Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

This article was downloaded by: [York University Libraries]

On: 12 August 2014, At: 09:29


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Journal of Engineering Design


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/cjen20

Semantic Modelling of Engineering Designs


a b b b
A. TALEB-BENDIAB , VINCENT OH , IAN SOMMERVILLE & MICHAEL FRENCH
a
Department of Mechanical Engineering , Manchester Metropolitan University , Manchester,
M1 5GD
b
School of Engineering, Computing and Mathematical Sciences, Lancaster University ,
Lancaster, LA1 4YR, UK
Published online: 23 Mar 2007.

To cite this article: A. TALEB-BENDIAB , VINCENT OH , IAN SOMMERVILLE & MICHAEL FRENCH (1993) Semantic Modelling of
Engineering Designs, Journal of Engineering Design, 4:1, 41-54, DOI: 10.1080/09544829308914771

To link to this article: http://dx.doi.org/10.1080/09544829308914771

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Journal of Engineering Design, Vol. 4, No.1, 1993 41

Semantic Modelling of Engineering Designs

A. TALEB-BENDIAB, VINCENT OH, IAN SOMMERVILLE & MICHAEL


FRENCH
Downloaded by [York University Libraries] at 09:29 12 August 2014

SUMMARY The entity-relationship (E-R) approach is a widely used technique for the
semantic modelling of commercial database systems. This paper demonstrates that the E-
R method can also be applied in the modelling of static design knowledge for knowledge-
based engineering design systems. It presents the basic definitions of the E-R data
modelling method and describes some extensions which are useful to model the function-
ality of system components. It demonstrates the use of this method to produce a generic
model of a centrifugal hydraulic pump, which is being used in the implementation of an
object-oriented intelligent CAD system. Finally, conclusions on the suitability of the
entity-relationship for engineering knowledge representation are drawn.

1. Introduction
Computer-aided design (CAD) systems are becoming an indispensable tool in mechan-
ical engineering design. The use of CAD systems has led to a considerable increase in
productivity in drafting, stress analysis, etc. However, the current generation of CAD
systems is mostly concerned with automating the drafting process with limited design
checking. They suffer from several deficiencies as far as providing support for the
design process is concerned [I]:
• Their user interface does not trap designer errors as they are made.
• They provide limited support for design problem solving.
• Different design tools such as drafting support and modelling tools are poorly
integrated.
These deficiencies all stem from the fact that most CAD systems do not provide any
means of providing information about the semantics or meaning of a design [2, 3].
There is no way for the system to relate the physical design representation to the
logical entity being designed.
The meaning of data may be expressed using a technique called entity-relationship
semantic data modelling which sets out the types of entity which are of interest to the
system and the relationships which may exist between these entity types [4, 5]. This
entity-relationship (E-R) method of data modelling is widely used to derive the
schemas for commercial database systems.
Providing designers with a means to define design semantics allows the creation of
an active rather than a passive CAD system. Our research is concerned with exploring
A. Taleb-Bendiab, Department of Mechanical Engineering, Manchester Metropolitan University, Man-
chester Ml SGD. Vincent Db, Ian Sommerville & Michael French, School of Engineering, Computing and
Mathematical Sciences, Lancaster University, Lancaster LAI 4YR, UK.
42 A. Taleb-Bendiab et al.

how a CAD system can automatically propose design improvements, and E-R
modelling is central to this system, as a means of defining semantics. The system
which we are developing is concerned with modelling and improving axisymmetric
assemblies such as hydraulic pumps. Using a semantic model, designs are checked as
they are input, and knowledge modules propose design improvements which are
validated using the semantic model.
In the remainder of this paper, we introduce the concepts of the E-R approach to
semantic data modelling and discuss extensions required for modelling engineering
designs. We then explain the E-R technique by means of a centrifugal hydraulic pump
example, and discuss the application of the technique in intelligent CAD systems.

2. Entity-Relationship Modelling
Downloaded by [York University Libraries] at 09:29 12 August 2014

E-R data modelling is a conceptual data modelling method for databases and informa-
tion systems design. It captures a system's generic entities and their relationships.
Traditionally, it is mainly used by software designers as an analytical tool to help
understand the entities within an enterprise and their interrelationships. E-R models
may be represented graphically in a form that is easy to understand, yet powerful
enough to model real problems.
The components of an E-R model are:
• Entities. An entity is a set of objects which have some common properties.
Examples of entities might be shafts, casings, bearings, etc.
• Attributes. An attribute is a characteristic of an entity such as its length, material,
etc.
• Relationships. A relationship is an association between two entities such as 'pan-
of' or 'enclosed-by'.
The graphical representation of an E-R model is called an entity-relationship diagram
(ERD). Several different graphical representations for ERDs have been invented
[6-8]. These are approximately equivalent and there is no standard representation. To
cope with the complexity inherent in engineering designs, we use an extended notation
for E-R description which borrows characteristics from various other notations.
In this section, we introduce, using examples, the basic components of E-R
modelling. The class of assembly we are currently studying is hydraulic pumps so the
examples in this section are drawn from that domain. However, the method is
applicable to any class of mechanical or electrical systems where components have
generic characteristics and participate in relationships with other components.
E- R diagrams are graphs where entities or components are represented as nodes
and their relationships (connectivities) as arcs. Figure I illustrates part of an E-R
model of a centrifugal hydraulic pump. The meaning of the entities should be obvious
and we discuss the relationship types together with the explanation of symbols used
later in this paper (Sections 2.2 and 2.3).

2.1 Entities
An entity represents a set of real or abstract objects which have common attitudes or
characteristis, which can exist and which can (somehow) be distinguished from other
entities. An entity is represented diagrammatically by a box with a name, where the
name is in the singular and shown in uppercase letters. The entity name must be one
that represents a type or class of thing, not a specific instance.
Semantic Modelling of Engineering Designs 43

Ml;CHANlCAL SEAL

···
·,,
Downloaded by [York University Libraries] at 09:29 12 August 2014

___________________________________________________________ 1

FIG. I. A section of the global ERD of a centrifugal hydraulic pump.

E-R diagrams may be presented at different levels of detail, in Fig. I, for example,
it is clear that the identified entities may themselves be complex components made up
of several parts. We therefore use structured E-R diagrams where the semantic model
of each part may be represented as a diagram in its own right. For example, Fig. 2
shows the semantic model of the IMPELLER entity used in Fig. I. In Fig. 2, we add
more detail and show the attributes of the entities.
In Fig. 2 the IMPELLER entity is delimited by a dashed line box, which means that
the entities enclosed by the box are 'parts of' the structured entity whose name appears
in the top-left corner of the box. The relationships in which the structured entity
participates are shown as arrows leaving the enclosing box. These should generally be
consistent with the relationships shown on the higher-level E-R diagram.
Entities are described in terms of attributes or properties. For instance, as shown in

KEY

IMPELLER I --------------------------------------

SHAFT, IMPELLER SEGMENT LOCKING DEVICE WEAR RING

FIG. 2. An E-R representation of the impeller entity.


44 A. Taleb-Bendiab et al.

Fig. 2, an IMPELLER WHEEL is a type of mechanical entity, which has for example a
set of associated attributes called dia (diameter), width and type. The type attribute
is used to classify the impeller wheel (for example axial or centrifugal), whereas the
width and dia attributes qualify the entity.
We emphasize that the entity concept used here is different from that used in most
conventional CAD drafting systems like AUTOCAD. The entity concept used in
conventional drafting systems is meant to represent graphical elements rather than the
semantic content of real world engineering object components, which the E-R ap-
proach described here focuses on. Our use of the entity concept here is at a much
higher level of abstraction. This is one reason why most present systems are not true
design aids since their support in the design process is only superficial. They support
the visualization aspect of the design very well and provide quick turnaround times for
drawings but they lack any semantic representation of the actual engineering compo-
Downloaded by [York University Libraries] at 09:29 12 August 2014

nents necessary for reasoning in design.

2.1.1 Attributes. An attribute represents properties of descriptions associated with an


entity. It represents any detail that serves to qualify, identify and/or classify an entity.
Attributes are identified by meaningful names such as diameter, length, etc. In an E-
R diagram, attributes are represented as an ellipse which includes the attribute's name
in lowercase letters.
Each attribute has an associated type and value. An attribute type constrains the
attribute value to be a numeric value, a character string, etc. An attribute value is
simply the value associated with a particular instance of an entity.
An entity SOLT for example might have some of its attributes described as follows:

• nominal diameter: type = fixed point number;


• length: type = fixed point number;
• type of thread: type = string (ISO, UNF, BA);
• pitch: type = fixed point number;
• head type: type = string (hexagon, countersunk, pan, recessed flange);
• length of thread: type = fixed point number.
Notice that no specific values are included in the specification of the entity attributes.
What follows the' =' sign serves to qualify the type of value the attribute will hold.
The items within the brackets, e.g. ISO, UNF, BA, form an enumerated set of string
items. This simply means that the attribute will take on a value equivalent to one of
the item members in the enumerated set. A fixed point number can also have its own
associated enumerated set or range, e.g, 10.0-100.0. Attribute values distinguish one
instance of an entity from another. For example, a specific instance of a bolt may have
the following attribute values:

• nominal diameter = 10.0 mm;


• length=50.0 mm;
• type of thread = ISO;
• pitch= 1.5 mm;
• head type = hexagon;
• length of thread = 26.0 mm.
So far the attributes described here specify the physical properties of entities. Function
attributes are another kind of attribute, and they are discussedin Section 3.2.
Semantic Modelling of Engineering Designs 45

2.2 Relationships
A relationship is a named binary association between two entities (or an entity and
itself). It is represented in our semantic modelling notation as an arrowed line
connecting two entities. A relationship has the following attributes:
• A name which is placed in square brackets at the middle of the arrowed line.
• A cardinality to show the number of associations between the related entities. The
cardinality can be one to one, one to many, many to many or one to an exact
number. For example, in Fig. 2, an IMPELLER WHEEL may have many blades so
the cardinality of the relationship between the IMPELLER WHEEL and the
IMPELLER BLADE entities is I :n. The cardinality may be omitted if it is 1: l.
The notion of relationships in E-R modelling is a simple one which is designed to cope
with the types of relationship which can exist between entities in a computer database.
Downloaded by [York University Libraries] at 09:29 12 August 2014

To represent engineering assemblies, we have extended the standard E-R model in two
ways:
(1) Relationships may be optional. In some assemblies, a relationship may exist
between two components; in other assemblies of the same class but with a
different configuration, there may be no requirement for a relationship. Manda-
tory relationships (relationships which must always exist) are represented by a
full line with arrows at each end. Optional relationships (which may but need
not exist) are represented as a dashed line.
(2) Multiple relationships of the same type between entities are supported. Because
of the geometry of an assembly, there may be several instances of the same
relationship between two components. For example, in Fig. 1, the relationship
cylindrical abutment [Cal exists four times between the entity SHAFT and the
MECHANICAL SEAL. Each relationship may therefore have an associated
multiplicity which indicates the number of instances of the relationship.
Generally, the relationship multiplicity is not shown on the ERDs if it is equal
to one. For example, Fig. 1 shows that a number of BEARING entities (n) may
participate in one cylindrical abutment [Ca] relationship with a single SHAFT
entity.
Relationships are used to represent different types of associations between entities.
Examples of these different types are:
• Generalizability/specialization. The kind.Lof relationship is used to represent
entity generalization. For example, a deep groove ball bearing is a kind.Lof
bearing. kind.Lof relationships are implicit when a set of entities is enclosed by a
dashed line box.
• Aggregation. The part-of relationship [9) indicates aggregation. In Fig. 2 the
IMPELLER BLADE is a part.Lof an IMPELLER WHEEL (as in a fan with
inserted blades).
• Association. This class encompasses all the domain-dependent relationships, such
as (for this work) physical and spatial relationships. These are particularly
important for semantic modelling and we discuss them in more detail below.

2.3 Classes of Engineering Relationships


Each application domain has its own particular types of relationships that must be
defined when building a semantic model. In mechanical engineering design, the
46 A. Taleb-Bendiab et al.
majority of relationships are physical relationships. For instance, WEAR RING
ABUTMENT is axially connected [Ac] to the IMPELLER WHEEL.
We classify mechanical relationships into 10 classes (Fig. 3). These relationships
represent a physical type of association between design entities. The integrity of the
design depends on this physical association being 'correct', and the explicit identifica-
tion of physical relationships simplifies the problem of identifying 'design rules'. Some
of these 'design rules' are shown in Fig. 3.

Class Schematic representation Symbol Rules Example

.. MaQl11 tude of the forces along the


x-axis for both objects are equal Fastener
against a
Faclnc;r V;;#@ Fa .. Both objects have a contact face
.. No relative notion. t .e. smne VX spaclno rln
Downloaded by [York University Libraries] at 09:29 12 August 2014

.. MaQrlitude of the forces along the JI:- Interference

~
and v-exes for both objects are equal ht of a
Cylindrical
Ca .. sane axis and equal naninal dterrerers bearing on
ahutnent its housino
.. No relat iva not ion vx. SX and Vy

... .- '" Hagnl tude of the forces alono the


y-axis for both objects are equal Piston and

~
Sliding Sf " sane axis cylinder
joint .. Relative notion VX and Sx

.. HaQl11tude of the forces 81ano the

~
_dod Key.
6l x-axre for both objects are equal
... No relative notion Vx and Sx clrcl1ps

.. MaQnl tude of the forces alono the


contact
~ C' y-axis for roth objects are equal
.. No relative not Ion Vy
VIce

SlldlnQ
... .- Sl
.. HaQl11 tude of the forces etonc the Lathe centre
~
y-exi s for both objects are equal post
• Relative rroUon VX

• 6aIre part
Axially Stepped
connected SSJfM- A< •

Ga:re
6M'e
axis
enerQY stft,te shaft

FIG. 3. A section of the table of classes of physical relationships.

A more detailed description of the knowledge structure of the relationships and


design rules is given in (10], where in an object-oriented fashion, classes of mechanical
relationships are structured as objects. Each relationship contains a set of protocols,
which is a collection of methods (operations) an object can perform. For instance,
setting a cylindrical abutment [Cal relationship between two entities (BEARING and
HOUSING) will invoke a plan of action.
plan:
name: cylindrical abutment
condition: user-initiated,
checkObjectsRound
planSchema:
actionSchema: haveSameAxis,
actionSchema: equalNominalDiameters
actionSchema: equalMagnitudeForces.
knowledgeRecovery: promptUser
This plan is invoked only if the condition is valid, that is, the user initiates the plan
and the two selected objects are round (axisymmetric). The planSchema is a
Semantic Modelling of Engineering Designs 47

conjunction of actions (or actionSchema) to be taken, such as checking that the


related objects have a common axis and that their nominal diameters and the
magnitude of the forces along both x- and y-axes are equal (Fig. 3).
A plan fails if at least one of the actions fails: in such a case a knowledge recovery
is suggested to the user to recover from the failure.
These design rules are embedded as behaviour of the relationship class and they
enable an automatic checking of design consistency.

2.4 Attribute Relationships


Relationships may exist between specific attributes entities and these are particularly
important for checking the consistency of assemblies represented as E-R models.
However, attribute relationships are not usually shown explicitly on ERDs because of
Downloaded by [York University Libraries] at 09:29 12 August 2014

the diagram clutter which would result. Rather, an ERD describing an entity should be
annotated with details of the specified attribute relationships as shown in Fig. 4.

INNER RING: dia ~ SHAff: STEP BEARING: dia

INNER RING: width --0--- SHAff: STEP BEARING: length

INNER RING: FILLET: radius ~ SHAFf: ALLET: radius

alITER RING: dia ~ HOUSING: dia

OUfER RING: width --0--- HOUSING: width

OlITER RING: FILLET: radius ~ HOUSING: FILLET: radius

FIG. 4. BEARING attribute relationship diagram.

An attribute relationship is a constraint, which governs the structural and behav-


ioural correctness of the considered relationship. For instance, the integrity of the
cylindrical abutment relationship existing between an instance of HUB and an instance
of SHAFT (Figs I and 2) is assumed by the equality constraint between the entities'
diameters:
IMPELLER: Hub: dia=SHAFT: dia
The colon : is used to represent both the attribute ownership and the entity structure
hierarchy. Thus in IMPELLER: HUB: dia, dia is an attribute of the entity HUB, which
in turn is a subcomponent (or feature) of the IMPELLER entity.
The diameter of the HUB must be the same as the diameter of the SHAFT
otherwise a design error will have occurred. We have yet to address the problem of
tolerances in our work because it does not affect it but we believe that when necessary
it can be tackled by redefining the' =' operator to be an inexact rather than an exact
operator with deltas used for tolerance specification. For example, we could write the
above relationship as follows:
IMPELLER: HUB: dia [~- =0; ~+ =T]=SHAFT: dia
This means that the diameter of the HUB may exceed the diameter of the SHAFT by a
maximum amount T but it must not be smaller than the SHAFT diameter; in other
words, the fit lies between 0 and + T.
It is generally the case that many different attribute relationships must be defined
48 A. Taleb-Bendiab et al.

to check the consistency of an engineering assembly. For example, Fig. 4 shows a


subset of the attribute relationships which may exist between a BEARING entity and
associated SHAFT and HOUSING entities.
These constraints are implied by the attribute relationships and the validity of the
attribute relationship depends on the satisfaction of the violated constraints. A network
of such constraints <ian form the underlying framework for an automatic constraint
propagation mechanism. The use of a semantic model helps engineers to identify and
define these constraints systematically.

3. Semantic Modelling of Engineering Assemblies


Like information systems, engineering assemblies are made up of a set of components
Downloaded by [York University Libraries] at 09:29 12 August 2014

which interchange some form of information, material, forces and/or energy input/
output arguments [11]. An engineering assembly may be viewed as an organization of
entities with more or less strict relationships existing between them. For instance, a
SHAFT (entity) can be 'supported by' (relationship) a BEARING (entity).
Developing semantic models of engineering assemblies has a number of advantages:
• These models can act as a means of communication between different engineering
disciplines. As mechanical systems develop to include electronic and software-
driven components as in the field of mechatronics, it is increasingly important
that mechanical, electrical and software engineers have a common language for
discussing a design. Conventional engineering drawings are not a suitable com-
mon language as relationships must be inferred by the engineer. While mechanical
engineers are trained in such inference, engineers from other disciplines cannot
fully understand engineering drawings.
• For CAD system designers, analysis of a set of semantic models of different
assemblies allows them to design effective data structures for design representa-
tion. As CAD systems develop, simple data structures which are derived from
geometrical relationships will be increasingly inefficient for the type of design
processing which is required. Effective design checking cannot simply be based on
geometric representations.
• A model of an assembly assists with the systematic derivation of design checks
which should be applied. These may be applied automatically by an intelligent
CAD system. However, even when no automated design checking is supported,
the designer may be given a 'checklist' of tests which should be applied to the
design.
• The model may be used as a framework for acquiring knowledge about the design
of specific types of assembly. As CAD systems evolve to include more 'intelli-
gence', the knowledge which is embedded in them must be captured from human
expert designers. The structured nature of a semantic model allows the knowl-
edge to be captured incrementally. Furthermore, the model may be analyzed and
'missing knowledge' may be identified.
• The models can be used in training. They allow complex assemblies to be
discussed at an abstract level and help structure explanations of why particular
components are used in a design and how these interact with other components.
Semantic modelling of assemblies has been explored by other researchers who have
used semantic nets to represent engineering systems. Semantic nets were originally
developed for natural language processing, that is, to represent the meanings of English
Semantic Modelling of Engineering Designs 49

words [4, 5] and were later extended to improve on the expressiveness of the
representation [12-14].
Semantic nets are generally used as a knowledge representation scheme in artificial
intelligence (AI). They have been used as a declarative knowledge representation
approach, and a basis to describe the relationship existing between the constituent
components of a design artefact. In the systems known as ICAD [15] and Concept
Modeller [9], a semantic network is constructed during the user's description of the
designed assembly to the system. The obtained concept model is used for consistency
checking every time changes are made to components.
As a modelling tool, semantic nets tend to produce cluttered diagrams and this adds
complexity to the modelling process. We considered the use of semantic nets in our
work but rejected them because of the complexity of the networks required to
represent the types of assembly and the relationships which were of interest to us.
Downloaded by [York University Libraries] at 09:29 12 August 2014

Rather, our starting point for semantic modelling was the E- R method which we
extended to support the particular requirements of the mechanical engineering applica-
tion domain.
The extended E- R method which we use can describe two aspects of an engineer-
ing assembly:

(I) Its structure. The ERD shows how components relate to each other.
(2) Its functionality. The ERD illustrates the functionality of each component.

An E-R model of an assembly is a static model so that this approach is not suitable for
modelling dynamic aspects of the behaviour of an assembly.

3.1 Structural Representation


An engineering assembly consists of components, connected to one another and to
external systems by relations. Each entity can be, in its turn, a component (subsystem)
of other structures (see Section 2.1.2 and [1]). The schema of components and the
aggregation type of relationship existing between the different design objects consti-
tute the structural representation.
Figure 5 illustrates the subpart representation and their relationships within the

BEARmGI------------------------------------, I

INNER RING

[Sl, C]

ROILING ElEMENT
[Sl, C]

GLAND HOUSING

FIG. 5. An E-R representation of the BEARING entity structure.


50 A. Taleb-Bendiab et al.

BEARING entity, in addition to its structural relationships with other structured


components within the hydraulic pump assembly.

3.2 Functional Representations


The functional representation of a design object can be illustrated by the association
existing between the design object and the set of operators (or functions) it can assume
[16].
Formally, an artefact's function can be defined as a relationship between the input
and the output arguments [II].
O=f(I), where 0 represents the output arguments and
I represents the input arguments
Downloaded by [York University Libraries] at 09:29 12 August 2014

To support the functional specification of an entity, the designer may introduce a


special entity attribute called Function, By convention, we do not use this name for
any other attributes. This attribute is represented as a 3-tuple 11, Fn, 01, where I is a
set of function input names, Fn is the name of the function and 0 is a set of function
output names. The meanings associated with all of these names must, of course, be
described separately. Examples of Function attributes are:
IlFx, Fy. s-. TxL channel. IFx, Fy, Sx, TxH
IlFx, Fy, Sx, TxL absorb, IFx. FyH
We illustrate these on an E-R diagram by showing inputs and outputs to the entity
attribute as shown in Fig. 6. The function attribute is distinguished by the input and
output arrows.

{F" Fy, S'. T,)

~~~ INNER RING



[Px. Fy, Sx. Tx)

FIG. 6. E-R extensions for functional specification.

This notion is further illustrated in Fig. 7 which shows the functional model of a
bearing. The INNER RING has a cylindrical abutment [Cal relationship with subpart
STEP_BEARING of the entity SHAFT. The STEP_BEARING object is that
segment of the SHAFT which is associated with the BEARING entity. The STEP_
BEARING's output arguments which represent respectively force along x- and y-axes
(Fx and Fy), the rotational speed (Sx) and the torque along the x-axis (Tx) are the
same as the INNER RING input arguments. In this example the bearing friction torque
is ignored because we considered it to be negligible compared to the other forces, and
its omission does not affect the E- R technique as an engineering design modelling
approach.
The INNER RING entity is applied to 'channel' forces and/or speed, so its output
arguments are equal to the input arguments [17, 18]. The ROLLING ELEMENT
entities only allow forces to be transmitted to the OUTER RING, thus the function is
'absorb' rotational speed about the x-axis. The OUTER RING has the same function as
Semantic Modelling of Engineering Designs 51

SHAFf: STEP_BEARlNG

I
BEARlNGI- - - - - - - - - - - - - - -

I INNER RING I ~
,
~-
{Px.Py. Sx}
_ •• _. -;

:
.
Ifx. fy, Sx}

ROUlNG EUMENf
Downloaded by [York University Libraries] at 09:29 12 August 2014

Ifx, fYI

I OUIERRING
t
I---y
: ~ )~X~f\ :
GLAND HOUSING

FIG. 7. A functional model of a BEARING.

the INNER RING, that is, to 'channel' the same state of input argument to other
entities, the HOUSING and GLAND entities.
This representation helps in the modelling of engineering assemblies for intelligent
CAD implementations' [I). In other work, functional modelling of this kind is already
being used, and this development could be used also in these design semantic models.

4. Applying Semantic Models in Intelligent CAD systems


As discussed in the introduction to this paper, the objective of our work is to develop
an intelligent CAD system which can automatically propose improvements to a
particular class of mechanical engineering designs. We originally investigated the use
of semantic models for assembly modelling because of the need to identify systemati-
cally design integrity checks which should be applied. The E-R modelling technique
allowed us to analyze graphically the component objects that make up the product
design and their relationships with one another; identify the physical attributes of the
components and their relationship with other attributes; and make explicit any implicit
relationships between components and their attributes that are normally taken for
granted in the minds of engineers and designers. The whole process of semantic
modelling via the E-R technique needs to be done iteratively and not as a one-shot
kind of activity. The resulting diagrams constitute the semantic model represented
graphically rather than in formal terms.
From here it was a simple matter of designing the data structures that encapsulate
the entities and their attributes, and the class of relationships. As defined in the E- R
diagrams, the kind of relationship established between two components decides the
relationships between their attributes. This in essence establishes a constraint between
52 A. Taleb-Bendiab et al.

the two components. For example, when a 'cylindrically abutted' relationship is


established between a shaft and bearing, it means that the nominal inner diameter of
the bearing has to be equal to the nominal shaft's diameter. This is explicitly
represented in the ERDs and implemented as a constraint in the computer. Fundamen-
tal laws of physics and design rules as embodied in the classes of engineering
relationships (Section 2.3) can be implemented in the same way. For instance, a shaft
entity can be viewed as a collection of shaft segments of differing diameters related to
the former by a 'part of' relationship. Each shaft segment is further related to each
other by a 'mating' relationship. So when a torque is applied at one end, this torque is
transmitted throughout the shaft via the 'mating' relationship that embodies a funda-
mental law of physics determining how energy or force should be transmitted from one
object to another.
The design improvement system uses embedded design knowledge to propose
Downloaded by [York University Libraries] at 09:29 12 August 2014

improvements to an existing design. Conceivably, the proposed designs will be


incorrect and it is important that these incorrect designs are identified as early as
possible in the improvement process. Otherwise, a great deal of computational effort
may be wasted in computing design transformations which are clearly not improve-
ments. Thus, a key module component of our intelligent CAD system is a checking
module which, at all stages, ensures that the model of the design maintained by the
system is free of errors.
We have used the semantic model of hydraulic pumps to formulate a set of
integrity rules, which may be used to check the correctness of a specific pump design.
The integrity rules are based on both the use of implicit constraints (Section 2.5)
existing between the attributes of the related object, and the behavioural representation
of a relationship (Section 2.4). A relationship may only exist if all the checks are valid,
and explict constraints (representing attribute relationship) produced from the associa-
tion of entities in an assembly are also used as a means of checking designs. Using the
design model, missing entities and relationships can be brought to the attention of the
user and actions suggested.
As well as checking the integrity of design improvements, the inclusion of a
detailed design checking facility also traps designer errors as the initial design which is
to be improved is input to the system. While the designer creates his or her design by
organizing pump components and giving values to their attributes, the design checker
continually monitors the design model. As soon as a design error is discovered, it is
brought to the engineer's attention so that it may be immediately corrected. Thus, we
avoid error propagation and unnecessary design work based on some initially incorrect
system.
The organization, structure and knowledge model of the design improvement
system will be described in a future publication. The current system prototype
includes the design checking facilities as already discussed and can propose simple
improvements to axisymmetric assemblies.

5. Conclusions
E- R modelling is a technique, mainly used by database designers, to develop the
conceptual understanding (conceptual schema) of the system to be modelled. We have
shown that its utility does not lie solely in the domain of information systems. The
method has proved to be extremely useful for acquIsition of static knowledge for the
design of a knowledge-based design system and as a basis for object-oriented modelling
of engineering design objects.
Semantic Modelling of Engineering Designs 53

We have shown how existing techniques of semantic modelling may be extended to


allow the modelling of the structure and function of complex mechanical engineering
assemblies. Such models contain an explicit description of a set of entities and the
relationships between them, along with the integrity rules (or constraints).
The advantages of building such a semantic model are four-fold:
(1) It exposes the essential relationships between components in an assembly and
allows the engineer to discover systematically the constraints which govern the
design of the assembly.
(2) It provides the basis for the implementation of automatic design checking in an
intelligent CAD system.
(3) It provides guidance to the intelligent CAD system designer as to the most
effective data structures which should be used to represent the internal model
of the design.
Downloaded by [York University Libraries] at 09:29 12 August 2014

(4) It provides a framework for knowledge acquisition from experts. The problems
of knowledge acquisition are recognized as a key bottle-neck in the develop-
ment of intelligent systems.
The disadvantage of using semantic modelling is that the method is not familiar to
most engineers. However, it is easy to learn and most engineers can understand (if not
actually create) a semantic model of an assembly after only a few minutes instruction.

Acknowledgements
The research described here was funded by the ACME Directorate of the UK Science
and Engineering Research Council. We would like to thank our industrial collaborators
British Aerospace pic, Gilbert Gilkes and Gordon Ltd, Angus Fire Ltd and Pafec Ltd
for their support.

REFERENCES
[1) TALAB-BENDIAB, A. (1990) The logic programming of conceptual design: selec-
tion processes by approximate matching, PhD Thesis, Mechanical Engineering
Department, Liverpool University, Liverpool.
[2) TOMIYAMA, T. & YOSHIKAWA, H. (1984) Requirements and principles for intelli-
gent CAD systems, in: GERO, J. S. (Ed.), IFIP WGS.2 Working Conference on
Knowledge Engineering in Computer-Aided Design, Budapest, pp. 1-24.
[3) OSHUGA, S. (1989) Toward intelligent CAD systems, CAD, 21 pp. 315-337.
[4) RAPHAEL, B. (1968) A computer program for semantic information retrieval, in:
MINSKY, M. (Ed.) Semantic Information Processing (Cambridge, MIT Press).
[5) QUILLIAN, R. (1968) Semantic memory, in: MINSKY, M. (Ed.), Semantic Infor-
mation Processing (Cambridge, MIT Press).
[6) CHEN, P.P. (1976) The entity-relationship model: toward a unified view of data,
ACM Transactions on Database Systems, 1, pp. 9-36.
[7) MARTIN, J. & MCCLURE, C. (1985) Diagramming Techniques for Analysts and
Programmers (New Jersey, Prentice-Hall).
[8) SHLAER, S. & MELLOR, S. (1985) Object-Oriented Systems Analysis: Modelling the
World in Data (New Jersey, Prentice-Hall).
[9) LUKAs, M.P. & POLLOCK, R.B. (1988) Automated design through artificial
intelligence techniques, Artificial Intelligence and Advanced Computer Technology
Conference, Long Beach, California, pp. 1-12.
54 A. Taleb-Bendiab et al.

[10] TALEB-BENDIAB, A. OH, V., SOMMERVILLE, I. & FRENCH, M.l (1992) Knowledge
representation for engineering design product improvement, 7th International
Conference on Applications of Artificial Intelligence in Engineering, ADEY, R. A.
(Ed.).
[11] PAHL, G. & BEITZ, P. (1984) Engineering Design, 1st edn (Berlin, Springer).
[12] WOODS, W.A.. (1975) What's in a link: foundations for semantic networks, in:
BOBROW, D. G. & COLLINS, A. M. (Eds) Representation and Understanding:
Studies in Cognitive Science (New York, Academic Press).
[13] SCHUBERT, L.K. (1987) Extending the expressive power of semantic networks,
Artificial Intelligence, 7, pp. 163-198.
[14] HENDRIX, G.G. (1977) Expanding the utility of semantic networks through
partitioning, Proceedings of the 4th International Joint Conference on Artificial
Intelligence.
Downloaded by [York University Libraries] at 09:29 12 August 2014

[15] MACILWAIN, C. (1988) CAD's fourth generation, The Engineer, 11, p. 33.
[16] TALEB-BENDIAB, A. & SHENTON, A.T. (1991) A conceptual design aid: selection
processes by approximate matching, In: HUBKA, V. (Ed.), International Confer-
ence on Engineering Design, Vol. 2, pp. 1042-1050 (Zurich, Heurista).
[17] KOLLER, R. (1985) Konstruktionslehre fur den Maschinenbau (Berlin, Springer).
[18] SHENTON, A.T., TALEB-BENDIAB, A. & CHEN, Y. (1989) Computer-aided con-
straint development systems for conceptual and embodiment engineering design,
3rd Eurographics Workshop on Intelligent CAD Systems, Amsterdam, pp. 65-97.

You might also like