Anda di halaman 1dari 10

Computers in Industry 44 (2001) 1±10

The application of KBE techniques to the FE model


creation of an automotive body structure
Martyn Pinfold*, Craig Chapman
Advanced Technology Centre, Warwick Manufacturing Group, School of Engineering, University of Warwick, Coventry CV4 7AL, UK
Received 3 February 2000; received in revised form 22 May 2000; accepted 11 August 2000

Abstract

An integral part of the design route for an automotive body structure is the analysis of it using ®nite element (FE)
techniques. The creation of a model for the FE analysis of such a structure is a very time consuming process. However, the
geometry, FE mesh and boundary conditions for the analysis are generically the same each time no matter the size or type of
the vehicle. This paper describes the design analysis response tool (DART) created to investigate the application of knowledge
based engineering (KBE) techniques for the automation of the FE model creation process for such a structure. # 2001 Elsevier
Science B.V. All rights reserved.

Keywords: Finite element analysis; Mesh generation; Knowledge based engineering; Automotive body structure

1. Introduction automotive structure consisting of the underlying


body structure made up of the body panels, chassis
Most authors (e.g. [1±3]), agree that the most time and structural members. An integral part of the design
consuming part of undertaking a ®nite element (FE) route for such structures is the analysis of them using
analysis is the creation of the analysis model. This FE techniques. The creation of a model for the FE
usually involves ®rst the transfer of the model from a analysis of such structures is a very time consuming
computer aided design (CAD) system into the analysis process that can currently take up to 15 man weeks of
system and then the simpli®cation of the geometry to effort. However, the geometry, FE mesh and boundary
remove some of the minute detail created. As Jones conditions for the analysis are generically the same
et al. [3] state the geometric model created in a CAD each time no matter the size or type of the vehicle. The
system can be incoherent and ambiguous and ``may DART system uses knowledge based engineering
require signi®cant cleaning and modi®cation in pre- (KBE) techniques to automate the FE model creation
paration for meshing and analysis''. To avoid these process. A rule base concerning the creation of ®rst the
problems analysts sometimes create their own model geometry of the vehicle structure and then from this a
thus, duplicating data [3,4]. simpli®ed model for mesh generation, and then ®nally
This paper describes the design analysis response the generation of the FE mesh itself has been devel-
tool (DART) created to aid in the analysis of an oped as a part of the KBE program.
With the use of the DART system it has been
*
Corresponding author. demonstrated that the time taken to create the analysis
E-mail address: m.k.pinfold@warwick.ac.uk (M. Pinfold). model can potentially be reduced from weeks to

0166-3615/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 6 - 3 6 1 5 ( 0 0 ) 0 0 0 7 9 - 8
2 M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10

minutes. This would allow engineers to try more applications have been aimed at routine and variant
structural combinations in a greatly reduced period designs for which the initial cost and time investment
of time and allow them to respond much more quickly can be justi®ed.
to potential design changes imposed by other depart- The method of creating models is very similar in
ments. As the models are all created from a uni®ed most commercial KBE systems. The adaptive model-
model description duplication of data is eliminated. ling language (AML) from TechnoSoft [9] has been
used within this project and has the ability to create
objects and write rules from templates, see Fig. 1 for a
2. Knowledge based engineering (KBE) layout of the AML knowledge based engineering
environment.
Unlike traditional computer aided design, KBE
systems can capture the intent behind the product 2.1. KBE applied to structural analysis
design by representing the `why' and `how' of a
design, in addition to the `what' of a design [5,6]. The goal of structural analysis is to understand the
The geometric description is only one piece of infor- physical response of structural systems when sub-
mation about the total product model. The focus of jected to loads. The geometry used for the creation
KBE is to provide an informationally complete of the ®nite element mesh is, in general an approx-
description or representation of a design as well as imation of the model geometry. The pre-processing
access to external databases. The knowledge based for an analysis, which includes the simpli®cation of
engineering environment is therefore, a framework for the CAD geometry to make it suitable for analysis and
capturing and de®ning the process of design creation. the generation of the FE mesh is often the most time
The inputs to the KBE model contain geometric and consuming and costly part of undertaking an analysis.
non-geometric attributes which can include design Desaleux and Fouet [10] state that the costs for the
speci®cations, design practices, engineers expertise, creation of an FE mesh represent about 80% of the
material properties and the boundary conditions for total analysis cost. Most of this time is spent in
structural analysis. When given a set of inputs, the correcting CAD model geometry problems [11].
KBE model can use the knowledge and rules to create Whilst many FE pre-processors will automatically
an instance of a design (note that readers requiring create an FE mesh, such automatic creation still
more information regarding KBE should see, for requires data such as mesh density, selection of ele-
example [7,8]). This can then be modi®ed by changing ments, and the position and type of boundary condi-
the input parameters to produce very quickly a new tions to be applied. However, if the same generic
instance of the design which allows for a signi®cant problem is to be analysed over and over again time
increase in engineering productivity [5,6]. Hence can be saved in the creation of the mesh if the rules
reduction in design cycle time through using a KBE regarding it are created once and then referred to. The
system is achieved through the complete automation simpli®cation of the CAD geometry to facilitate mesh
of any tedious, time consuming, error prone and generation can be regarded in a similar manner and
repetitive tasks [5,6]. In addition, the model can be therefore, meets all the criteria for the application of
made to run external programs, such as ®nite element knowledge based engineering technology [12,13].
analysis [6]. A KBE system therefore, allows for the From the available literature (for example [1,12±
rapid evaluation of design alternatives and the inte- 18]) it can be seen that whilst there are many exam-
gration of islands of technology. ples, or frameworks, discussed for the linking of KBE
Development of a KBE application requires a lot of or expert systems to structural analysis, these are often
up front investment in time and effort to gather knowl- done in two different systems. The output from the
edge, rules, dependencies and constraints that are KBE system is a set of geometry, maybe with attached
applicable to a family of parts. However, once this parameters, which will guide the creation of an FE
information has been acquired and incorporated into mesh in another software package. In addition, it is
the model, it can then be used repeatedly to create new clear from the literature that the creation of the FE
designs [5,6]. Due to this initial investment, KBE mesh is not seen as a major part of work. More time
M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10 3

Fig. 1. The AML knowledge based engineering environment.

and effort would appear to be spent upon optimising the analysis requirements. At any point within the
the FE mesh or upon optimising the design. design session the engineer has the ability to simplify
What is described in this paper is an example where the design model automatically for input to analysis.
the geometry creation and FE mesh generation are The concept structure can be veri®ed at any point
both done in the same system, which then outputs the within the session with respect to manufacturability,
data in a format suitable for the analysis package. styling and packaging con¯icts. The knowledge base,
Whilst what is proposed here is essentially a much which contains the design rules, is used to construct
simpler problem to be addressed than those discussed the product within the imposed constraints.
in most of the literature it is felt that this area of linking The model created in DART breaks down into ®ve
KBE and FE should not be ignored if only for the main parts: the style, packaging, structural members,
potential time savings achievable for the FE mesh joints and panels. These parts can then be decomposed
generation. further, each stage of decomposition varying the
intelligence of the objects such as joints. For example,
primitive objects might only know how to draw
3. The DART system themselves, whereas higher level objects will use
the knowledge base to make speci®c decisions regard-
The DART application is designed to allow the ing their shape, relationships, manufacturing process
engineer to interact through a set of menus and and cost. The collection of the knowledge base for
graphical interface tools, to specify the requirements DART from the engineers was not necessarily
of the desired product to rapidly create an automotive straightforward as many different departments had
structure with panelling in a three dimensional solid/ to be consulted as the knowledge base was fragmented
surface representation. The model can be created in a and largely based on past experience. Thus, the knowl-
non-structured way with information, such as material edge required often existed as `rules of thumb' with
data, added at any point in the modelling session. the engineers not realising that this constituted the
Structure sub-parts such as beams, joints and panels design rules. In addition, in the case of the vehicle
can be modi®ed in real time without consideration of structure rules are not only required for mesh genera-
4 M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10

tion but also for the creation of the geometry. This is which forms the outer working constraint to which the
not easy for such a complex structure, and is a very structural package must adhere to. ``No-go'' areas are
large problem to undertake. Thus, the geometry cre- de®ned by the importing of packaging objects, for
ated for the body structure was that of a relatively example, mannequins and powertrain components.
simple space-frame type of construction and used to The input to the DART system could take many forms
demonstrate the feasibility of the approach and the use as follows.
of KBE technology for this type of application.
 Imported data via standard formats such as IGES.
The ®rst step in the creation of DART was to create
 The specification of structured point data sets,
a conceptual product model. The conceptual product
defining the vehicle outer surfaces.
model can be thought of as a schematic of the com-
 The import of models via native CAD database
pleted system. A particular design instance of a pro-
formats.
duct model is described by a product model tree. This
 The import of previously stored models retrieved
structure describes the hierarchical relationships
from the object database.
between the various components (parts and processes)
of the model. The model also determines the methods Alternatively, if a model was not available at this
of how the objects will obtain the correct information, level of concept design the system allows the engineer
in order to carry out their speci®c tasks and as to the to sketch his own space frame type of vehicle structure
information they will provide to other objects. This by creating the beams, joints and panels in the form of
hierarchical abstraction means the decomposition of solid models, without reference to an initial outer
information into levels of increasing detail [19]. The surface de®nition.
decisions made by the system in creating a model are
based on the user inputs and the knowledge bases. The 3.2. Structure development Ð members, panels and
system is extended by increasing the part library, and joints
by expanding the object class de®nitions. The stages
that the user goes through in the DART system are At this point the body structure in terms of the
described in the following sections. structural members, the panels and joints are created
taking into account the packaging `no-go' areas such
3.1. Detailed geometry creation as the engine and passenger compartment. Figs. 2 and
3 show the build up of frame structures using ``adap-
The initial input to the DART system is the vehicle tive beam'' objects and automatic ``cast node joint''
style in terms of a vehicle outer surface de®nition, objects for two different vehicle styles. Fig. 4 shows

Fig. 2. Frame structure around a vehicle style model.


M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10 5

Fig. 3. Frame structure around an alternative vehicle style model.

the DART system main menu form and Fig. 5 shows a


beam creation interface. Fig. 6 shows an example of Fig. 5. Beam creation interface.
the product model tree.
The user can edit, if necessary, any of the member be extracted. This model can be used to provide many
or panel details or positions. From a standard library of different types of information as non-geometric as
existing designs general stiffness targets for each well as geometric information can be contained within
member can be obtained based on the overall size the KBE database. The model could be used to provide
of vehicle being analysed. The DART system will manufacturing information or costing information for
automatically check the member geometries and example, and can be output through neutral ®le for-
materials against the required stiffness targets and mats to external systems. However, the aim of the
make alterations if necessary. This is a very quick DART system was to enable a detailed ®nite element
®rst shot analysis to ensure that the vehicle being analysis of the structure to be undertaken.
designed is realistic. In addition, if relevant manufac-
turing information is stored within the database, and 3.3. Simpli®ed geometry for analysis
the user selects a particular type of material and
manufacturing method, the system will highlight Upon receipt of a detailed CAD model of an auto-
any case where the rules are broken. For example, motive body structure in the analysis area, the analyst
if the radius of curvature is too small to allow the will ®rst seek to simplify that model. The analyst will
component to be manufactured using the speci®ed then start to create the FE mesh based mainly upon
method or material. past experience of how various components and fea-
The model created at this stage in the DART system tures should be represented in the FE model. Some
is a complete solid model from which the surfaces can items, such as the engine, may be represented as a

Fig. 4. Main DART menu.


6 M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10

of components and jointing methods are used. Thus,


the types of elements used, the element size, the
methods of modelling the joints etc. are the same
each time. The panels are represented as shell ele-
ments and thus, the solid model needs to be simpli®ed
such that a panel is represented by a surface for
example, the bottom surface. This, in some areas
could have meant that when the surface data for the
panels was extracted from the detailed solid model,
gaps between adjacent surfaces may have appeared, or
some surfaces may have overlapped. Thus, before any
subsequent creation of an FE mesh this surface repre-
sentation could need `mending'. If the methods of
doing this `mend' are known they can be stored as
rules within the KBE system and can be undertaken
automatically. However, in this case there was no need
to `mend' the surfaces as with the space frame type of
structure created the surfaces are bounded by a beam
structure.
The structural members for analysis are also repre-
sented as shell elements. This means that the solid
model again needs to be transformed into shells and
the appropriate constants such as thickness applied,
see Fig. 7 for an example of the simpli®cation under-
taken. In some areas the members are represented as
Fig. 6. Example product model tree.
beam elements. In this case the solid model is trans-
formed into a beam representation and the appropriate
constants, such as cross-sectional area and moments of
lumped mass rather than in any detail, and these are inertia, are calculated from the solid model and
obtained from a library of such standard parts. The applied to the beams. The rule sets for how the model
manner in which the analyst simpli®es the initial CAD is to be transformed have been created such that this is
model is the same regardless of the size or type of done automatically by the DART system and an
vehicle being analysed as, for example, the same types appropriate mesh created.

Fig. 7. Simpli®ed beam model.


M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10 7

Fig. 8. Example FE mesh.

The various panels and structural members are DART system regarding the creation of the mesh
joined in particular ways. They may be spot welded, and hence this is done automatically and takes only
bonded or a mixture of both depending upon the a few minutes. Figs. 8 and 9 show examples of the
material selected. Rule sets have been created for FE meshes created automatically within the DART
how these joints should be represented within the system.
FE model. The particular types of joints to be used The vehicle structure usually requires two different
will be determined by the DART system depending types of analysis models to be created. One for stiff-
upon the material type. If spot welds are used for ness analysis and one for impact modelling. To
example, then the DART system will automatically demonstrate the feasibility of the approach adopted
determine the pitch of them based on its rule sets. If a here only the rule sets for the generation of the FE
cast aluminium joint is to be used, the DART system mesh for the stiffness analysis have been created. The
automatically determines the allowable mating length
and adjusts the connecting members automatically.
This information is used when simplifying the model
for analysis such that the appropriate position and
types of element to represent the joints is chosen and
their associated constants such as stiffness are applied.

3.4. Mesh generation

Once the simpli®ed representation of the design


model suitable for the creation of an FE mesh is
created, as discussed in Section 3.3, it could be passed
to an external system for mesh generation. However,
as the AML software used in the project has a module
for mesh generation the FE mesh can be automatically
generated within AML. Rules have been created in the Fig. 9. Example FE mesh.
8 M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10

output ®le from DART is in a format suitable for use in modelling route is automated. From an understanding
the ANSYS analysis software [20] with various types of the design and analysis process it can be seen that
of elements such as beams and shells being used. This there is a stage hardly discussed, that of simpli®cation
output ®le contains all of the ANSYS commands or model transformation. This step is critical for the
required to undertake a stiffness analysis of the struc- engineering process and often creates extra models
ture. The boundary conditions for the analysis are that have no associativity with the original CAD
applied within DART from a database of standard models. The DART system described here overcomes
boundary conditions for different sizes and types of these limitations by having the model transform itself
vehicles. Similarly, a database of material properties is automatically. This is possible because of the dynamic
available. The DART system automatically chooses nature of the object oriented representation used to
the element types based on the rules such that, for model the design process within the DART system.
example, a different shell element is used if the Thus, a material change, such as a change from a steel
structure is composite to that used if the structure is member to aluminium, could entirely restructure the
steel. The user does have the ability to override the design. This method of generative modelling allows
system such that solid elements may be used in place the designer to use the KBE model and rapidly per-
of shells for example, if they are more appropriate. As form a series of ``what-if'' analyses allowing a number
already mentioned, the element properties required by of alternative design solutions to be considered.
the analysis software are computed automatically. In The time taken to gather and create a rule base for a
addition, the elements are grouped in such a manner to KBE system can be justi®ed only if the same generic
facilitate ease of manipulation. problem is faced many times. This is the case in the
Such automatic creation of an FE mesh based on analysis of the automotive structure described here
rule sets does tend to formalise this information and where essentially the same generic components are
allows a standardised mesh to be created time after analysed for the same types of boundary conditions.
time. This ensures that the same methods of modelling All that may differ between analyses are the effects of
are always used for particular items such as spot the particular vehicle styling and the values of the
welds, the same types of elements are used, the same applied loads.
element properties are used and the geometry is KBE systems allow for easily customised user
simpli®ed in the same manner. However, within a interfaces to be created and the product model created
KBE system there is always the possibility of allowing by the rule base to be presented to the user in various
the users to override defaults such that design ways. There are two main ``views'' of the product
creativity is not sti¯ed and other possible solutions process required here, a full product representation
studied. and the simpli®ed FE analysis representation. Tradi-
tionally the analyst needs to modify and simplify the
design. This will either be done by editing the CAD
4. Discussion data or by creating a completely new model within the
analysis software. The use of the DART system
CAD and analysis software companies have recog- developed here negates the need to create a new model
nised the need for integration and have either inte- for analysis. The detailed geometry created is auto-
grated with other products or have modelling and matically transformed into the form required for
analysis software within one system environment. analysis purposes hence reducing duplication of the
This, however, does not go far enough and may well CAD data. Geometry simpli®cation for the creation
lead to rework. A design system must have within it of an FE mesh is achieved using the designers knowl-
the ability to represent a single product model as edge that is stored in the DART system as rule sets.
various alternative engineering views. There is a great The knowledge used in KBE systems can be obtained
deal of confusion in this area, because CAD systems from many different sources. However, in this case,
create a geometric model and through data transfer or the rules regarding the simpli®cation of the geometry
direct integration pass the model to other systems and and the subsequent generation of the FE mesh for
thus, CAD vendors suggest that the design analysis the structure are obtained mainly from past experience
M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10 9

as few formal design guidelines exist and few 5. Conclusions


design rules for this type of activity are actually
documented. The DART system created here has demonstrated
the potential bene®ts achievable by the use of KBE
4.1. Advantages of DART technology in the area of FE model creation. The KBE
system can automatically transform the detailed geo-
The number of elements used in the analysis of the metry into the form required for analysis purposes
type of automotive structure considered here is typi- hence reducing duplication of the CAD data. Thus, by
cally between 80,000 and 250,000, and the mesh using the KBE system the analysis department can
generation can entail up to 15 man weeks of effort respond much more quickly to potential design
upon receipt of the CAD model. Currently, due to the changes imposed by other departments. Whilst the
time and cost required to generate the analysis models, system described here is speci®c to an automotive
they are often used in a post-design phase to evaluate a structure the methodology used is generic and can be
®nal design that will only be modi®ed if the results are applied to any structure where the same steps in the
unacceptable. This work has demonstrated that if the analysis are undertaken time after time.
geometry simpli®cation and mesh generation for a
structure as shown in Fig. 2 is achieved in a KBE
system then the FE mesh can be generated in minutes. Acknowledgements
The meshing is done with respect to heuristic material
and analysis solution rule bases at an object level. The This work was undertaken as part of a UK Govern-
FE mesh created in DART is directly associated with ment EPSRC sponsored research programme under
the model geometry and thus, any alteration of the the IMI Land Transport programme entitled Structu-
model due to packaging problems for example, does rally Advanced Lightweight Vehicle Objective Project
not entail costly and time consuming rework. The new 4 (SALVO4) Structural Analysis of Hybrid Structures.
mesh is automatically regenerated in minutes thus, the The partner companies to the University of Warwick
analysis department could respond much more in this programme are Rover Group, British Steel, and
quickly to potential design changes imposed by other Ove Arup Ltd. The AML software has been provided
departments. by TechnoSoft Inc.
The use of a KBE system also means that many
more design options can be studied in the same time as
was previously used and thus, the use of new materials References
to optimise the solution can be investigated. In a
traditional engineering environment more material [1] T. Robertson, B. Prasad, R. Duggirala, A knowledge based
options tends to lead to slower overall design and engineering method to integrate metal forming process design
analysis times. Thus, these material options may not and simulation, in: P. Bocks, B. Prasad (Eds.), Proceedings of
the 1994 ASME Database Symposium, Engineering data
be investigated at all due to the need to do things Management: Integrating the Engineering Enterprise, ASME,
quickly. Use of different materials in a hybrid material 1994, pp. 41±50.
structure may mean that different element types are [2] K. Haghighi, E. Kang, A knowledge based approach to the
needed, e.g. composites may use a laminated shell adaptive ®nite element analysis, in: I. Babushka, J. Flaherty,
element, and the joint types are also different. This W. Henshaw, J. Hopcroft, J. Oliger, T. Tezduyar (Eds.),
Modelling, Mesh Generation and Adaptive Numerical
selection of element types is done automatically in the Methods for Partial Differential Equations, The IMA
DART system once the material type is applied as the Volumes in Mathematics and its Applications, Vol. 75,
system has rule sets for the simpli®cation of the Springer, Berlin, 1995, pp. 267±276.
geometry and the types of element to be used. Thus, [3] M. Jones, M. Price, G. Butlin, Geometry management support
for auto-meshing, in: Proceedings of the 4th International
the use of the DART system means that the engineers
Meshing Roundtable, 16±17 October, Sandia National
can investigate many more `what-if' scenarios than Laboratories, New Mexico, 1995, pp. 153±164.
would traditionally be possible due to time con- [4] M. Sheehy, I. Grosse, An object oriented blackboard based
straints. approach for automated ®nite element modelling and analysis
10 M. Pinfold, C. Chapman / Computers in Industry 44 (2001) 1±10

of multichip modules, Engineering with Computers 13 (1997) cation, Education and Control, Computational Mech. Publish-
197±210. ers, 1987, pp. 209±220.
[5] M. Saxena, R. Irani, A knowledge based engineering [16] J. Mackerle, K. Orsborn, Expert systems for ®nite element
environment for automated analysis of nozzles, Concurrent analysis and design optimisation Ð a review, Engineering
Engineering: Research and Applications 2 (1994) 45±57. Computers 5 (1988) 90±102.
[6] C. Chapman, The design process: a need to rethink the [17] G. Miller, An object oriented approach to structural analysis
solution using knowledge based engineering, MSc Thesis, and design, in: B.H. Topping (Ed.), Arti®cial Intelligence
University of Warwick, 1997. Techniques and Applications for Civil and Structural
[7] J. Patel, S. Singhal, B. Prasad, A. Vertiz, Towards the use of Engineers, Civil-Comp Press, 1989, pp. 149±155.
knowledge based engineering system in a concurrent [18] I. Taig, Expert aids to ®nite element system applications, in:
engineering environment, in: S. Ganesan (Ed.), Proceedings D. Sriram, R.A. Adey (Eds.), Proceedings of the 1st
of the Conference on Advances in Concurrent Engineering Ð International Conference on Applications of Arti®cial
CE97, Technomic Publishers Co., 1997, pp. 379±385. Intelligence to Engineering Problems, Computational Mech.
[8] S. Cooper, I.-S. Fan, R. Roy, K. Sehdev, Reuse of machining Publishers, 1986, pp. 759±770.
knowledge in aircraft design, in: S. Sivaloganathan, T. Shahin [19] W. Zeiler, State of the art, object oriented hybrid intelligent
(Eds.), Proceedings of Conference on Design Reuse Ð CAD system, Computers in Industry 20 (1) (1992) 1±9.
Engineering Design'98, 1998, pp. 263±270. [20] ANSYS Inc., Southpointe, 275 Technology Drive, Canons-
[9] Adaptive Modelling Language (AML), TechnoSoft Inc., 4424 burg, PA 15317, USA.
Carver Woods Drive, Cincinnati, OH 45242, USA.
[10] T. Desaleux, J.-M. Fouet, Expert systems for automatic Dr Martyn Pinfold is a Senior Fellow in
meshing, in: K. Bathe, D. Owen (Eds.), Proceedings of the the Warwick Manufacturing Group at the
International Conference on Reliability of Methods for University of Warwick. He obtained his
Engineering Analysis, Swansea, 1986, pp. 503±514. Ph.D. in the area of the analysis of com-
[11] D. Cheney, Model quality testing: the key to system posite components. He has been involved
interoperability, NAFEMS Benchmark Ð The International in collaborative research for several years
Magazine for Engineering designers and Analysts, October, through UK Government funded projects
1997, pp. 26±29. with the automotive industry concentrat-
[12] S. Fenves, A framework for a knowledge based ®nite element ing on the analysis of composite auto-
analysis assistant, in: C. Dym (Ed.), Proceedings of the motive structures. Prior to this he worked
Applications of Knowledge Based Systems to Engineering in the CAD/CAM centre at the University.
Analysis and Design, ASME, FL, 1985, pp. 1±7.
[13] S. Fenves, A framework for cooperative development of a Craig Chapman is a Senior Fellow with
®nite element modelling assistant, in: K. Bathe, D. Owen a long industrial track record working
(Eds.), Proceedings of the International Conference on within various industrial sectors provid-
Reliability of Methods for Engineering Analysis, Swansea, ing design engineering solutions in an
1986, pp. 475±486. international consulting environment. He
[14] A. Kissil, H. Kamel, An expert system ®nite element joined WMG after working in the USA
modeller, in: D. Sriram, R.A. Adey (Eds.), Proceedings of on KBE projects for companies such as
the 1st International Conference on Applications of Arti®cial Allied Signal, Boeing, Goodyear and
Intelligence in Engineering Problems, Computational Mech. Sangiorgio System Technology. Prior
Publishers, Vol. 2, 1986, pp. 1179±1186. to this he worked for British Aerospace
[15] J. Baugh, D. Rehak, FOOBAR: an arti®cial intelligence based as a Principal Production Engineer. He
®nite element system, in: D. Sriram, R.A. Adey (Eds.), has actively participated in KBE research over a number of years
Knowledge Based Expert Systems for Engineering: Classi®- and is now leading the KBE initiative at Warwick.

Anda mungkin juga menyukai