Handbook of Reliability, Availability, Maintainability and Safety in Engineering Design - Part 5 potx

20 1 Design Integrity Methodology
Designing for maintainability, as it is applied to an item of equipment, includes the
aspects of testability, repairability and inter-changeabilityof an assembly’s inherent
components. In general, the concept of designing for maintainability is concerned
with the restoration of equipment that has failed to perform over a period of time.
The performance variable used in the determination of maintainability that is con-
cerned with the measure of time subject to equipment failure is the mean time to
repair (MTTR).
Thus, besides providing for visibility, accessibility, testability, repairability and
inter-changeability, designing for maintainability also incorporates an assessment
of expected performance in terms of the measure of MTTR in relation to the per-
formance capabilities of the equipment. Designing for maintain a bility during the
preliminary design phase would be to minimise the MTTR of a system by ensuring
that failure of an inherent assembly to perform a specific duty can be restored to its
expected performance over a period oftime. Similarly, designing for maintainability
during the detail design phase would be to minimise the MTTR of an assembly by
ensuring that failure of an inherent component to perform a specific function can be
restored to its expected initial state over a period of time.
d) Designing for Safety
Traditionally, assessments of the risk of failure are made on the basis of allow-
able factors of safety obtained from previous failure experiences, or from empirical
knowledge of similar systems operating in similar anticipated environments. Con-
ventionally, the factor of safety has been calculated as the ratio of what are assumed
to be nominal values of demand and capacity. In this context, demand is the resul-
tant of many uncertain variables of the system under consideration, such as loading
stress, pressures and temperatures. Similarly, capacity depends on the properties of
materials strength, physical dimensions, constructability,etc. The nominal values of
both demand and capacity cannot be determined with certainty and, hence, their ra-
tio, giving the conventional factor of safety, is a random variable. Representation of
the values of demand and capacity would thus be in the form of probability distribu-
tions whereby, if maximum demand exceeded minimum capacity, the distributions

would overlap with a non-zero probability of failure.
A convenient way of assessing this probability of failure is to consider the differ-
ence between the demand and capacity functions, termed the safety margin,aran-
dom variable with its own probability distribution. Designing for safety, or the mea-
sure of adequacy of a design,whereinadequacy is indicated by the measure of the
probability of failure, is associated with the determination of a reliability index for
items at the equipment and component levels. The reliability index is defined as the
number of standard deviations between the mean value of the probab ility d istribu-
tion of the safety ma rgin, where the safety margin is zero. It is the reciprocal of the
coefficient of variation of the safety margin.
Designing for safety furthermore includes analytic techniques such as genetic al-
gorithms and/or artificial neural networks (ANN) to perform multi-objective op timi-
1.2 Artificial Intelligence in Design 21
sations of engineering design problems. The use of genetic algorithms in designing
for safety is a new approach in determining solutions to the redundancy allocation
problem for series-parallel systems design comprising multiple components. Artifi-
cial neural networks in designing for safety offer feasible solutions to many design
problems because of their capability to simultaneously relate multiple quantitative
and qualitative variables, as well as to form models based solely on minimal data.
1.2 Artificial Intelligence in Design
Analysis of Target Engineering Design Projects
A stringent approach of objectivity is essential in implementing the theory of design
integrity in any target engineering design project, particularly with regard to the
numerous applications of mathematical models in intelligent computer automated
methodology. Selection of target engineering projects was therefore based upon il-
lustrating the development of mathematical and simulation models of process and
equipment functionality, and development of an artificial intelligence-based (AIB)
blackboard model to d etermine the integrity of process engineering design.
As a result, three different target engineering design projects are selected that
relate directly to the progressive stages in the developmen t of the theory, and to the

levels of modelling sophistication in the practical application of the theory:
• RAMS analysis model (product assurance) for an engineering design project
of an environmental plant for the recovery of sulphur dioxide emissions from
a metal smelter to produce sulphuric acid as a by-product. The purpose of im-
plementing the RAMS analysis model in this target engineering design project
is to validate the developed theory of design integrity in designing for reliabil-
ity, availability, maintainability and safety, for eventual inclusion in intelligent
computer automated methodology using artificial intelligence-based (AIB) mod-
elling.
• OOP simulation model (process analysis)for an engineeringdesign super-project
of an alumina plant with establishment costs in excess of a billion dollars. The
purpose of implementing the object oriented programming (OOP) simulation
model in this target engineering design p roject was to evaluate the mathemati-
cal algorithms developed for assessing the reliability,availability, maintainability
and safety requirements of complex process systems, as well as for the complex
integration of process systems, for eventual inclusion in intelligent computer au-
tomated methodology using AIB modelling.
• AIB blackboard model (design review) for an engineeringdesign super-project of
a nickel-from-laterite processing plant with establishment costs in excess of two
billion dollars. The AIB blackboard model includes intelligent computer auto-
mated methodologyfor application of the developed theory and the mathematical
algorithms.
22 1 Design Integrity Methodology
1.2.1 Development of Models and AIB Methodology
Applied co mputer modelling inclu des up-to-date object oriented software program-
ming applications inco rporatin g integrated systems simulation modelling, and AIB
modelling including knowledge-based expert systems as well as blackboard mod-
elling. TheAIB modelling provides for automated continualdesign reviews through-
out the engineeringdesign p rocess on the basis of concurrentdesign in an integrated
collaborative engineering design environment. Engineering designs are composed

of highly integrated, tightly coupled components where interactions are essential to
the economic execution of the design.
Thus, concurrent, rather than sequential consideration of requirements such as
structural, thermal, hydraulic, manufacture, construction, operational and mainte-
nance constraints will inevitably result in superior designs. Creating concurrent de-
sign systems for engineering designers requires knowledge of downstream activi-
ties to be infused into the design process so that desig ns can be generated rapidly
and correctly. The design space can be viewed as a mu lti-dimensional space, in
which each dimen sion has a different life-cycle objective such as serviceability or
integrity.
An intelligent design system should aid the designer in understanding the in-
teractions and trade-offs among different and even conflicting requirements. The
intention of the AIB blackboard is to surround the designer with expert systems that
provide feedback on continual design reviews of the design as it evolves throughout
the engineering design process. These experts systems, termed perspectives,must
be able to generate information that becomes part of the design (e.g. mass-flow bal-
ances and flow stresses), and portions of the geometry (e.g. the shapes and dimen-
sions). The perspectives are not just a sophisticated toolbox for the designer; rather,
they are a group of advisors that interact with one another and with the designer, as
well as identify conflicting inputs in a collaborative design environment.Implemen-
tation by multidisciplinary remotely located groups of designers inputs design data
and schematics into the relevant perspectives or knowledge-based expert systems,
whereby each design solution is collaboratively evaluated for integrity. Engineering
design includes important characteristics that have to be considered when develop-
ing design models, such as:
• Design is an optimised search of a number of design alternatives.
• Previous designs are frequently used during the design process.
• Design is an increasingly distributed and collaborative activity.
Engineering design is a complex process that is often characterised as a top-down
search of the space of possible solutions, considered to be the general norm of

how the design process should proceed. This process ensures an optimal solution
and is usually the construct of the initial design specification. It therefore involves
maintaining numerous candidate solutions to specific design problems in parallel,
whereby designers need to be adept at gen erating and evaluating a range of candi-
date solutions.
1.2 Artificial Intelligence in Design 23
The term satisficing is used to describe how designers sometimes limit their
search of the design solution space, possibly in response to technology limitations,
or to reduce the time taken to reach a solution because of schedule or cost con-
straints. Designers may opportunistically deviate from an optimal strategy, espe-
cially in engineering d esign where, in many cases, the design may involve early
commitment to and refining of a sub-optimal solution. In such cases, it is clear that
satisficing is often advantageous due to potentially r educed costs or where a satis-
factory, rather than an optimal design is required. However, solving complex design
problems relies heavily on the designer’s knowledge, gained through experience, or
making use of previous design solutions.
The concept of reuse in design was traditionally limited to utilising personal ex-
perience, with reluctance to copy solutions of other designers. The modern trend in
engineering design is, however, towards more extensive design reuse in a collabo-
rative environment. New computing technology provides greater opportunities for
design reuse and satisficing to be applied, at least in part, as a collaborative, dis-
tributed activity. A large amount of current research is concerned with developing
tools and methodologies to support design teams separated by space and time to
work effectively in a collaborative design environment.
a) The RAMS Analysis Model
The RAMS analysis model incorporates all the essential preliminaries of systems
analysis to validate the developed theory for the determination o f the integrity of
engineering design. A layout of part of the RAMS analysis model of an environ-
mental plant is given in Fig. 1.1.
The RAMS analysis model includes systems breakdown structures, process func-

tion definition, determination of failure consequences on system performance, de-
termination of process criticality, equipment function s definition, determination of
failure effects on equipmen t functionality, failure modes effects and criticality anal-
ysis ( FMECA), and determination of equipment criticality.
b) The OOP Simulation Mode l
The OOP simu lation model incorporates all the essential preliminaries of process
analysis to initially determine process characteristics such as process throughput,
output, input and capacity. The application of the model is primarily to determine its
capability of accurately assessing the effect of complex integrations of systems, and
process output mass-flow balancing in preliminary engineering design of large inte-
grated processes. A layout of part of the OOP simulation model is given in Fig. 1.2.
24 1 Design Integrity Methodology
Fig. 1.1 Layout of the RAM analysis model
c) The AIB Blackboard Model
The AIB blackboard model consists of three fundamental stages of analysis for de-
termining the integrity of engineering design, specifically preliminary design pro-
cess analysis, detail design plant analysis and commissioning operations analysis.
The preliminary design process analysis incor porates the essential prelim inaries of
design review, such as process definition, performance assessment, process design
evaluation, systems definition, functions analysis, risk assessment and criticality
analysis, linked to an inter-disciplinary collaborative knowledge-based expert sys-
tem. Similarly, the detail design plant analysis incorporates the essential prelimi-
naries of design integrity such as FMEA and plant criticality an alysis. The applica-
tion of the model is fundamentally to establish automated continual design reviews
whereby the integrity of engineering design is determined concurrently throughout
the engineering design process. Figure 1.3 shows the selection screen of a multi-user
interface ‘blackboard’ in collaborative engineering design.
1.2 Artificial Intelligence in Design 25
Fig. 1.2 Layout of part of the OOP simulation model
1.2.2 Artificial Intelligence in Engineering Design

Implementation o f the various models covered in this handbook predominantly fo-
cuses on determining the applicability and benefit of automated continual design
reviews throughout the engineering design p rocess. This hinges, however, upon
a broader understanding of the principles and philosophy of the use of artificial
intelligence (A I) in engineering design, pa rticularly in which new AI modelling
techniques are applied, such as the inclusion of knowledge-based expert systems
in blackboard models. Although these modelling techniques are described in d etail
later in the handbook, it is essential at this stage to give a brief account of artificial
intelligence in engineering design.
The application of artificial intelligence (AI) in engineering design, through ar-
tificial intelligence-based ( AIB) computer modelling, enables decisions to be made
about acceptable d esign performance by considering the essential systems design
criteria, the functionality of each particular system, the effects and consequences of
potential and functional failure, as well as the complex integration of the systems as
a whole. It is unfortunate that the growing number of unfulfilled promises and ex-
pectations about the capabilities o f artificial intelligence seems to have damaged the
credibility of AI and eroded its true contributions and benefits. The early advances
26 1 Design Integrity Methodology
Fig. 1.3 Layout of the AIB blackboard model
of expert systems, which were based on more than 20 years of research, were over-
extrapolated by many researchers looking for a feasible solution to the complexity
of integrated systems design. Notwithstanding the problems of AI, recent artificial
intelligence research has produced a set of new techniques that can usefully be em-
ployed in determining the integrity of engineering design. This does not mean that
AI in itself is sufficient, or that AI is mutually exclusive of traditional engineering
design. In order to developa proper perspective on the relationship between AI tech-
nology an d engineering design , it is necessary to establish a framework that provides
the means by which AI techniques can be applied with conventional engineering de-
sign. Knowledge-based systems provide such a framework.
a) Knowledge-Based Systems

Knowledge engineering is a problem-solving strategy and an approach to program-
ming that characterises a problem principally by the type of knowledge involved.
At one end of the spectrum lies conventional engineering design technology
based on well-defined, algorithmic knowledge. At the other end of the spectrum lies
AI-related engineering design technology based o n ill-defined heuristic knowledge.
1.2 Artificial Intelligence in Design 27
Among the problems that are well suited for knowledge-based systems are design
problems, in particular engineering design. As engineering knowledge is heteroge-
neous in terms of the kinds of problems that it encompasses and the methods used
to solve these, the use of heterogeneous representations is necessary. Attempts to
characterise engineering knowledge have resulted in the following classification of
the properties that are essential in constructing a knowledge-based expert system:
• Knowledge representation,
• Problem-solving strategy, and
• Knowledge abstractions.
b) Engineering Design Expert Systems
The term ‘expert system’ refers to a computer program that is largely a collection of
heuristic rules (rules of thumb) and detailed domain facts that have proven useful in
solving the special problems of some or other technical field. Expert systems to date
are basically an outgrowth of artificial intelligence, a field that has for many years
been devoted to the study of problem-solving using heuristics, to the construction of
symbolic representations of knowledge, to the process of communicating in n atural
language and to learning from experience.
Expertise is often defined to be that body of knowledge that is acquired over
many years of experience with a certain class of problem. One of the hallmarks
of an expert system is that it is constructed from the interaction of two types of
disciplines: domain experts, or practicing experts in some technical domain, and
knowledge engineers,orAI specialists skilled in analysing processes and problem-
solving approaches, and encoding these in a computer system.
The best domain expert is one with years, even decades, of practical experience,

and the best expert system is one that hasbeen created through a close scrutiny of the
expert’s domain by a ‘knowledgeable’ knowledge engineer. However, the question
often asked is which kinds of problems are most amenable to this type of approach?
Inevitably,problemsrequiringknowledge-intensiveproblemsolving,where years
of accumulated experience produce good performance results, must be the most
suited to such an approach. Such domains have complex fact structures, with large
volumes of specific items of information, organised in particular ways. The domain
of engineering design is an excellent example of knowledge-intensiveproblem solv-
ing for which the application of expert systems in the design process is ideally
suited, even more so for determining the integrity of engineering design. Often,
though, there are no known algorithms for approaching these problems, and the do-
main may be poorly formalised. Strategies for approaching design problems may
be diverse and depend on particular details of a problem situation. Many aspects of
the situation need to be determined during problem so lving, usually selected from
a much larger set of possible needs of which some may be expensive to determine—
thus, the significance of a particular need must also be considered.
28 1 Design Integrity Methodology
c) Expert Systems in Engineering Design Project Management
The advantages of an expert system are significant enough to justify a major effort
to develop these. Decisions can be obtained more reliably and consistently, where an
explanation of the final answers becomes an important benefit. An expert system is
thus especially useful in a consultation mode of complex engineering designs where
obscure factors may be overlooked, and is therefore an ideal tool in engineering
design project management in which the following important areas of engineering
design may be impacted:
• Rapid checking of preliminary design concepts, allowing more alternatives to be
considered;
• Iteration over the design process to improve on previous attempts;
• Assistance with and automation of complex tasks and activities of the design
process where expertise is specialised and technical;

• Strategies for searching in the space of alternative designs, and monitoring of
progress towards the targets of the design process;
• Integration of a diverse set of tools, with expertise applied to the problem of
engineering design project planning and control;
• Integration of the various stages of an engineering design project, inclusive of
procurement/installation, construction/fabrication, and commissio ning/warranty
by having knowledge bases that can be distributed for wide access in a collabo-
rative design environment.
d) Research in Expert Systems for Engineering Design
Within the past several years, a number of tools have been developed that allow
a higher-levelapproach to building expert systems in general, although most still re-
quire some programming skill. A few provide an integrated k nowledge engineering
environment combining features of all of the available AI languages.
These languages (CLIPS, JESS, etc.) are suitable and efficient for use by AI pro-
fessionals. A number of others are very specialised to specific problem types, and
can be used without programmingto build up a knowledge base, including a number
of small tools that run on personal computers (EXSYS, CORVID, etc.). A common
term for the more powerful tools is shell, referring to their origins as specialised
expert systems of which the knowledge base has been removed, leaving only a shell
that can perform the essential functions of an expert system, such as
• an inference engine,
• a u ser interface, and
• a knowledge storage medium.
For engineering design applications, however, good expert system development
tools are still being conceptualised and experimen ted with. Some of the most recent
techniques in AI may become the basis for powerful design tools. Also, a number
of the elements of the design process fall into the diagnostic–selection category,and
1.2 Artificial Intelligence in Design 29
these can be tackled with existing expert system shells. Many expert systems are
now being developed along these limited lines. The development of a shell that has

the basic ingredients for assisting or actually doing design is still an open research
topic.
e) Blackboard Models
Early expert systems used rules as the basic data structure to address h euristic
knowledge. From the rule-based expert system, there has been a shift to a more
powerful architecture based on the notion of cooperating experts (termed black-
board models) that allows for the integration of algorithmic design approaches with
AI techniques. Blackboard models provide the means by which AI techniques can
be applied in determining the integrity of engineering designs.
Currently,oneof the main areas of developmentis to p rovide integrativemeans to
allow various design systems to communicate with each other both dynamically and
cooperatively while working on the same design problem from different viewpoints
(i.e. concurrent design). What this amounts to is having a diverse team of experts or
multidisciplinary groups of design engineers, available at all stages of a design, rep-
resented by their expert systems. This leads to a design process in which technical
expertise can be shared freely in the form of each group’s expert system (i.e. col-
laborative design). Such a design process allows various groups of design engineers
to work on parts of a design problem independently, using their own expert sys-
tems, and accessing the expert systems of other disciplinary groups at those stages
when group cooperation is required. This would allow one disciplinary group (i.e.
process/chemical engineering) to produce a design and obtain an evaluation of the
design from other disciplinary groups (i.e. mechanical/electrical engineering), with-
out involving the people concerned. Such a design process results in a much more
rapid consideration of major design alternatives, and thus improves the qu ality of
the result, the effectiveness of the design review process, and the integrity of the
final design.
AclassofAI tools constructed along these lines is the blackboard model,which
provides for integrated design data management, and for allowing various knowl-
edge sources to cooperate in data development, verification and validation, as well
as in information sharing (i.e. concurrent and collaborative design). The blackboard

model is a paradigm that allows for the flexible integration of modular portions of
design code into a single problem-solving environment. It is a general and simple
model that enables the representation of a variety of design disciplines. Given its
nature, it is prescribed for problem solving in knowledge-intensive domains that
use large amounts of diverse, error-full and incomplete knowledge, therefore requir-
ing multiple cooperation between knowledge sources in searching a large problem
space—which is typical of engineering designs. In termsof the type of problems that
it can solve, there is only one major assumption—that the problem-solving activity
generates a set of intermediate results that contribute to the final solution.