

N-339B

N339 A

N338/1

N-338B

N-339C

N-339D

solvent/powder separation

V-309

V-367

105

V-326

N-355A/1

N-355B/1

F-380A

V-346

F-380B

N-356B

N-356A

N-354 A/1

N-354 B/1

H-333A/4

V-328

N-329B/1

N-329A/1

N-359

N-359B

V-360/1

M-390A

M-390B

V-329

solvent recovery system

H-309 C

H-309 B

H-329

Appendix E Selection modeled sections


Assetnr.
trans/weigh
F-3107A
F-3107C
F-3107B
M-3107A
M-3107B
M-3107C
F-306A
F-306B
M-309A
M-307B
M-309B
V-354
V-354B
V-352B
formulering
V-312
M-312
V-369
M-369
reactie
N-316A
N-316B
N-316C
N-316D
V-313A
M-313A
N-317A
H-302A
H-303A
V-313B
M313B
N-317B
H-302B
H-303B
V-370
M-370
N-317C
H-335
H-336
V-314A/1
M-314A/1
V-314B/1

WO's EMPAC

Output dents

+
+
+
+
+
+
+
+
+

+
+
+

+
+
-

+
+
+
+
+
+
+

+
+

+
+
+
+
+

106

M-314B/1
V-371A
M-371A
V-371B
M-371B
extractie
N-318A
N-318B
F-317A
F-317C
F-317E
H-307A
F-315A
F-315B
F-315C
F-315E
F-315D
F-315F
V-325A
N-327A
N-327B
N-327E
N-358A
N-358B
F-317B
F-317D
F-317F
H-307B
F-315G
F-315H
F-315J
F-315L
F-315K
F-315M
V-325B
M-325B
N-327C1
N-327D1
concentratie
H-310B1/1
H-310B2/1
N-335B
N-335C
V-332B
H-310C1
H-310C2
N-335D
N-335E

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-

+
+
+
+
+
+

+
+
+
+
+
+

+
+

+
+

107

H-312C
V-332C
V-333B
H-310A1/2
H-310A2/3
N-335A
H-312A
H-312D
V-332/1
V-333A
precipitatie
N-336D
N-336E
N-336F/1
H-3063B
H-3063C
M-316B1
M-316B2
M-316C1
M-316C2
M-318B1
M-318B2
F-323B
M-321-B/1
M-318C1
M-318C2
F-323C
M-321-C/1
N-336A
N-336B/1
N-336C/1
H-3063A
M-316A1
M-316A2
M-316A3
M-318A1
M-318A2
F-323A/1
M-321-A/1
drogen
M-320-B
V-342B
M-322-B
M-384B
M-320-C
V-342C
M-322-C
M-384C

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

+
+

+
+
+

+
+
+
+

+
+
+
+
+
+
+
+

108

F-327B/1
V-340B/1
M-331B/1
V-385/1
V-388
M-380/2
F-373/1
K-313A/1
K-313B/1
M-320-A/1
V-342A
M-322-A/1
M-384A/1
F-326A/1
M-331/1
V-341/1
M-341/2
M-374/1
F-333
K-308A/1
K-308B/1
air blower syst
F-3103C
K-311C
F-342B
H-3062B
F-3103D
F-3103E
K-311D
K-311E
F-342C
H-3062C
F-3103B
K-311A
K-311B
F-342A
H-3062A
drying air syst
F-328A
F-328B
K-305A/1
F-329
N-340A
N-340B
H-318
K-306A/1
H-319A
F-328C

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

+
+
+
+

+
+

+
+

+
+

+
+
+
+
+
+
+
-

109

F-328D
K-305B/1
F-329B
N-340C
N-340D
H-318B
K-306B/1
H-319B
vent gas syst
N-349A
N-349B
V-324
N-357A
N-357B
V-368
V-315/1
V-317
H-340/1
K-303A
K-303B
V-319A/2
V-319B/1
H-305/1
V-320/1
N-346
V-321/1
K-314A
K-314B
V-381A
coolant water
V-349
N-343A
N-343B
N-3107A
N-3107B
Z-308A/1
Z-308B/1
solv/powder sep.
N-337A
N337B
N-337C
N-337D
N-337E
F-341A/1
F-341B/1
F-341C/1
H-316
H-316B

+
+
+
+
+
+
+

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

+
+

+
+
+
+
+
+
+
+
-

110

H-317
V-345B
N-339C
N-339D
N-338B
H-340/1
H-341/1
V-345A
N-339A
N-339B
N-338/1
V-367
V-309
solv.rec.syst.
V-346
F-380A
F-380B
N-354A/1
N-354B/1
N-355A/1
N-355B/1
V-326
N-356A
N-356B
H-333A/4
V-328
N-329A/1
N-329B/1
V-360/1
H-309B
H-309C
N-359A
N-359B
M-390A
M-390B
H-329
V-329

+
+
+
+
+
+
+
slecht
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
slecht
slecht
+
+
slecht

111

Appendix F Input functions simulation model


Asset

Failure distribution Repair time distribution

Transportation & Weiging


F-3107A Weibull
Normal
Beta:0.8475
Mean: 0.8426
Eta:214.0386
Std: 0.4495
Gamma:0.825
F-3107B Weibull
Lognormal
Beta:0.7098
Mean:-1.3461
Eta:127.093
Std:1.2257
F-3107C Weibull
Normal
Beta:0.833
Mean:0.8958
Eta:213.0194
Std:0.2018
Gamma:3.1875
M-3107A Weibull
Normal
Beta:1.2929
Mean:0.1563
Eta:104.3325
Std:0.0202
M-3107B Weibull
Normal
Beta:0.7758
Mean:0.1563
Eta:109.7437
Std:0.0202
M-3107C Weibull
Lognormal
Beta:1.0033
Mean:-1.2715
Eta:100.5923
Std:0.94
F-306-1 Exponential
Lognormal
Lambda:0.0071
Mean:-1.3286
Std:1.5219
F-306-B Exponential
Normal
Lambda:0.0071
Mean:0.4861
Std:0.4705
M-309
Weibull
None
Beta:0.8796
Eta:72.5232
M-307-B Weibull
Normal
Beta:0.9602
Mean:0.1571
Eta:97.3924
Std:0.0195
M-309B Lognormal
None
Mean:3.4069
Std:1.482
V-354
Weibull
Exponential
Beta:0.7109
Lambda:5.0646
Eta:165.558
Gamma:4
V-352B Weibull
Lognormal
Beta:0.893
Mean:-1.6524

112

V-354B

Eta:67.2743
Lognormal
Mean:2.4304
Std:1.3003

Std:0.2725
Exponential
Lambda:8.0124

Precipitation
N-336A Weibull
Exponential
Beta:0.8372
Lambda:3.9429
Eta:59.9555
N-336B Weibull
Normal
Beta:0.8832
Mean:0.4927
Eta:44.9848
Std:0.1866
N-336C Weibull
Normal
Beta:0.8599
Mean:0.4467
Eta:36.733
Std:0.2914
N-336D Weibull
Normal
Beta:0.7081
Mean:0.345
Eta:33.0142
Std:0.2445
N-336E Weibull
Normal
Beta:0.7998
Mean:0.3894
Eta:51.6608
Std:0.2393
N-336F Weibull
Normal
Beta:0.8645
Mean:0.373
Eta:40.2167
Std:0.2157
H-3063A Exponential
Lognormal
Lambda:0.0071
Mean:-2.8191
Std:0.6697
H-3063B Weibull
Lognormal
Beta:0.5534
Mean:-1.8636
Eta:180.7551
Std:0.7732
H-3063C Weibull
Lognormal
Beta:0.5534
Mean:-1.8636
Eta:180.7551
Std:0.7732
M-316A Weibull
Normal
Beta:0.8544
Mean:0.2778
Eta:146.6138
Std:0.0516
M-316B Weibull
Normal
Beta:0.7893
Mean:0.2778
Eta:57.3566
Std:0.0516
M-316C Weibull
Normal
Beta:0.7893
Mean:0.2778
Eta:57.3566
Std:0.0516
M-318A Weibull
Lognormal
Beta:0.7346
Mean:-1.7329
Eta:50.7667
Std:0.6364
M-318B Weibull
Lognormal
Beta:0.7346
Mean:-1.683

113

Eta:50.7667
M-318C1 Exponential
Lambda:0.0084
M-318C2 Weibull
Beta:0.947
Eta:268.1823
F-323A/1 Weibull
Beta:0.9224
Eta:172.7266
F-323B
Exponential
Lambda:0.0038
Gamma:45.8548
F-323C Weibull
Beta:0.9224
Eta:172.7266
M-321A/1 Exponential
Lambda:0.0195
M-321B/1 Exponential
Lambda:0.0221
M-321C/1 Weibull
Beta:0.7645
Eta:42.9548

Std:0.3578
Lognormal
Mean:-2.2101
Std:0.9204
Lognormal
Mean:-2.2101
Std:0.9204
Lognormal
Mean:-0.6931
Std:0.8265
Lognormal
Mean:-0.6931
Std:0.8265
Lognormal
Mean:-0.6931
Std:0.8265
Normal
Mean:0.4107
Std:0.1413
Normal
Mean:0.4107
Std:0.1413
Lognormal
Mean:-2.4849
Std:1.371

Drying L2
M-320A1 Lognormal
Normal
Mean:3.6902
Mean:0.0917
Std:1.4769
Std:0.0445
V-342A Weibull
Normal
Beta:0.7972
Mean:0.125
Eta:68.5461
Std:0.0324
M-322-A/1 Weibull
Lognormal
Beta:0.8633
Mean:-1.6695
Eta:39.3662
Std:0.2067
Gamma:0.2425
H-3062A Weibull
Exponential
Beta:0.6313
Lambda:6.2792
Eta:634.7064
M-384A/1 Weibull
Lognormal
Beta:0.9332
Mean:-1.431
Eta:36.9103
Std:0.5193
F-326A/1 Exponential
Fixed: 0.3333
Lambda:0.0157
Gamma:252.9797
M-331/1 Weibull
Normal

114

Beta:0.9224
Eta:186.9352
Gamma:63.8175
V-341/1 Weibull
Beta:0.5153
Eta:145.5177
Gamma:15.15
M-341/2 Weibull
Beta:0.5153
Eta:145.5177
Gamma:15.15
K-308A/1 Normal
Mean:182.1428
Std:129.4963
K-308B/1 Weibull
Beta:0.9775
Eta:192.9384
Gamma:36.26
F-333
Exponential
Lambda:0.0076

Mean:0.2250
Std.:0.0913
Lognormal
Mean:-2.2541
Std:0.4235
Normal
Mean:0.1515
Std:0.0675
Lognormal
Mean:-2.7237
Std:0.6471
Lognormal
Mean:-2.7237
Std:0.6471
Normal
Mean:0.0833
Std:0.08

Drying L1
Weibull
Normal
Beta:0.7225
Mean:0.1563
Eta:46.3603
Std:0.0193
M-320C Weibull
Normal
Beta:1.5043
Mean:0.1597
Eta:70.7792
Std:0.0132
V-342B Weibull
Lognormal
Beta:0.7972
Mean:-1.7918
Eta:68.5461
Std:0.6405
V-342C Weibull
Normal
Beta:0.9394
Mean:0.1966
Eta:98.2474
Std:0.0733
M-322B Weibull
Lognormal
Beta:0.8633
Mean:-1.4367
Eta:39.3662
Std:0.3468
Gamma:0.2425
M-322C Exponential
Normal
Lambda:0.0172
Mean:0.1458
Std.:0.0672
H-3062B Lognormal
Exponential
Mean:4.0859
Lambda:6.9844
Std:1.54
H-3062C Exponential
Exponential
Lambda:0.0019
Lambda:6.9844

M-320B

115

M-384B

M-384C

F-327B/1

V-340B/1

M-331B/1

V-385/1

K-313A/1

K-313B/1

F-373/1

Weibull
Beta:1.1563
Eta:32.2013
Weibull
Beta:0.7383
Eta:34.6465
Gamma:0.535
Weibull
Beta:1.1832
Eta:133.5133
Weibull
Beta:0.6014
Eta:130.3585
Gamma:1.4575
Weibull
Beta:0.7594
Eta:56.6242
Weibull
Beta:1.1891
Eta:204.1017
Weibull
Beta:0.659
Eta:89.8338
Gamma:5.23
Weibull
Beta:0.6268
Eta:62.9173
Weibull
Beta:0.433
Eta:71.5015
Gamma:88.75

Lognormal
Mean:-0.9939
Std:0.5242
Lognormal
Mean:-0.9940
Std:0.5243
Fixed:0.3333

Normal
Mean:0.2361
Std:0.178
Normal
Mean:0.3021
Std:0.0392
Lognormal
Mean:-2.2541
Std:0.4235
Lognormal
Mean:-2.7368
Std.:0.4501
Lognormal
Mean:-2.4846
Std:0.6114
Fixed:0.3333

116

Appendix G Model simulation output


Abbreviations used in tables with results of the model runs:
FCI
DECI
Mean Av.
Mean Av. (W/o PM & Insp.)
Expected NOF
SD Events
CM
PM

Failure Criticality Index


Down coming Event Criticality Index
Mean availability
Mean availability without preventive maintenance and
inspections
Expected number of failures
Shutdown events
Corrective Maintenance
Preventive Maintenance

Yearly Shutdown
Transportation Weighing section
System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9657
0
0.9998
1
0.094
2.32
729.31
1586.6069
0.3931
56
56.3931
2.312
2.312
4
6.312

117

Block Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

SB
M-309A
M-307B
M-309B
F-306A
F-306B
V-352B
V-354B
V-354
EB
BPAfilters
M-3107A
F-3107A
F-3107C
M-3107C
M-3107B
F-3107B
BPAfilters_S
witch

0.00%
7.41%
5.34%
7.16%
4.66%
3.45%
7.24%
2.93%
4.74%
0.00%
57.07%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%

1
0.964
0.9644
0.9636
0.9627
0.9626
0.9632
0.9652
0.9648
1
0.9657
0.9648
0.9626
0.9625
0.9627
0.9649
0.9628
1

0.00%
10.80%
8.37%
9.32%
5.64%
5.64%
11.15%
4.28%
4.40%
0.00%
40.40%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%

Mean
Av.
(w/o
PM &
Insp.)
1
0.998
0.9984
0.9976
0.9967
0.9964
0.9972
0.9993
0.9989
1
0.9998
0.9988
0.9967
0.9964
0.9967
0.999
0.9968
1

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0
21.342
16.312
24.942
10.624
11.17
22.75
9.34
9.42
0
1.324
12.104
6.576
6.68
12.328
10.94
9.47
0

0
21.342
16.312
24.942
10.624
11.17
22.75
9.34
9.42
0
0
12.104
6.576
6.68
12.328
10.94
9.47
0

0
0.682
0.528
0.588
0.356
0.356
0.704
0.27
0.278
0
2.55
0
0
0
0
0
0
0

119

0
59.1916
58.4566
59.8371
61.2118
61.4996
60.4385
57.1428
57.8109
0
56.2815
57.8059
61.4106
61.6385
61.2817
57.7172
61.0653
0

1643
1583.8084
1584.5434
1583.1629
1581.7882
1581.5004
1582.5615
1585.8572
1585.1891
1643
1586.7185
1585.1941
1581.5894
1581.3615
1581.7183
1585.2828
1581.9347
1643

0
3.3596
2.5686
3.9211
5.3518
5.8916
4.5225
1.1428
1.8669
0
0
1.8899
5.4666
5.9185
5.4777
1.7172
5.2613
0

0
3.988
3.992
3.994
3.99
3.972
3.994
4
3.996
0
4
3.994
3.996
3.98
3.986
4
3.986
0

0
55.832
55.888
55.916
55.86
55.608
55.916
56
55.944
0
56
55.916
55.944
55.72
55.804
56
55.804
0

Precipitation Section
System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of Inspections:
Number of PMs:
Total Events:

0.9501
0.0003
0.9844
0.986
0
103.188
13.4235
1561.087
25.6418
56.2711
81.913
102.996
102.996
0
4.054
107.05

120

Block
RS FCI RS
Name
DECI
(Diagram)

Mean
Av.
(All
Events)

N-336L1
N-336E
N-336F
N-336D
SB
H-3063B
H-3063C
H-3063A
M-316B2
M-316A3
M316B/1
M-318B1
M-318B2
M-318C1
M-318C2
M-318A1
M-318A2
F-323B
F-323C
F-323A
M-321B
M-321C
M-321A

0.11%
0.00%
0.00%
0.00%
0.00%
0.04%
0.06%
0.00%
24.80%
0.00%
24.62%

0.64%
0.00%
0.00%
0.00%
0.00%
0.09%
0.09%
0.11%
24.11%
0.05%
23.87%

0.9658
0.9632
0.9835
0.962
1
0.9648
0.9648
0.9655
0.9621
0.5387
0.9622

Mean
Av.
(w/o
PM &
Insp.)
0.9999
0.994
0.9974
0.9923
1
0.9988
0.9989
0.9995
0.9961
0.9988
0.9961

0.12%
0.01%
0.10%
0.04%
0.00%
0.00%
0.03%
0.08%
0.00%
0.16%
0.26%
0.00%

0.37%
0.03%
0.22%
0.10%
0.25%
0.03%
0.11%
0.16%
0.06%
0.45%
0.51%
0.26%

0.9627
0.9657
0.9646
0.9653
0.9624
0.9648
0.9632
0.9629
0.963
0.9577
0.9618
0.9586

0.9967
0.9997
0.9986
0.9994
0.9964
0.9989
0.9972
0.9969
0.9969
0.9915
0.9957
0.9925

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0.436
25.174
11.298
36.012
0
9.036
8.906
10.862
25.66
7.958
25.456

0.684
0
0
0
0
0.096
0.098
0.122
25.806
0.058
25.558

56.1675
60.4746
27.1405
62.5093
0
57.8473
57.7849
56.7655
62.2651
757.8484
62.1677

1586.8325
1582.5254
1615.8595
1580.4907
1643
1585.1527
1585.2151
1586.2345
1580.7349
885.1516
1580.8323

0
25.174
11.298
36.012
0
9.036
8.906
10.862
25.66
7.958
25.456

0
9.9066
4.2925
12.6973
0
1.9033
1.8409
0.8215
6.4051
1.997
6.3637

4
3.612
1.632
3.558
0
3.996
3.996
3.996
3.99
54.21
3.986

56
50.568
22.848
49.812
0
55.944
55.944
55.944
55.86
755.8513
55.804

27.7
2.074
13.04
6.046
27.724
8.496
9.006
9.18
9.046
34.116
33.154
30.072

0.392
0.03
0.236
0.106
0.268
0.036
0.12
0.174
0.068
0.48
0.544
0.282

61.3205
56.4208
58.1638
57
61.7997
57.7959
60.3993
60.9447
60.8292
69.4228
62.7975
67.9533

1581.6795
1586.5792
1584.8362
1586
1581.2003
1585.2041
1582.6007
1582.0553
1582.1708
1573.5772
1580.2025
1575.0467

27.7
2.074
13.04
6.046
27.724
8.496
9.006
9.18
9.046
34.116
33.154
30.072

5.4885
0.4208
2.2198
1
5.9957
1.7959
4.6513
5.1127
5.0532
14.0388
7.1335
12.2893

3.988
4
3.996
4
3.986
4
3.982
3.988
3.984
3.956
3.976
3.976

55.832
56
55.944
56
55.804
56
55.748
55.832
55.776
55.384
55.664
55.664

121

EB
M-316A2
M-316A1
M-316C2
M-316C1
N-336L2
N-336B
N-336A
N-336C

0.00%
0.00%
0.00%
24.63%
24.93%
0.00%
0.00%
0.00%
0.00%

0.00%
0.00%
0.00%
23.88%
24.22%
0.37%
0.00%
0.00%
0.00%

1
0.9643
0.9643
0.9623
0.9621
0.9657
0.9657
0.9658
0.9552

1
0.9984
0.9984
0.9961
0.9961
0.9998
0.9998
0.9999
0.9888

0
10.768
10.586
25.474
25.76
1.342
0.7
0.632
40.176

0
0.004
0.002
25.566
25.928
0.392
0
0
0

122

0
58.6313
58.6265
62.021
62.3385
56.2902
56.3514
56.2036
73.584

1643
1584.3687
1584.3735
1580.979
1580.6615
1586.7098
1586.6486
1586.7964
1569.416

0
10.768
10.586
25.474
25.76
0
0.7
0.632
40.176

0
2.6873
2.6545
6.329
6.4225
0
0.3514
0.2036
18.368

0
3.996
3.998
3.978
3.994
4
4
4
3.944

0
55.944
55.972
55.692
55.916
56
56
56
55.216

Drying Section Line 1


System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9426
0.0003
0.9766
0.986
0
190.678
13.9905
1548.6169
38.3864
55.9967
94.3831
188.418
188.418
4.01
192.428

123

Block
Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

SB
H-3062A
M-341/1
K-311AB
K-311B
K-311A
F-333
K308
K-308B
K-308A
F-326A/1
M384A
M-384A
BypassA
F-342A
M-331/1
V-341/1
F-3103B
EB
M-322A
V-342A
M-320A
K311AB_Sw

0.00%
1.91%
4.81%
0.38%
0.00%
0.00%
6.18%
0.04%
0.00%
0.00%
1.61%
34.38%
0.00%
0.00%
0.00%
2.69%
4.91%
0.00%
0.00%
20.96%
11.53%
10.61%
0.00%

1
0.9656
0.9651
0.9658
0.9661
0.9658
0.9654
0.9659
0.9683
0.9687
0.9653
0.9542
0.9557
0.7957
1
0.9652
0.9653
1
1
0.9614
0.9643
0.9648
1

0.00%
1.91%
4.79%
0.61%
0.00%
0.00%
5.36%
0.11%
0.00%
0.00%
1.65%
34.53%
0.00%
0.00%
0.00%
2.73%
4.93%
0.00%
0.00%
21.21%
11.67%
10.49%
0.00%

Mean
Av.
(w/o
PM &
Insp.)
1
0.9996
0.9992
0.9999
0.9995
0.9995
0.9994
0.9999
0.9998
0.9998
0.9994
0.988
0.9894
0.7957
1
0.9993
0.9993
1
1
0.9953
0.9983
0.9989
1

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0
3.694
9.208
2.32
13.086
12.608
11.842
1.064
3.828
4.302
3.066
73.182
40.75
0.372
0
5.138
9.386
0
0
40.266
22.124
20.346
0

0
3.694
9.208
0
13.086
12.608
11.842
0
3.828
4.302
3.066
0
40.75
0
0
5.138
9.386
0
0
40.266
22.124
20.346
0

0
3.678
9.222
1.176
0
0
10.314
0.22
0
0
3.172
66.452
0
0
0
5.246
9.49
0
0
40.814
22.458
20.186
0

0
56.5217
57.3344
56.1357
55.6812
56.2071
56.9121
56.0839
52.0251
51.3682
56.9659
75.218
72.7972
335.5836
0
57.16
57.0854
0
0
63.423
58.6575
57.8467
0

124

1643
1586.4783
1585.6656
1586.8643
1587.3188
1586.7929
1586.0879
1586.9161
1590.9749
1591.6318
1586.0341
1567.782
1570.2028
1307.4164
1643
1585.84
1585.9146
1643
1643
1579.577
1584.3425
1585.1533
1643

0
0.5777
1.3904
0
0.8012
0.7671
1.0521
0
0.3091
0.3522
1.0219
0
17.3572
0
0
1.16
1.0854
0
0
7.759
2.7695
1.8747
0

0
3.996
3.996
4
3.92
3.96
3.99
4
3.694
3.644
3.996
3.986
3.96
0
0
4
4
0
0
3.976
3.992
3.998
0

0
55.944
55.944
56
54.88
55.44
55.86
56
51.716
51.016
55.944
55.4438
55.44
0
0
56
56
0
0
55.664
55.888
55.972
0

itch
K308_Swit
ch
M384A_S
witch

0.00%

0.00%

1643

0.00%

0.00%

1643

125

Drying section Line 2


System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9592
0.0001
0.9933
0.994
0
52.47
39.4437
1576.0271
10.9756
55.9973
66.9729
51.24
51.24
4
55.24

126

Block Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

SB
H-3062B
H-3062C
M384C
bypassC
M-384C
K-313
K-313B/1
K-313A/1
M384B
M-384B
BypassB
F-373/1
F-327B/1
M-322B
V-340B/1
M-331B/1
V-385/1
V-338
EB
F-342B
K-311C
F-3103C

0.00%
0.64%
0.11%
2.10%
0.00%
0.00%
0.42%
0.00%
0.00%
1.71%
0.00%
0.00%
13.78%
24.19%
1.63%
23.12%
10.31%
14.39%
0.00%
0.00%
0.00%
0.74%
0.00%

1
0.9646
0.9657
0.9543
1
0.9561
0.9659
0.9652
0.9655
0.9863
0.9557
1
0.9656
0.9634
0.9608
0.9642
0.965
0.9654
1
1
1
0.9644
1

0.00%
0.81%
0.12%
2.87%
0.00%
0.00%
0.80%
0.00%
0.00%
1.64%
0.00%
0.00%
13.02%
23.35%
2.56%
20.21%
9.94%
13.85%
0.00%
0.00%
0.00%
1.15%
0.00%

Mean
Av.
(w/o
PM &
Insp.)
1
0.9987
0.9997
0.9881
1
0.9899
1
0.9993
0.9995
0.9869
0.9893
1
0.9996
0.9974
0.9947
0.9982
0.999
0.9995
1
1
1
0.9984
1

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0
15.05
2.938
78.04
0
39.034
0.218
11.91
11.302
86.354
41.232
0
7.228
12.714
44.722
12.186
5.416
7.556
0
0
0
20.386
0

0
15.05
2.938
0
0
39.034
0
11.91
11.302
0
41.232
0
7.228
12.714
44.722
12.186
5.416
7.556
0
0
0
20.386
0

0
0.448
0.064
1.586
0
0
0.442
0
0
0.904
0
0
7.19
12.896
1.412
11.162
5.49
7.648
0
0
0
0.636
0

127

0
58.1156
56.4255
75.1162
0
72.1839
56.0089
57.1931
56.7579
22.575
72.7566
0
56.5723
60.0971
64.3971
58.8592
57.5045
56.8246
0
0
0
58.5085
0

1643
1584.8844
1586.5745
1567.8838
1643
1570.8161
1586.9911
1585.8069
1586.2421
1620.425
1570.2434
1643
1586.4277
1582.9029
1578.6029
1584.1408
1585.4955
1586.1754
1643
1643
1643
1584.4915
1643

0
2.1436
0.4255
0
0
16.5759
0
1.1931
0.8139
0
17.5126
0
0.6003
4.2371
8.7891
2.9432
1.6445
0.8806
0
0
0
2.6205
0

0
3.998
4
3.982
0
3.972
4
4
3.996
3.946
3.946
0
3.998
3.99
3.972
3.994
3.99
3.996
0
0
0
3.992
0

0
55.972
56
55.6062
0
55.608
56
56
55.944
0.9865
55.244
0
55.972
55.86
55.608
55.916
55.86
55.944
0
0
0
55.888
0

V-342B
M-320B
K-311DE
K-311E
F-3103E
K-311D
F-3103D
F-342C
M-322C
V-342C
M-320C
M384C_Switch
K-313_Switch
M384B_Switch
K311DE_Switch

0.49%
1.25%
1.63%
0.00%
0.00%
0.00%
0.00%
0.00%
1.52%
0.83%
1.14%
0.00%
0.00%
0.00%
0.00%

0.67%
1.91%
1.89%
0.00%
0.00%
0.00%
0.00%
0.00%
2.14%
1.24%
1.86%
0.00%
0.00%
0.00%
0.00%

0.9646
0.9628
0.9657
0.9742
0.8125
0.969
0.9197
1
0.9633
0.9638
0.9635
1
1
1
1

0.9986
0.9967
0.9998
0.9996
0.8125
0.9998
0.9197
1
0.9973
0.9979
0.9975
1
1
1
1

11.462
34.516
7.17
11.352
0.338
10.898
0.146
0
30.628
17.252
25.93
0
0
0
0

0.37
1.056
1.046
0
0
0
0
0
1.18
0.684
1.026
0
0
0
0

128

58.2171
61.175
56.3538
42.3787
307.9879
50.8927
131.8765
0
60.3404
59.4466
60.0303
0
0
0
0

1584.7829
1581.825
1586.6462
1600.6213
1335.0121
1592.1073
1511.1235
1643
1582.6596
1583.5534
1582.9697
1643
1643
1643
1643

11.462
34.516
0
11.352
0
10.898
0
0
30.628
17.252
25.93
0
0
0
0

2.3571
5.399
0
0.6867
0
0.4087
0
0
4.4804
3.5306
4.1423
0
0
0
0

3.99
3.984
4
2.978
0
3.606
0
0
3.99
3.994
3.992
0
0
0
0

55.86
55.776
56
41.692
0
50.484
0
0
55.86
55.916
55.888
0
0
0
0

4 yearly shutdown
Transportation weighing section
System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9828
0
0.9998
0.998
0.16
1.89
880.4566
1614.7033
0.2967
28
28.2967
1.872
1.872
1
2.872

129

Block Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

SB
M-309A
M-307B
M-309B
F-306A
F-306B
V-352B
V-354B
V-354
EB
BPAfilters
M-3107A
F-3107A
F-3107C
M-3107C
M-3107B
F-3107B
BPAfilters_Switch

0.00%
12.06%
5.82%
6.98%
7.30%
4.23%
8.36%
3.81%
4.13%
0.00%
47.30%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%

1
0.9809
0.9814
0.9809
0.9795
0.9795
0.9802
0.9824
0.982
1
0.9828
0.9818
0.9801
0.9799
0.9799
0.9821
0.9802
1

0.00%
13.02%
7.59%
7.52%
7.17%
4.18%
10.86%
3.62%
4.60%
0.00%
41.43%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%

Mean
Av.
(w/o
PM &
Insp.)
1
0.998
0.9984
0.9979
0.9965
0.9965
0.9973
0.9994
0.999
1
0.9999
0.9989
0.9971
0.9969
0.9969
0.9991
0.9972
1

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0
21.154
16.342
21.814
11.294
10.934
22.68
8.566
8.416
0
0.894
11.852
5.616
5.584
11.664
9.486
8.194
0

0
21.154
16.342
21.814
11.294
10.934
22.68
8.566
8.416
0
0
11.852
5.616
5.584
11.664
9.486
8.194
0

0
0.374
0.218
0.216
0.206
0.12
0.312
0.104
0.132
0
1.19
0
0
0
0
0
0
0

131

0
31.3217
30.516
31.4194
33.696
33.6991
32.5175
28.9962
29.6506
0
28.1887
29.861
32.7659
33.0307
32.9882
29.4268
32.5457
0

1643
1611.6783
1612.484
1611.5806
1609.304
1609.3009
1610.4825
1614.0038
1613.3494
1643
1614.8113
1613.139
1610.2341
1609.9693
1610.0118
1613.5732
1610.4543
1643

0
3.3217
2.572
3.4194
5.808
5.7551
4.5175
1.0522
1.6506
0
0
1.861
4.7659
5.0307
5.0442
1.4828
4.5457
0

0
1
0.998
1
0.996
0.998
1
0.998
1
0
1
1
1
1
0.998
0.998
1
0

0
28
27.944
28
27.888
27.944
28
27.944
28
0
28
28
28
28
27.944
27.944
28
0

Precipitation section
System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9679
0.0002
0.9849
0.98
0
100.244
13.2884
1590.1871
24.7903
28.0226
52.8129
100.068
100.068
1.014
101.082

132

Block Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

N-336L1
N-336E
N-336F
N-336D
SB
H-3063B
H-3063C
H-3063A
M-316B2
M-316A3
M-316B/1
M-318B1
M-318B2
M-318C1
M-318C2
M-318A1
M-318A2
F-323B
F-323C
F-323A
M-321B
M-321C
M-321A
EB

0.03%
0.00%
0.00%
0.00%
0.00%
0.06%
0.09%
0.00%
24.98%
0.00%
24.58%
0.16%
0.02%
0.14%
0.07%
0.00%
0.00%
0.04%
0.06%
0.00%
0.24%
0.30%
0.00%
0.00%

0.9829
0.9797
0.9923
0.9781
1
0.9821
0.9821
0.9824
0.9792
0.3931
0.9793
0.9798
0.9826
0.9816
0.9823
0.9795
0.9821
0.9802
0.9803
0.9801
0.9745
0.979
0.9754
1

0.20%
0.00%
0.00%
0.00%
0.00%
0.07%
0.10%
0.03%
24.77%
0.01%
24.38%
0.21%
0.03%
0.18%
0.09%
0.06%
0.02%
0.05%
0.07%
0.02%
0.32%
0.37%
0.08%
0.00%

Mean
Av.
(w/o
PM &
Insp.)
1
0.9941
0.9983
0.9924
1
0.9991
0.9991
0.9995
0.9962
0.9991
0.9963
0.9968
0.9997
0.9986
0.9993
0.9964
0.9991
0.9972
0.9972
0.997
0.9914
0.9958
0.9923
1

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0.07
24.746
7.278
35.394
0
6.8
6.684
11.422
25.076
5.802
24.686
26.218
2.74
13.532
6.186
27.336
6.8
8.098
8.31
8.916
34.49
32.194
30.736
0

0
24.746
7.278
35.394
0
6.8
6.684
11.422
25.076
5.802
24.686
26.218
2.74
13.532
6.186
27.336
6.8
8.098
8.31
8.916
34.49
32.194
30.736
0

0.202
0
0
0
0
0.068
0.104
0.028
25.038
0.014
24.644
0.21
0.032
0.182
0.092
0.06
0.016
0.05
0.072
0.022
0.328
0.374
0.076
0

133

28.0264
33.3926
12.7293
35.9636
0
29.4343
29.427
28.8537
34.1238
997.1438
33.9583
33.1334
28.537
30.2448
29.07
33.7
29.4359
32.5727
32.3846
32.6523
41.9031
34.4892
40.4147
0

1614.9736
1609.6074
1630.2707
1607.0364
1643
1613.5657
1613.573
1614.1463
1608.8762
645.8562
1609.0417
1609.8666
1614.463
1612.7552
1613.93
1609.3
1613.5641
1610.4273
1610.6154
1610.3477
1601.0969
1608.5108
1602.5853
1643

0
9.7606
2.7613
12.4996
0
1.4343
1.427
0.8537
6.2358
1.4495
6.1263
5.1894
0.537
2.3008
1.07
5.868
1.4919
4.6287
4.5526
4.9323
14.1831
6.8252
12.6387
0

1
0.844
0.356
0.838
0
1
1
1
0.996
35.894
0.994
0.998
1
0.998
1
0.994
0.998
0.998
0.994
0.99
0.99
0.988
0.992
0

28
23.632
9.968
23.464
0
28
28
28
27.888
995.6944
27.832
27.944
28
27.944
28
27.832
27.944
27.944
27.832
27.72
27.72
27.664
27.776
0

M-316A2
M-316A1
M-316C2
M-316C1
N-336L2
N-336B
N-336A
N-336C
N336L1_Switch
N336L2_Switch

0.00%
0.00%
24.80%
24.42%
0.00%
0.00%
0.00%
0.00%
0.00%

0.00%
0.00%
24.62%
24.24%
0.08%
0.00%
0.00%
0.00%
0.00%

0.9814
0.9815
0.9792
0.9792
0.9828
0.9828
0.9828
0.9725
1

0.9984
0.9985
0.9962
0.9963
0.9998
0.9998
0.9999
0.989
1

10.306
9.988
24.92
24.544
1.22
0.604
0.608
39.872
0

0
0
24.886
24.506
0.078
0
0
0
0

30.565
30.4726
34.1883
34.1262
28.2595
28.2917
28.1857
45.2303
0

1612.435
1612.5274
1608.8117
1608.8738
1614.7405
1614.7083
1614.8143
1597.7697
1643

10.306
9.988
24.92
24.544
0
0.604
0.608
39.872
0

2.565
2.4726
6.1883
6.1262
0
0.2917
0.1857
18.0143
0

1
1
1
1
1
1
1
0.972
0

28
28
28
28
28
28
28
27.216
0

0.00%

0.00%

1643

134

Drying section line 1


System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9596
0.0002
0.9767
0.984
0
185.628
14.2877
1576.6757
38.3249
27.9994
66.3243
183.41
183.41
1.002
184.412

135

Block Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

SB
H-3062A
M-341/1
K-311AB
K-311B
K-311A
F-333
K308
K-308B
K-308A
F-326A/1
M384A
M-384A
BypassA
F-342A
M-331/1
V-341/1
F-3103B
EB
M-322A
V-342A
M-320A
K311AB_Switch

0.00%
1.35%
3.94%
0.13%
0.00%
0.00%
6.66%
0.01%
0.00%
0.00%
2.18%
36.18%
0.00%
0.00%
0.00%
3.12%
4.06%
0.00%
0.00%
21.94%
11.49%
8.94%
0.00%

1
0.9827
0.9823
0.9829
0.9842
0.9835
0.9823
0.983
0.9873
0.9878
0.9822
0.9709
0.9724
0.8116
1
0.9822
0.9824
1
1
0.9783
0.9813
0.982
1

0.00%
1.37%
3.94%
0.20%
0.00%
0.00%
5.74%
0.03%
0.00%
0.00%
2.21%
36.54%
0.00%
0.00%
0.00%
3.13%
4.10%
0.00%
0.00%
22.20%
11.63%
8.89%
0.00%

Mean
Av.
(w/o
PM &
Insp.)
1
0.9998
0.9993
1
0.9996
0.9996
0.9993
1
0.9998
0.9998
0.9992
0.9877
0.9893
0.8116
1
0.9992
0.9995
1
1
0.9952
0.9984
0.9991
1

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0
2.552
7.35
0.558
11.574
11.768
12.406
0.092
3.11
3.624
4.05
75.028
41.424
0.358
0
5.8
7.564
0
0
40.96
21.506
16.712
0

0
2.552
7.35
0
11.574
11.768
12.406
0
3.11
3.624
4.05
0
41.424
0
0
5.8
7.564
0
0
40.96
21.506
16.712
0

0
2.522
7.27
0.376
0
0
10.594
0.058
0
0
4.082
67.386
0
0
0
5.778
7.564
0
0
40.938
21.446
16.398
0

137

0
28.3938
29.1221
28.0314
25.9751
27.0928
29.1208
28.0071
20.8005
20.0601
29.2939
47.8638
45.3545
309.5137
0
29.2569
28.8663
0
0
35.7263
30.6895
29.5307
0

1643
1614.6062
1613.8779
1614.9686
1617.0249
1615.9072
1613.8792
1614.9929
1622.1995
1622.9399
1613.7061
1595.1362
1597.6455
1333.4863
1643
1613.7431
1614.1337
1643
1643
1607.2737
1612.3105
1613.4693
1643

0
0.3938
1.1221
0
0.7191
0.7168
1.1208
0
0.2485
0.2921
1.3499
0
17.6345
0
0
1.3129
0.8663
0
0
7.8943
2.6895
1.5307
0

0
1
1
1
0.902
0.942
1
1
0.734
0.706
0.998
0.998
0.99
0
0
0.998
1
0
0
0.994
1
1
0

0
28
28
28
25.256
26.376
28
28
20.552
19.768
27.944
27.7215
27.72
0
0
27.944
28
0
0
27.832
28
28
0

K308_Switch
0.00%
M384A_Switch 0.00%

0.00%
0.00%

1
1

1
1

0
0

0
0

0
0

138

1643
1643

0
0

0
0

0
0

0
0

Drying section line 2


System Overview
General
Mean Availability (All Events):
Std Deviation:
Mean Availability (w/o PM & Inspection):
Point Availability (All Events) at 1643:
Reliability at 1643:
Expected Number of Failures:
MTTFF:
System Uptime/Downtime
Uptime:
CM Downtime:
PM Downtime:
Total Downtime:
System Downing Events
Number of Failures:
Number of CMs:
Number of PMs:
Total Events:

0.9764
0.0001
0.9934
0.99
0
50.592
39.478
1604.1958
10.8043
27.9998
38.8042
49.548
49.548
1
50.548

139

Block Name
(Diagram)

RS FCI RS
DECI

Mean
Av.
(All
Events)

SB
H-3062B
H-3062C
M384C
bypassC
M-384C
K-313
K-313B/1
K-313A/1
M384B
M-384B
BypassB
F-373/1
F-327B/1
M-322B
V-340B/1
M-331B/1
V-385/1
V-338
EB
F-342B
K-311C
F-3103C
V-342B

0.00%
0.54%
0.19%
1.93%
0.00%
0.00%
0.21%
0.00%
0.00%
1.56%
0.00%
0.00%
13.33%
26.23%
1.83%
19.83%
11.81%
16.10%
0.00%
0.00%
0.00%
0.83%
0.00%
0.35%

1
0.9819
0.9827
0.9716
1
0.9734
0.983
0.9824
0.9826
0.9871
0.9724
1
0.9826
0.9803
0.9777
0.9814
0.9819
0.9824
1
1
1
0.9815
1
0.9819

0.00%
0.61%
0.19%
2.18%
0.00%
0.00%
0.32%
0.00%
0.00%
1.57%
0.00%
0.00%
13.17%
26.34%
2.17%
18.10%
11.86%
16.16%
0.00%
0.00%
0.00%
0.96%
0.00%
0.38%

Mean
Av.
(w/o
PM &
Insp.)
1
0.999
0.9997
0.9883
1
0.9901
1
0.9994
0.9996
0.9872
0.9893
1
0.9997
0.9973
0.9946
0.9985
0.9989
0.9994
1
1
1
0.9985
1
0.9989

Individual Block Summary


Expected SD
Block
Block
NOF
Events Downtime Uptime

Number CM
Number PM
of CMs Downtime of PMs Downtime

0
11.866
3.102
76.662
0
38.334
0.108
9.386
9.284
83.872
41.454
0
6.748
13.286
45.104
10.106
5.976
8.152
0
0
0
19.59
0
8.762

0
11.866
3.102
0
0
38.334
0
9.386
9.284
0
41.454
0
6.748
13.286
45.104
10.106
5.976
8.152
0
0
0
19.59
0
8.762

0
0.306
0.098
1.1
0
0
0.162
0
0
0.794
0
0
6.658
13.316
1.098
9.148
5.994
8.168
0
0
0
0.484
0
0.194

140

0
29.7036
28.4449
46.6055
0
43.664
28.0046
28.9589
28.6608
21.2155
45.4261
0
28.5642
32.3719
36.7151
30.487
29.7505
28.9334
0
0
0
30.3689
0
29.7772

1643
1613.2964
1614.5551
1596.3945
1643
1599.336
1614.9954
1614.0411
1614.3392
1621.7845
1597.5739
1643
1614.4358
1610.6281
1606.2849
1612.513
1613.2495
1614.0666
1643
1643
1643
1612.6311
1643
1613.2228

0
1.7036
0.4449
0
0
16.224
0
0.9589
0.6608
0
17.6501
0
0.5642
4.4279
8.8831
2.487
1.8065
0.9334
0
0
0
2.4809
0
1.7772

0
1
1
0.982
0
0.98
1
1
1
0.99
0.992
0
1
0.998
0.994
1
0.998
1
0
0
0
0.996
0
1

0
28
28
27.44
0
27.44
28
28
28
0.2475
27.776
0
28
27.944
27.832
28
27.944
28
0
0
0
27.888
0
28

M-320B
K-311DE
K-311E
F-3103E
K-311D
F-3103D
F-342C
M-322C
V-342C
M-320C
M384C_Switch
K-313_Switch
M384B_Switch
K311DE_Switch

1.21%
0.44%
0.00%
0.00%
0.00%
0.00%
0.00%
1.48%
0.83%
1.29%
0.00%
0.00%
0.00%
0.00%

1.44%
0.54%
0.00%
0.00%
0.00%
0.00%
0.00%
1.64%
0.91%
1.47%
0.00%
0.00%
0.00%
0.00%

0.98
0.9828
0.9898
0.7902
0.9855
0.909
1
0.9802
0.9809
0.9804
1
1
1
1

0.9969
0.9998
0.9997
0.7902
0.9998
0.909
1
0.9972
0.9978
0.9973
1
1
1
1

32.586
5.684
9.124
0.378
9.93
0.152
0
31.234
17.264
27.346
0
0
0
0

0.726
0.272
0
0
0
0
0
0.828
0.46
0.742
0
0
0
0

141

32.9209
28.2759
16.8004
344.6324
23.8968
149.5742
0
32.4504
31.436
32.1933
0
0
0
0

1610.0791
1614.7241
1626.1996
1298.3676
1619.1032
1493.4258
1643
1610.5496
1611.564
1610.8067
1643
1643
1643
1643

32.586
0
9.124
0
9.93
0
0
31.234
17.264
27.346
0
0
0
0

5.0889
0
0.5604
0
0.3768
0
0
4.5624
3.548
4.3613
0
0
0
0

0.994
1
0.58
0
0.84
0
0
0.996
0.996
0.994
0
0
0
0

27.832
28
16.24
0
23.52
0
0
27.888
27.888
27.832
0
0
0
0

Appendix H Output dents 2002


Asset
K311
V313
K313
M384A
M316A
M384C
F323A
V352B
Fenol
F329B
H302
F315 algemeen
M384B
M316C
M316B
Tailgasleiding
M313B
F323C
F3107
F315M
H310
Z308AB
M321B
V3118AB
M321C
PCP
N336
F306AB
V370
F315L
M341
K308
V342A
H333
V312

Tons
1931
858
681
667
565
564
477
462
459
455
439
407
404
343
332
324
312
273
261
249
232
229
223
222
190
167
165
163
138
136.1
118
115
113
112
100

142

Resin deuken pareto 2002


2500
2000
1500
1000

143

F329B

Fenol

V352B

F323A

M316A

M384A

K313

V313

K311

M384C

500

You might also like