Evolution in Mechanical Design Automation and Engineering Knowledge Management

Abstract

Design activity consists, strictly speaking, in synthesizing something new (or arranging existing things in a new way) to satisfy a recognized need. This activity is accomplished through an iterative, knowledge-based, decision-making process. One of the goals pursued by IT applied to product development has been Design Automation that is the execution of all tasks of the design process of a product by a software application. In this context, engineering knowledge elicitation, representation and management are key issues to achieve the so called “right the first time” design. This chapter summarizes author’s research in Mechanical Design Automation (MDA) domain during the last two decades; it focuses the evolution from CAD centered applications, to Object-Oriented ones up until the more recent issues related to storing, sharing and reusing of engineering knowledge, the role in PLM approach, the support to decision making. Finally, the fundamental aspects related to development of real MDA applications are dis-cussed, on the basis of personal experiences of the authors

Full text

DISCLAIMER
This is a pre-copyedited version of the article submitted for publication.
The definitive publisher-authenticated version is available from the publisher.

Chapter 4
Evolution in Mechanical Design Automation
and Engineering Knowledge Management
Giorgio Colombo1 and Ferruccio Mandorli2
Abstract Design activity consists, strictly speaking, in synthesizing something
new (or arranging existing things in a new way) to satisfy a recognized need. This
activity is accomplished through an iterative, knowledge-based, decision-making
process. One of the goals pursued by IT applied to product development has been
Design Automation that is the execution of all tasks of the design process of a
product by a software application. In this context, engineering knowledge elicita-
tion, representation and management are key issues to achieve the so called “right
the first time” design. This chapter summarizes author’s research in Mechanical
Design Automation (MDA) domain during the last two decades; it focuses the
evolution from CAD centered applications, to Object-Oriented ones up until the
more recent issues related to storing, sharing and reusing of engineering
knowledge, the role in PLM approach, the support to decision making. Finally, the
fundamental aspects related to development of real MDA applications are dis-
cussed, on the basis of personal experiences of the authors.
4.1 Introduction
During his/her design activities, mechanical designers must have the skill to syn-
thesize into a final solution, conflicting issues related to required functions, prod-
uct shape, manufacturing techniques and materials, economic constraints.
The design process, as well as the production processes, has been deeply stud-
ied during the past: the traditional actions-based representation of design, based on
a sequence of linked activities, has been changed in a product-centered representa-
tion, able to convey new paradigms like concurrent engineering and co-design.
1 G. Colombo ()
Politecnico di MIlano
e-mail: giorgio.colombo@polimi.it
2 F. Mandorli ()
Università Politecnica delle Marche
e-mail: f.mandorli@univpm.it

2 Giorgio Colombo Ferruccio Mandorli
In the last decade, others product related aspects, like distribution, mainte-
nance, disposal and recycling, came up to the design and production and, together
with the use phase, lead to the concept of product life cycle.
As this evolution took place, the product development process shifted from a
production-driven process to a design-driven process, based on engineering
knowledge management.
In this context, Design Automation (DA) has played an important role to effec-
tively innovate and improve design procedures; in industrial engineering we can
more properly speak of Mechanical Design Automation (MDA).
MDA can be defined as a set of methods, tools and applications that permit to
automate the design process; it can be applied to all the phases of the process,
from the conceptual to final one related to the production of technical documenta-
tion. However, in our opinion, MDA well fits the phases of parametric and de-
tailed design during which a technical solution is completely formalized. In par-
ticular, more important benefits from a MDA application can be achieved when
dealing with products and parts characterized by a well defined architecture and
design process. Some examples of products falling within this category are heat
exchangers, gears, machine tools, industrial mixers and so on. For these types of
products, a completely automated procedure can be implemented and the configu-
ration of a new product can be executed by a software application.
DA has a history that matches with that one of IT; first applications were de-
veloped with general purpose programming languages and concerned specific as-
pects of design process such as kinematic analyses or structural calculation. A re-
markable improvement was gained when CAD techniques was developed; in fact,
graphics and geometric modeling play a fundamental role in product design. The
integration between CAD models and programming languages permitted the de-
velopment of automatic procedure to configure parts and simple products; there
are recent examples of this approach [1]. This approach was enhanced by the de-
velopment of parametric CAD [2]; a lot of applications to configure parts and
products have been developed by using parametric models and programmable
tools such as spreadsheet. In SMEs (Small Medium Enterprises) this is actually
one of more diffuse method to develop simple DA applications.
At the same time, a number of methods and tools derived from Artificial Intel-
ligence appeared; in particular, Knowledge Based Engineering (KBE) tools based
on Object-Oriented approach which constitute also today a valid approach to DA.
They had a relevant impact within aeronautical and automotive companies; some
industrial applications have been developed in these contexts [3-4]. Side by side,
researches on methods to acquire and formalize knowledge have been started; in
fact, knowledge formalization is a fundamental issue, and several efforts [5] have
been carried out in this direction. KBE tools evolved from initial elementary im-
plementation to the actual one characterized by simple programming language,
powerful tools to define customized GUI and to integrate external programs (CAD
systems, FE solvers, spreadsheet, data base, etc.).

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 3
In this chapter we present our experiences in the last two decades in developing
real industrial DA applications and we discuss topics in this context and in Engi-
neering Knowledge Management (EKM) in general. In detail, we present and dis-
cuss the evolution from first applications centered in parametric CAD integrated
with programming languages or calculus tools, then to ones based in O-O ap-
proach and finally the new approach based on agents assembled to carry out all the
design tasks.
Moreover, attention is also focused on engineering knowledge; it is very com-
plex, differentiated and interrelated, involving explicit and implicit information,
often based on tacit assumptions. A key aspect in the development of MDA appli-
cations is the elicitation and the formalization of the design knowledge and the
representation of such knowledge into the design support system. This issue is at
higher level than ones of MDA and will determine functionalities of IT systems
finalised to design process.
4.2 A brief history of our works in MDA and EKM
As mentioned, the history of DA and EKM matches practically with that one of
IT; first applications, finalized to mechanism or structural analyses, focused spe-
cific problems without considering the complete process; only in recent years at-
tention is focused on a global approach to design.
CAD techniques development represented a milestone in MDA; we highlighted
the fundamental role played by graphics and geometric modeling in design pro-
cess. In particular, parametric CAD has represented, and represents also today, a
powerful tool to implement MDA. The first our contribution in DA domain date
from 1989; in [2] Colombo et al., proposed an approach to represent design rules
directly related to parametric drawing of mechanical parts. Innovations presented
in that paper were: a 2D parametric CAD system based on representation of geo-
metric entities and constraints in a graph, a spreadsheet-like tool to manage geo-
metric and functional parameters and a simple programming language to define
sizing or analyses rules. A lot of applications to configure parts and products have
been developed by using parametric models and programming languages or tools
such as electronic spreadsheets. Today this approach is the most used in SMEs
(Small Medium Enterprises) to develop MDA applications; the evolution, in this
case, is not in the approach but in the better tools functionalities.
At the same time, methods and tools developed in Artificial Intelligence do-
main were applied to Engineering Design; in particular, Object-Oriented (O-O)
programming furnished a great contribute. The first KBE (Knowledge Based En-
gineering) development systems were base on an O-O programming language [6].
We approached this technology and we realized some prototypes by using a sys-
tem named The Concept Modeller® from Wisdom Systems. In [7] we presented a
MDA application to configure a family of shearing dies starting from the drawing

4 Giorgio Colombo Ferruccio Mandorli
of the punch sheet. In this application significant contributions were the O-O rep-
resentation of the product structure and the use of a graphic language proposed by
Coad and Yourdon [8] to synthesize object, parameters and methods implemented.
This last aspect is important in EKM; UML was further developed starting from
this first approach to the problem of representing object, data, and relations.
A first discussion in differences between approaches to MDA based on CAD
models and programming languages or tools and O-O programming was presented
in [9].
In the last decade we concentrated our attention in dissemination of MDA
technologies and we developed some applications by using different tools which
can be referred to the two previously cited approaches; some methodological as-
pects are presented in [10-11]; while examples of implementation with different
KBE tools are summarized in [12]. Susca et al. in [13] presented an application to
automatically calculate mass properties of a racing car, developed by using the
KBE tool named Selling Point® by Oracle. Just to the new century, MDA was ap-
plied mainly in aerospace and automotive companies, i.e., industrial contexts with
relevant human resources and know-how (a selection of KBE publications pro-
duced by the Design of Aircraft and Rotorcraft group at Delft University is col-
lected in the booklet "Knowledge Based Engineering Supported Design", availa-
ble on the DAR web site).
In the last years our attention was addressed to implement MDA in SMEs, and,
moreover, we identified issues of MDA: product modeling, process modeling, in-
tegration with PDM/PLM, and knowledge reusability and sharing and support to
decision making. Examples of MDA applications developed for SMEs are report-
ed in [14-15].
Experiences in developing MDA in SMEs were synthesized in a methodology
named MEDEA, which considers all the tasks of the development of an applica-
tion and proposes methods and tools for each of them [16].The issue of the inte-
gration with PLM/PDM systems was discussed in [17]; in particular, the part cod-
ing was used to retrieve existing components in company data base. In various
contexts we highlighted relevance of product and process representations; exam-
ples are in [16] and [18]. An example of knowledge reusability and sharing was
presented in [19].
In the following sections, by using examples, we will clarify our approach to
MDA and the results obtained in our work.
4.3 Methods and Tools for Mechanical Design Automation
The overall objective of MDA is to develop computerized systems able to sup-
port the engineer in performing his/her design related tasks, generally within the
context of product or process design.

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 5
Although MDA systems can be very different to each other, they are usually
generative systems, i.e. they use a set of predefined procedures to automatically
(or semi-automatically) generate different outputs starting from different inputs.
Fully automatic systems just require to the user to specify the inputs; semi-
automatic systems can require additional interactions with the user in order to
solve special cases and exceptions, not covered by the predefined procedures.
Generally speaking, the inputs are the design specifications; the generative pro-
cedures are the design rules and the outputs are the design solutions.
In order to implement such kind of systems, both methodological and techno-
logical issues must be addressed and solved.
From the methodological point of view, the issue concerns the definition of a
set of appropriated procedures and guidelines to identify, to capture, to formalize
and to maintain the knowledge related to the problem under inspection.
From the technological point of view, the issue concerns the definition of a set
of tools and functions suitable to implement and manage the data structures and
the procedures used to represent the design process knowledge.
Several methodologies to support MDA implementation in real industrial con-
text have been proposed: some of them are particularly suitable for the DA im-
plementation in a Small/Medium Enterprise (SME), others have been proposed by
consortium of large companies.
A common feature of all the proposed methodologies is the Object Oriented
approach to the problem analysis: the product (or process) to be designed is ana-
lyzed and represented as a hierarchical structure made of objects that are defined
by means of a set of properties. The instantiation of the properties values define a
specific design solution. The procedures to compute the properties values repre-
sent the design rules.
The MEDEA (Methodology per Design Automation) methodology focused the
attention on some characteristics of the design process in SMEs [16]. It proposes a
step by step roadmap and suggests methods and tools finalized to developers more
skilled on products and design process than on IT technologies. It is based on five
main steps:
 Specs definition: identification of DA application specs and the crite-
ria to make re-usable and sharable blocks of the product and process
data;
 Knowledge acquisition: collection of the knowledge related to the
product architecture and the design process;
 Knowledge formalization: representation of product architecture (tree
diagram and UML class diagram [20]) and of the process model
(IDEF0 and IDEF3 diagrams [21]);
 Integration with PDM/PLM system: definition of the interactions be-
tween DA application and company’s PDM/PLM system;
 Implementation of DA application using a KBE system

6 Giorgio Colombo Ferruccio Mandorli
An alternative methodology has been proposed as the result of the EU project
MOKA "Methodology and tools Oriented to Knowledge based Applications" [5].
The proposed methodology focuses on knowledge formalization and data structur-
ing, mainly from an implementation point of view.
MOKA identifies four key technical roles in developing DA applications: the
domain experts (they provide main source of knowledge and specification); the
knowledge engineers (players that orchestrate the capturing and formalizing activ-
ities); the software engineers (the people who builds software, based on require-
ments set by knowledge engineers); the end users (the ultimate customers of the
application). The MOKA roadmap defines the development paths and who does
what, when, and what to do when things go wrong.
Additional proposed methodologies are: KOMPRESSA (Knowledge-Oriented
Methodology for the Planning and Rapid Engineering of Small-Scale Applica-
tions) and DEKLARE (Design Knowledge Acquisition and Redesign Environ-
ment) [22].
From the implementation point of view, the issues are related to the definition
of appropriated data structures and algorithms to represent and manage the prod-
uct/process model as well as the implementation of the procedures to compute the
model properties.
Although MDA can be very different from each others, some common features
and functionalities can be identified: the product/process model is represented as a
hierarchical structure made of parts and sub-parts, defined by means of a set of
properties, related to each other in a parametric way.
The properties are used to store all the different types of information that are
relevant to consolidate the design knowledge into the product model. This infor-
mation can be stored in form of explicit knowledge (i.e. properties values, like
standard dimensions, materials, physical properties, etc.) or procedural knowledge
(i.e. procedures to compute the properties values, like dimensioning algorithms,
configuration and compatibility rules, check and analysis procedures, etc.).
In order to manage such kinds of models, the system must incorporate some
dependency backtracking or history-based mechanism to keep updated the proper-
ties values.
DA systems are focused on knowledge capture and on the definition and im-
plementation of design procedures. In the mechanical field, it often happens that a
significant part of the information to be managed during the analysis and synthesis
design phases is related to components shape, position and dimensions. For this
reason, it is quite common that a significant module of a DA system is the geomet-
ric kernel.
Different approaches and IT tools can be used to implement DA applications
and the last two decades have shown the impact of IT evolution on DA applica-
tions development. In the following, we summarize the three main approaches for
the development of DA applications, based on different technologies: CAD cen-
tered approach, KBE approach and mixed approach

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 7
The CAD centered approach is particularly suitable to develop DA applications
where the largest part of the information to be managed is somehow related to the
geometric properties of the product model. The CAD model, enriched with appro-
priated customized properties, represents the product model. The use of parametric
CAD model templates allows the definition of generic models. Traditional pro-
gramming languages can then be used to implement algorithms that, by using the
functions provided by the CAD Software Development Kit (SDK), allow to in-
stantiate the parameters stored into the model templates, in order to generate a
specific product model configuration.
The KBE approach is particularly suitable to develop DA applications where
geometry is one of many kinds of information to be managed. By following this
approach, 2D or 3D models are usually outputs, automatically generated by the
system. It is based on the use of a KBE development framework. These frame-
works are based on a simplified Object Oriented programming language and pro-
vide functionality to handle the hierarchical product model structure, including
procedure to generate shapes, to set dimensions, to apply assembly rules, to access
to external data bases, etc.
Finally, the mixed approach is particularly suitable when the generative meth-
odology, provided by the KBE framework, needs to be coupled with the interac-
tion provided by the use of traditional CAD systems. The main feature of a KBE
development framework able to support this approach is that the geometric kernel
it is not embedded into the framework, but the framework provides functions to
interface and drive an external CAD system.
Theoretically speaking, the implementation of a DA system can be considered
a technicality. However, in a real industrial context, the selection of the most ap-
propriated approach and developing tool, is directly related to aspects such as (i)
the capability to foresee the effort and the expertise required to develop, to test, to
validate and to maintain the system; (ii) the capability to predict the trade-off be-
tween development costs and use benefits; (iii) the possibility for the domain ex-
pert and/or the knowledge engineer to directly participate to the system implemen-
tation; (iv) the possibility for the domain expert to maintain and upgrade the
system, by-passing the software engineer and/or the knowledge engineer.
In the following we introduce some examples of applications developed adopt-
ing the mentioned approaches.
4.4 DA Applications based on CAD models and software
components
The application example reported in this section refers to a KBE application
developed to support the design and production of gas turbine ducts. The research
was carried on in collaboration with an Italian company, supplier of General Elec-
tric.

8 Giorgio Colombo Ferruccio Mandorli
As shown in Figure 4.1, the input of the ducts design process is the 2D drawing
of the draft layout of the plant. The results of the design activity will be: the 3D
detailed design of the ducts, the 2D drawings required for the ducts production,
the bill of material (BOM) and the cost estimation of the ducts.
Fig. 4.1 Tasks share
During the design activity, the designer is constrained by economical consider-
ations (i.e., the use of standard components and semi-finished materials), techno-
logical aspects (i.e. the manufacturing processes), and limits imposed by standard
regulations (i.e., acoustic, structural and thermodynamic tests).
The traditional ducts design is a trial and test process: the preliminary design
solution is obtained by mapping the generic ducts parts, reported in the 2D draft
drawings, into standard ducts modules, on the basis of the designer experience.
The preliminary solution is, then, tested by means of the thermodynamic,
acoustic and structural tests imposed by standard regulations and design specifica-
tions. If the tests are not passed, the design must be modified and then the tests
need to be re-executed.
When the final design solution is achieved, all the documents required for the
production must be compiled, including BOMs made of several thousands of
components.
The efficiency of the described process is clearly affected by three main con-
straints: the capability of the designer to rapidly converge to a sound solution; the
time required to validate the solution (i.e. to perform the required tests); the time
required to compile the production documentation (i.e. drawings and BOMs).
In the described context, the objectives of a KBE design support system will
be: to capture and to consolidate the multidisciplinary know-how of the designer,
in order to let the system generate sound design solutions (from the manufactura-
bility and assimilability point of view); to automate as much as possible the repeti-
tive and time-consuming tasks, in order to shorter the time required to validate the
generated design solution and to compile the production documents.

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 9
The design knowledge elicitation started with the functional analysis of several
plant typologies. This analysis allowed identifying the modularity of the systems
and enabled the synthesis of the different types of plants into a default template
and a set of rules to manage the product variants.
The plant template was then represented with a hierarchical model structure by
using an Object-Oriented approach. Such model contains all the information and
rules required to generate the detailed geometric configuration of the different
parts of the ducts.
Figure 4.2 shows part of the hierarchical structure of the duct: the duct is an or-
dered set of modules, called items, along a 3D mean axis made of line segments.
Each item is generally made of four walls. A wall is made of the structural part,
the casing wall, and insulation. Casing wall consists on sheet metal panels, some
stiffeners and parts to support insulation (studs and scallop bars). The insulation is
made of claddings, edge trims and batten channels which all are sheet metal parts
that contain insulation fibers.
Fig. 4.2 Part of the duct hierarchical structure
From the methodological point of view, the results of the knowledge elicitation
phase have been logically organized by using formal structures like Configuration
Virtual Prototypes (CVP) and Design Structure Matrix (DSM), as explained in
[23]. In particular, the DSM approach has been used to identify and relate to each
other the functional requirements, as the CVP model has been used to define the
overall product structure.

10
Fro
m
gratin
g
p
rovid
e
ager.
T
of the
d
The
p
ut int
compo
n
Fig. 4.3
The
of the
d
model
t
5.000
p
If p
a
CAD
m
Sui
t
metric
a
thermo
ductio
n
m
the imple
m
g
in house d
e
e
d by a com
m
T
he user inter
a
d
esign proced
u
overall struc
t
erface, wher
e
n
ents of each
Duct modules
d
system takes
d
uct from
t
he
t
e
m
p
lates. A
s
p
arts.
a
rticular exc
e
m
odel in orde
r
t
able procedu
r
a
l informatio
n
dynamic and
n
documents.
m
entation poi
n
e
veloped algo
m
ercial 3D M
C
a
c
t
s with the
u
re.
t
ure of the du
c
e
he/she can
module of th
e
d
ata entr
y
few minutes
data specifie
d
s
an example,
e
ptions need
t
r
to make loc
a
r
es are then c
n
in order to
structural an
a
n
t of view, t
h
o
rithms and
d
C
AD syste
m
,
a
system at di
ff
c
t is defined
b
specify num
b
e
duct, as sho
w
to automatic
a
d
by the user
a
the model sh
t
o
b
e manag
e
a
l modificatio
n
c
harged to ac
c
extract the
d
a
lysis (see Fi
g
Giorgio Co
l
h
e system has
d
ata structure
s
a
relational d
a
f
erent levels
a
b
y the user th
a
b
er, type, di
m
w
n in Figure
4
a
lly generate
t
a
nd the
p
rede
f
own in Figur
e
e
d, the user c
a
n
s.
c
ess both geo
m
d
ata required
g
ure 4.5) as
w
l
ombo Ferrucci
o
been develo
p
s
with data s
t
a
tabase and
G
a
nd at differe
n
a
nks to a tabl
e
m
ensions and
4
.3.
t
he detailed 3
D
f
ined
p
arame
t
e
4.4 is made
a
n directly a
c
m
etrical and
n
to perform
a
w
ell as to gene
o
Mandorli
p
ed in
t
e-
tructures
G
UI man-
n
t stages
e
like in-
optional
D model
t
ric CAD
of about
c
cess the
n
on geo-
acoustic,
e
rate p
r
o-

4 Evolut
i
Fig. 4.4
A
Fig. 4.5
G
It is
w
tion is t
h
approa
c
linked
t
databas
e
The
d
ometric
In a
the hier
b
ly is
m
MCAD
The
u
ing exa
m
The
l
When
a
rithm tr
a
ing rel
a
i
on in Mechanic
A
utomatically g
e
G
eometrical an
d
w
ell known t
h
h
e syste
m
ma
c
h is based o
n
t
o part data (
e
e
.
d
efinition of
a
template and
similar way,
archical struc
m
anaged tha
n
system and t
h
u
se of assem
b
m
ples.
l
eft pa
r
t of Fi
a
specific inst
a
a
verses the a
s
a
tionships. Th
e
a
l Design Auto
m
e
nerated 3D du
c
d
non geometric
a
h
at a critical i
s
i
ntenance an
d
n
the wide u
s
e
.g. code, sta
n
a
new compo
n
the input of t
h
assembly te
m
t
ure of the d
u
n
ks to the tr
a
h
anks to the u
b
ly templates
g
ure 4.6 sho
w
a
nce of the te
m
s
sembly temp
l
e
final assem
b
m
ation and Engi
c
t mode
a
l data automati
c
s
sue related t
o
d
upgrade. In
o
s
e of parts te
m
n
dard dimens
i
n
en
t
requires
t
he data into t
h
m
plates and
d
u
cts. The posi
t
a
ditional mat
i
u
se of 2D sket
c
and sketches
w
s the templa
t
m
plate needs
t
l
ate and colle
c
b
ly is
t
hen a
u
n
eering Knowl
e
c
ally extracted
fr
o
the develop
m
o
rder to face
t
m
plates (para
m
i
ons, material
the interactiv
e
h
e database.
d
atabase data
t
ioning of the
i
ng relations
h
c
hes.
is summariz
e
t
e of the silen
t
o be generat
e
c
ts compone
n
u
tomatically g
dge Manageme
n
r
om the duct m
o
m
ent of KBE
a
h
is issue, the
m
etric CAD
m
, etc.) stored
e
modeling o
f
a
re used to r
e
parts into th
e
h
ips available
e
d by the two
c
er panel fra
m
d,
t
he assem
b
n
t references
a
e
nerated by c
nt
11
o
de
applica-
adopted
models)
into the
f
the ge-
e
present
e
assem-
e
in any
follow-
m
ework.
b
ly algo-
a
nd mat-
c
hecking

12 Giorgio Colombo Ferruccio Mandorli
the components that must be present, by generating the models of the parts to be
included in the assembly and automatically applying the mating constraints (as
shown in the right part of Figure 4.6).
Fig. 4.6 Example of assembly template (left) and assembly instance (right)
Simplified layouts using 2D sketches can be successfully used to add flexibility
in the management of part positioning.
In case of insulation panel definition, the basic components to be assembled are
studs, scallop bars and claddings (see lower part of Figure 4.7).
Fig. 4.7 Example of 2D sketches to drive the automatic assembly procedure
Studs are characterized by revolution geometry and are welded on panels.
Then, only points are necessary for their position: circle centers will represent
those points. Scallop bars are represented by segments which, combined with stud
circles, provide all necessary geometrical parameters. Claddings are represented
by closed paths passing through stud centers.

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 13
The insulation panel layout can then be defined by using 2D sketches where
different types of 2D geometrical entities are used to identify the component type
and position (see upper part of Figure 4.7).
The default layout can be computed on the basis of known rules and specific
panel dimensions. The sketch corresponding to the default layout is then automati-
cally generated by the system when an insulation panel needs to be instantiated. If
a different layout is required, the user can interactively modify the 2D sketch and
ask the system to re-compute the configuration.
The automatic procedure analyzes the modified sketch and extracts the infor-
mation for the instantiation and generation of the parts and the subassemblies. Fi-
nally, the components are assembled to their final configuration, as shown in Fig-
ure 4.8
Fig. 4.8 Example of instantiated assembly
The adopted approach has proven to be a good compromise between a fully au-
tomated procedure and the need to manage highly configurable product variants.
4.5 DA applications with KAE tools
This section describes two industrial applications developed adopting the
MEDEA methodology. The first is related to the configuration of fired heaters,
while the second concerns an industrial mixer and is focused on process represen-
tation and integration with PDM system.

14
4.5.1
F
The
to mar
k
config
u
concer
n
p
roach
,
icated
m
and tra
n
to gen
e
requir
e
(KB) a
p
The
p
artial
l
an ima
g
Fig. 4.
9
A f
i
flowin
g
leavin
g
the en
v
Ge
n
cabin
a
Thi
s
warme
d
mally,
b
ox fo
r
Acc
and pr
o
develo
p
F
ired Heat
e
term “produ
c
k
eting or to p
r
u
ration and t
o
n
s the secon
d
,
where the c
u
m
achines or i
n
n
slate the cu
s
e
rate configur
e
ments. In su
c
p
plication to
a
fired heate
r
i
l
y vaporize oi
l
g
e of a chemi
c
9
Petroleum ref
i
i
red heater c
o
g
inside the t
u
g
the radiant
s
v
ironment thr
o
n
erally, more
a
nd cylindrica
l
s
classificatio
n
d
up. This v
a
vertical cylin
d
r
higher value
ording to M
E
o
cess and fol
l
p
ment of the
K
e
r Con
f
i
g
u
r
c
t configurati
o
r
oduct devel
o
o
engineering
d
case and, in
u
stomer need
s
n
dustrial pla
n
tomer’s nee
d
s
a
tions of the
p
c
h a context,
a
utomatically
i
s a sub-syste
m
l
in order to s
c
al plant and
o
i
nery plant
o
nsists of a r
a
u
be coils by
r
s
ection give
u
o
ugh the
s
tac
k
common fire
l
.
n
is based on
a
lue permits t
o
d
rical type is
s
.
E
DEA metho
d
l
owing forma
l
K
B applicatio
n
r
ation
o
n” means di
ff
o
pment that r
e
or technical
c
n
particular, t
o
s
highly custo
n
ts. In this ca
s
s
using an en
g
p
roduct and
o
it has been
configure fir
e
m of petrole
u
s
eparate it fro
m
o
f a fired hea
t
a
dian
t
sectio
n
r
adiation, a c
o
u
p their heat
b
k
.
e
heaters can
n
the amount
o
t
o determine
t
adequate up
t
d
ology, kno
w
lization have
n.
Giorgio Co
l
ff
erent activiti
e
e
spectively re
f
c
onfiguration
o
an Enginee
r
m
ized produ
c
s
e, company
s
g
ineering con
f
o
f its compon
e
implemente
d
e
d heater for
a
u
m refinery p
l
m
the hydroc
a
t
er.
n
whose heat
i
o
nvection se
c
b
y convectio
n
be grouped
i
o
f heat absor
b
t
he most con
v
t
o 20-25 106
w
ledge acquis
i
been two im
p
l
ombo Ferrucci
o
e
s that can be
f
ers to the co
m
.
The first ap
p
r
To Order (
E
c
ts, such as sh
s
taff has to u
n
f
iguration sys
e
nts that satis
a Knowled
g
a
chemical pl
a
l
ants used to
a
rbons. Fig. 4
.
i
s transmitte
d
c
tion where fl
u
n
and then re
l
i
n (Figure 4.
1
b
ed by the fl
u
v
enient soluti
o
k
cal/h; while
i
tion for both
p
ortant phase
o
Mandorli
oriented
m
mercial
plication
E
TO) ap-
h
ips, ded-
n
derstand
s
tem able
s
fy initial
g
e Based
a
n
t
[24].
heat and
.9 shows
d
to fluid
l
ue gases
l
eased in
1
0): box,
u
id to be
on. Nor-
cabin or
h
product
e
s for the

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 15
Fig. 4.10 Fired heater: A) Box, B) Cabin and C) Cylindrical
Knowledge has been acquired using different sources: interviews with technical
staff, company technical manuals, standards [25-26] and scientific books.
.For the formalization step, different techniques and models have been adopted.
First the product structure has been represented as a tree where it is possible to see
the mentioned subsystems and their main sub-components; then, it has been trans-
lated into UML Static Class Diagrams where subsystems, components, and rela-
tionships are properly represented. Figure 4.11 portrays the main UML diagram
where the three mentioned subsystems (radiant, convenction and stack) are repre-
sented.
Process knowledge has been formalised adopting IDEF0 and IDEF 3 Process
Flow (PF) models (www.idef.com). The design process of a fired heater is based
on two interrelated principal activities, partially carried out in parallel:
 parts are dimensioned applying design rules derived from the thermody-
namics;
 designer’s choices and calculated parameters values are verified on the
base of mechanical and fluid dynamics principles.
Design rules that permit both to size correctly each component and generate the
product configuration have been associated to the IDEF0 and UML diagrams.
Rules have been derived both from thermodynamics principles and designers’ ex-
perience. As an example, consider activity “Characterize thermodynamic process”
or UML class Fire heater, related design rules to calculate heat fired are:
 HFired = Duty Tot /η
where:
HFired: = Heat Fired [MW];
DutyTot = Total Duty [MW].
 H Losses = HFired . ηLosses
where:
HLosses = Heat Losses [MW];
ηLosses = Efficiency Losses.

16
 ηt
o
Fig. 4.1
The
RuleSt
r
Fig
u
where
p
UML
d
ture.
It is 2,5
%
o
t = η + ηLos
s
where:
ηtot = Tota
l
η = efficie
n
1 Main UML
C
DA applica
t
r
ea
m
.
u
re 4.12
p
ort
r
p
arts are ord
e
d
iagrams. Th
e
%
when a pre
h
s
es
l
efficiency;
n
cy.
C
lass Diagram
t
ion has bee
n
r
ays the pro
d
e
red accordin
g
e
refore, a set
o
h
eated system
n
realized us
d
uct tree stru
c
g
to assembl
y
of elements l
i
Giorgio Co
l
is used other
w
ing the com
m
c
ture (represe
n
y
and sub-ass
e
i
nked by rule
l
ombo Ferrucci
o
w
ise it is 1,5
%
m
ercial KBE
n
ting its arc
h
e
mbly logic a
n
s
composed t
h
o
Mandorli
%
syste
m
,
h
itecture)
n
d to the
h
e struc-

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 17
Fig. 4.12 Implemented product structure and design process
To implement the design process, it has been necessary to define the rules and
relationships among the parts and link them to the associated parameters. All ruled
related to the design process (formalized with IDEF diagrams) have been imple-
mented linking to each part properties a value derived from an analytical formula
or geometric relationship.
The design of a fired heater, such as many other products, requires the execu-
tion of specific design activities and in parallel the definition of the product archi-
tecture. Therefore the KB application has been based on two separate models: the
first for the plant components while the second for the configuration/design pro-
cess. Therefore, a communication channel to transfer data and information from
the process to the product and vice versa has been implemented.
The realised application automatically executes the configuration process, per-
forming all activity usually carried out by the designer. It starts with the acquisi-
tion of design specs and subsequent steps are executed in accordance with the im-
plemented process.
In collaboration with the technical staff of the involved company the prototype
has been tested with two test cases:
 A cabin fired heater with horizontal tubes that can generate a total duty equal
to 11.6 MW. It was already developed by the company and the goal has been
to demonstrate that the final results generated by the automatic configurator
are the same of those obtained following the traditional process within the
company.
 A cylindrical and a box fired heater (both able to generate a total duty equal to
10MW) applying the same data input in order to demonstrate the possibility
to rapidly generate and compare different design solutions.

18 Giorgio Colombo Ferruccio Mandorli
Figure 4.13 portrays the acquisition of customer’ reqs while Figure 4.14 the
configuration obtained for cylindrical fire heater (Figure4.14a) and for a box one
both with vertical tubes (Figure 4.14b).
Fig. 4.13 The KB application: entering design specs for thermodynamic characterization
A. B.
Fig. 4.14 Automatic fired heater configuration: A) Cylindrical; B). Box
All data generated, from the 3D CAD models to the final documentation sum-
marizing fire heater characteristics, were in accordance with those calculated by
the company.
This permitted to demonstrate either the accordance of the obtained results with
those obtained with the traditional process or the possibility to rapidly generate
and compare different design solutions.
Company staff considered DA application an optimal tool especially during or-
der acquisition to present the offer to the potential customer and generate technical
drawing of the plant, named general arrangements. In this case the introduction

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 19
within the company could be done in a short time. On the other hand, the introduc-
tion of such a technology in the technical departments requires deeper changes and
could be done at medium long term.
4.5.2 Integration KBE-PDM
This case study regards the design of a family of industrial mixers and is fo-
cused on the integration between KBE and companies’ data repositories (PDM
system) to improve knowledge reuse in mechanical companies [16-17].
The design process is addressed to produce 3D models, drawings and BOM,
with updating the company PDM system. This leads to a partial process re-
engineering to fulfill the complete integration between KBE/PDM. The designer
searches in the PLM/PDM models/documents already realized and tries to reuse it,
to avoid the redesign of existing components or products.
As in the previous case, before implementing the application with the KBE
kernel, knowledge acquisition and formalization are needed. Product tree structure
and UML static class diagrams have been developed to describe relationships
among parts and properties. The system is composed of three main sub-systems:
engine, rotor and stator. Figure 4.15 describes the rotor group, the relation with the
engine and the properties of the top block.
As mentioned, for this test case it has been also necessary to study the interac-
tion among DA application, PDM system and end-users. Accordingly to
MEDEA, it has been described with an UML Activity Diagram (Figure 4.16),
which represents the sequence of the steps necessary to complete the configuration
and archiving of a new mixer.
Also in this case we adopted Rulestram since it also enables designers to share
their knowledge and reuse their applications or part of them in other applications.
The source code of RuleStream applications lets the designers personalize their
applications, thus permitting the development of a module to retrieve data and
documents stored into the company’s PDM.
Figure 4.17 shows a snapshot of the application while Figure 4.18 the function-
al scheme.
The application allows the designer to automatically design and configure an
industrial mixer. It can be used in two different ways:
 To configure a new product following all steps defined to design a mixer,
such as the calculation of the forces acting on the blades, the sizing of the
shaft, and so on.
 To query the PDM to acquire product components already developed, using
their drawings, codes, and data in general. The application generates the code
for every component by which it can query the PDM and discover if a part
with the same code already exists. On positive case, it returns the component
and its drawings and avoids the next design and review steps; otherwise, it

20 Giorgio Colombo Ferruccio Mandorli
lets the designer to decide weather to proceed to design a new component
with that code, or to investigate into the PDM to find a similar component.
The coding system adopted was similar to that one used in the involved com-
pany but simplified according to the nature of DA prototype. In fact, the company
uses a coding system that describes very particularly the components, reaching a
very high level of description. Anyway, the authors think that using a more simpli-
fied coding system does not affect the validation of the prototype.
The validation phase with various case studies and the direct involvement of
SMEs’ technical staff permitted to verify the effectiveness of the implemented DA
application. Meaningful result has been the reduction of the product development
time. Typically, within the SME that provided the study case of the heat exchang-
er, the product design requires about 30 hours of an expert designer. This devel-
opment time is mainly due to the frequent iterative cycles required by some steps
of the design process even if they use electronic datasheets for some tasks. In-
stead, the developed application takes about 15 minutes ensuring also the respects
of standards and design rules, thus quality of the product.
Fig. 4.15 Simplified UML class diagram of the rotor group
.

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 21
Fig. 4.16 Operations to configure the product with new parts storing
Fig. 4.17 User Interface for the mixer DA application

22 Giorgio Colombo Ferruccio Mandorli
Fig. 4.18 Application functioning scheme
4.6 Discussion and Conclusions
This chapter has presented authors’ contribution to the domain of MDA; their
experiences started from first examples based on parametric CAD integrated with
programming languages or calculus tools to those based on an O-O approach. In
last two decades more interesting contributions to real industrial MDA are refera-
ble to these two methodologies. Other tools have been tested, such as rule based
systems or soft computing, but their relevance in industrial context was very light.
Our experience demonstrates that if the basic methods are not changed, the com-
plexity of the applications is really increased. We highlighted the key issues of the
current MDA applications: product and process modelling, PLM integration,
knowledge sharing and reuse.
CAD and O-O approaches focus on product modelling. We think that this is a
good approach for products without high complexity that do not require numerous
and articulated tasks to reach the final solution. In fact, only in these cases it is
easy to embed design process in parameters and methods of the objects represent-
ing the parts. For complex situations, for example integrating numerical simula-
tion, the design process is the kernel of the MDA application. We are not saying
that the product model is not necessary, but it is necessary to represent design pro-
cess to accurately control design choices.
The problems of integration within PLM/PDM solutions, sharing and reusing
knowledge are strictly related and very complex and we think that MDA method-
ologies must play the fundamental game for its own future. Practically, all the

4 Evolution in Mechanical Design Automation and Engineering Knowledge Management 23
MDA applications we have developed are “black box”, i.e., they encapsulate all
knowledge necessary to solve specific problems. But a lot of this knowledge can
be used in other situations; for example, a procedure to choice rolling bearing is a
very general procedure for engineering design. Thus, the problems are: i) how to
represent this procedure and ii) where store it to make it accessible to all people
that have necessity to use. This means that the engineering knowledge must be
stored in PLM/PDM systems; however today these systems do not provide func-
tions and methods to do it. In our works we investigated methods to inquire in au-
tomatic way a product data base to retrieve parts to implement a function and store
reusable knowledge for different MDA applications but in the same industrial con-
text and not in general way.
Finally, we think that MDA is an important tool to improve and innovate de-
sign procedures, both in big companies and SMEs. At present, MDA is accessible
also to SMES in terms of human resources, ICT tools investments and scien-
tific/technical competences. Moreover, EKM related to product and process is a
powerful tool to disseminate best practices within company departments making
people aware of company intellectual property and at the same time a way to take
care of the company knowledge and know-how.
References
1. Choi J C, Kim C (2001) A Compact and Practical CAD/CAM System for the Blanking or
Piercing of Irregular Shaped-sheet Metal Products for Progressive Working. Journal of Ma-
terials Processing Technology, 110:36-46
2. Colombo G, Ferretti D and Cugini U (1989) How to Represent Design Rules in a Paramet-
ric CAD System. In proceedings of International Symposium on Advanced Geometric
Modelling for Engineering Applications, Berlin, Germany
3. La Rocca G, Krakers L, van Tooren M J L (2002) Development of an ICAD generative
model for blended wing body aircraft design. In proceedings 9th Symposium on multidisci-
plinary analysis and optimization, AIAA/ISSMO, Atlanta, USA
4. Craig B, Chapman M P (2001) The Application of a Knowledge Based Engineering Ap-
proach to the Rapid Design and Analysis of an Automotive Structure. Advances in Engi-
neering Software, Elsevier, 32:903-912
5. Stokes M (2001) Managing Engineering Knowledge – MOKA: Methodology for
Knowledge Based Engineering Applications. Professional Engineering Publishing, London
and Bury St Edmunds, UK
6. Rosenfeld L W (1995) Solid Modeling and Knowledge-Based Engineering. Handbook of
Solid Modeling, Eds. Donald E. LaCourse, McGraw-Hill Inc, New York, USA
7. Colombo G, Cugini U and Mandorli F (1992) An Example of Knowledge Representation in
Mechanical Design Using the Object-Oriented Methodology. In proceedings of The 1992
European Simulation Symposium, Dresden, Germany
8. Coad P, Yourdon E (1990) Object-Oriented Analysis. Yourdon Press Computing Series,
Prentice Hall
9. Colombo G, and Cugini U (1992) Knowledge Aided Design: Requirements, Systems and
Applications. Revue Internationale de CFAO e d’Infographie, 7:3:293-309

24 Giorgio Colombo Ferruccio Mandorli
10. Germani M and Mandorli F (2004) Self-configuring components approach to product vari-
ant development. AIEDAM Special Issue: Platform Product Development for Mass Cus-
tomization, 18:1:41-54
11. Raffaeli R, Germani M, Graziosi S, Mandorli F (2007) Development of a Multilayer
Change Propagation Tool for modular product. In proceedings of the ICED 2007, 16th In-
ternational Conference on Engineering Design , Paris, France
12. Mandorli F, Rizzi C, Susca L, Cugini U (2001) How to Implement Feature-Based Applica-
tions using KBE Technology. In poceedings of the IFIP - FEATS 2001 International Con-
ference on Feature Modeling and Advanced Design-For-The-Life-Cycle Systems, Valenci-
ennes, France
13. Raffaeli R, Alfaro D, Germani M, Mandorli F, Montiel E (2006) Innovative Design Auto-
mation Technologies for Corrective Shoes Development. In proceedings of the 9th Interna-
tional Design Conference DESIGN 2006, Edited by D. Marjanovic, Dubrovnik, Croatia
14. Germani M, Mandorli F, Otto H E (2003) Product families: a cost estimation tool to support
the configuration of the solution phase. In proceedings of the ICED 2003, 14th International
Conference on Engineering Design, A Folkeson, K Gralen, M. Norell, U. Sellgren Editors,
Published by Design Society, Stockholm, Sweeden
15. Susca L, Mandorli F, Rizzi C and Cugini U (2000) Racing Car Design Using Knowledge
Aided Engineering. AIEDAM, Cambridge University Press, 14:3:235-249
16. Colombo G, Pugliese D and Rizzi C (2008) Developing DA Applications in SMEs Indus-
trial Context. In proceedings of IFIP-CAI 2008, Milan, Italy
17. Colombo G, Pugliese D and Saturno Spurio M (2007) About the Integration between KBE
and PLM. In proceedings of CIRP-LCE 2007, Springer Verlag, Tokyo, Japan
18. Colombo G, Facoetti G, Gabbiadini S, Rizzi C (2010) Virtual Configuration of Lower Limb
Prosthesis. In proceedings of the ASME 2010 International Mechanical Engineering Con-
gress & Exposition IMECE2010, Vancouver, British Columbia, Canada
19. Colombo G and Pugliese D (2008) Improving Product Design by Using Design Automa-
tion. In Proceedings of TMCE 2008,Edited by I Horvath and Z Rusak, Izmir, Turkey
20. Eriksson H, Penker M (2000) Business modeling with UML. John Wiley and Sons
21. www.idef.com
22. Forster J, Arana I & Fothergill P (1996) Re-design knowledge representation with
DEKLARE. In proceedings of KEML'96 6th Workshop on Knowledge Engineering: Meth-
ods & Languages, C Pierret-Golbreich, D Fensel, E Motta and M Willems editors, Paris,
France
23. Raffaeli R, Mengoni M, Germani M, Mandorli F (2009) An Approach to Support the Im-
plementation of Product Configuration Tools. In proceedings of the ASME International
Design Engineering Technical Conferences & Computers and Information in Engineering
Conference, IDETC/CIE 2009, San Diego, California, USA
24. Colombo G, Rizzi C, Scotto M (2010) A knowledge based application to enhance fired
heater design. In proceedings of IDMME - Virtual Concept 2010, Bordeaux, France
25. ANSI/API STANDARD 560/ISO 13705. Petroleum, petrochemical and natural gas indus-
tries – Fired heaters for general refinery service. Fourth edition, August 2007
26. ANSI/API STANDARD 530/ISO 13704. Petroleum, petrochemical and natural gas indus-
tries – Calculation of heater-tube thickness in petroleum refineries. Sixth edition, Septem-
ber 2008