Avgoustinov Nikolay. Modelling in Mechanical Engineering and Mechatronics

Подождите немного. Документ загружается.

192 4 Towards Better Product and Process Modelling

• can complete some simple self-maintenance activities;

• is capable of learning;

• can represent both static and dynamic data (i.e. product and process related).

Therefore, MCA not only relies on use of knowledge and intelligence, but re-

commends incorporating them directly where they are needed. A simplified ver-

sion of the algorithm used to create intelligent models is presented in Figure 4.13.

Observation and analysis

Recognition of recurring elements:

features, events, states and (re)actions

Giving appropriate names

to the recognized elements

Recognition of patterns of

recurring elements

Determining (sets of) elements

that can be factored out

Determining their attributes, possible states,

events to react to and essence of the reactions

Defining the interfaces of their models

Implementation, testing

Use, reuse

Figure 4.13.

The steps towards incorporation of intelligence into models, after Avgoustinov

and Bley (2006)

4.6 Inherencies of MCA 193

It is important to note that when following this approach, sooner or later –

depending on the model granularity (

i.e., on the models' average size) – name

collisions can occur. These are normal for bottom-up development and formally

could be avoided with careful choice of all names and suitable application of

namespaces. The real problem, though, is not how to avoid a name collision in the

technical sense, but how to ensure that in a (future) world with unlimited number

of models, exactly the desired model is unambiguously referred to. On the other

hand, the opposite problem is also possible: to know that a model with the desired

qualities exists, but to be unable to find it because its name is unknown. These two

problems will receive a satisfactory solution only with the development of

taxonomies and domain ontologies for the concerned domains. The introduction of

domain ontologies would also improve the development of the user interfaces, as

well as the human-to-human communication.

With respect to globalization and the international cooperation we consider also

very important the employment of thesauri. It would increase the comfort of use

and the probabilities for finding the searched. According to our approach in

Avgoustinov and Bley (2003) the intelligence is distributed within the process

model (in our case, of assembly and assembly planning) across several hierarchy

levels. The behaviour making the models appear intelligent is implemented on the

lowest level with the help of features, patterns and procedures, integrated into

autonomous intelligent entities (AIEs). On the upper levels the AIEs can be

combined in more and more sophisticated models, up to distributed intelligent

virtual enterprises. Very interesting possibilities are offered by the combined use of

(assembly) patterns and (assembly) features – they have many similarities, but also

many differences that can nicely complement each other.

Summing up, with the concepts for separation of authoring from use, separate

modelling, model organization according to Figure 4.6 and incorporation of

intelligence, the MCA offers much higher flexibility of the prepared models.

4.6.8 Cooperative Work and Distributed Authoring

The separation of authoring from use, together with the separate modelling, create

prerequisites for easier cooperation, since it makes no difference where the models

are created. When the cooperating authors (or modellers) are distributed in

different points of the world we can speak of

distributed authoring. An illustration

of how cooperation on the basis of distributed authoring can be organized is

presented in Figure 4.14. Note the difference between Internet and (manufacturer's)

intranet. The idea is that instead of using the conventional electronic catalogues it

would be much more convenient if the manufacturers of different equipment – in

this example workpieces, tools, machine tools, and fixtures – provide functional

models of their products to be used for testing and planning of the future

production. Of course, these models can involve some kind of protection against

theft of intelectual properties – e.g., authorisation of the use only after

identification, or even for a small fee. Nevertheless, the possibility to experiment

and play different scenarios with the whole equipment and prove that it is exactly

what is needed can enormously improve the productivity, the economical

efficiency and probably even the quality of the production yet during its planning.

And such way of cooperation could provide for much earlier and more intensive

feedback concerning the quality of the involved models and their respective

modellees.

194 4 Towards Better Product and Process Modelling

rocess engineer's workbenc

Manufacturer's Intranet

Designer's workbench

Manufacturer's Intranet

Interne

Manufacturer's Intranet

.......................

Server for

communication,

control and

integration

Web-services

Product Mode

Intelligent

tool models

ibrary

Intelligent

part models

ibrary

...

Software

componens

Workpieces

manufacturer

Design

features

Authoring

environment

Authoring

environment

Authoring

environment

Tools

manufacturer

Intelligent

ixture models

ibrary

Authoring

environment

Fixture

manufacturer

Intelligent

machine tool

odels librar

Machine Tools

manufacturer

Manufact.

features

Assembly

features

Functional

features

Models, cooperatin

over network and

epresented within

he same web-based

application

Figure 4.14.

A scheme for cooperative work over Internet, advanced from Avgoustinov

(1999) and Avgoustinov and Bley (2003)

4.6.9 Lean Modelling

As already mentioned in Section 4.6.1, MCA can be viewed as a new organisation

of the modelling and the resulting models. This new organisation could play for the

modelling the same role, which the concept of lean manufacturing, Womack and

Jones (1994), lean thinking, Womack and Jones (2003) and lean solutions,

Womack and Jones (2005) played during the last decades for the manufacturing:

minimising the waste and increasing the efficiency.

And if the approach is successful, it is not important, whether its name is lean

modelling, model centred approach or something else; important is to make the

next step on the way of continuous improvement.

4.7 Comparison of MCA and SCA 195

4.7 Comparison of MCA and SCA

A comparative assessment of the most important traits of the product and process

models, developed by means of the two approaches, is presented in Table 4.6 and

Table 4.8. The comparison is intended just to give some orientation and first

impressions about the MCA.

For SCA, the theoretical possibility for a certain parameter to be employed

(either 0 or 100%) and an estimation of how many CAx-systems are making use of

it really are given. The value 5% denotes that probably there is a CAx-system,

which makes use of this parameter, but it is not known to us.

Table 4.6. Modelling parameters (strategic and conceptual) in use by SCA and MCA (cf.

Table 4.4, UL part). Parameters with the same use in both SCA and MCA are not shown.

MCA

theoret.

possible

practice

theoret.

possible

100% 5% 100%

0% 0% 100%

100% 30% 100%

100% 60% 100%

5% 100%

100% 5% 100%

100% 10% 100%

100% 80% 100%

100% 50% 100%

100% 80% 100%

0% 100%

100% 75% 100%

100% 10% 100%

100% 70% 100%

100% 50% 100%

100% 70% 100%

100% 90% 100%

100% 75% 100%

100% 5% 100%

SCA

use of patterns

choice of appropriate

authoring tools

tool performance

quality of the delivered

models

tool lifetime

tool reliability

new versions backward

compatible

interoperability

low price of the tool

standardization

approp-

riate

hosting

long living hosts

high performance

low price of the host

new versions backward

compatible

Correlation

Modelling parameters

strategic & conceptual activities

separation of authoring from use

shifting the focus from use of tools

towards ade

uate models/results!

(modellee) interaction resemblance

parameterization

pursuing optimal granularity

platform-independent model design

integra-

bility

The column “in practice” is not given for MCA, because it is still in its

introductory phase. Neither Table 4.6, nor Table 4.8, which is the continuation, can

give a final, qualified answer to the question of which approach is better. What

196 4 Towards Better Product and Process Modelling

they are giving you is a basis for comparison, showing which modelling

parameters lead to differences when used/changed.

Table 4.7. Modelling parameters (concerning organization and implementation) in use by

SCA and MCA (

cf. Table 4.4c and Table 4.4e). Parameters with the same use are not shown.

MCA

theoret.

possible

practice

theoret.

possible

Correlatio

SCA

clear

interfaces

100% 5% 100%

libraries

100% 5% 100%

broker

100% 5% 100%

capability to

answer

pre

defined

ques

tions

about

itsel

0% 100%

capability for

self-

maintenance

0% 100%

capability to

take

decisions

0% 100%

capability for

self-

improvement

0% 100%

100% 80% 100%

100% 70% 100%

100% 30% 100%

100% 5% 100%

0% 100%

100% 60% 100%

100% 40% 100%

low

decentralised

autonomy and intelligence

incorpo-

ration o

self-

aware

ness

structure

hierarchical

layered

Modelling parameters

model organisation

architecture

compo-

nent-based

open

modula

fractal

Modelling parameters

implementation modalities

object-

oriented

allow user-defined types

employment of recursion

proper choice of the level o

standardization

user qualification

high

inimization of the information content

mnemonics

factoring out reusable content

incorpo-

ration o

intelli-

gence

max. independence from the host

4.7 Comparison of MCA and SCA 197

These two tables could be used together with Table 4.4 for determining which

customer demands would be influenced. An attempt to evaluate which of the two

approaches is better suited to a particular situation can be carried out after

introducing weight factors for the modelling factors or even better – for the

(adapted to the specific situation and needs) functional requirements. We leave this

task as an exercise to you – probably after considering the remainder of the book.

4.7.1 Modelling Efficiency

Since most systems have a hierarchical structure, the lower in the hierarchy we go

the higher is the number of the components on the respective level and the lower is

their complexity –

cf. Equation 3.10 and Figure 3.16. Let us consider how this can

influence the modelling efficiency.

The efficiency of a process is usually defined as the ratio of its output to its

input. The output of the modelling process is a model. Assuming that the

requirements for the model define the modelling process, but do not belong to it,

what remains on the input are the modelling techniques and tools. Therefore, the

modelling efficiency for the creation of just one model can be expressed as

follows:

efficiency

modelling

toolstechniques

ellmod

priceprice

value

(4.1)

According to this equation (which neglects the value of time savings) the

efficiency would tend to infinity when the divisor tends to zero. Supposing that

smaller and less complex models have lower value, it appears more efficient to use

cheap tools and techniques for small models, and more expensive tools and

techniques for larger, more complex models. The only fact that should not be

disregarded is that the use of more than one tool for authoring can cause

incompatibilities or problems during the integration of the separate models.

Of course, no tool is used for only one model. To take this into consideration

we should estimate how many models (denoted by

N below) could be produced

during the lifetime of the respective tool and what their value is.

efficiency

modelling

toolstechniques

ellmod

priceprice

value

∑

(4.2)

If we assume that the frequency of the need to edit (i.e., create or modify)

component (without its sub-components!) does not depend on its size, and that

each modification is performed within a CAx-system, it follows that each system

will be used more frequently to modify small components than large ones.

4.7.2 Systems for Modelling (Authoring Tools) vs. (Systems of) Models

In Table 4.8 is presented a short comparison between the systems for modelling

and groups of models, integrated into systems.

198 4 Towards Better Product and Process Modelling

Table 4.8. Systems for modelling vs. systems of models: a short comparison

Trait Systems for modelling Systems of models

Aim To satisfy the needs within

a given domain (abstract

aim!); needs are

permanently growing.

Moreover, to be flexible the

systems have to posses

reserve functionality

another cause for growing.

Each model should be, by

definition, adequate but

still finite representation of

the modellee

(concrete/tangible aim). A

system of models is a

(finite) number of models.

Aim abstractness abstract/fuzzyÎ more

difficult to pursue.

concrete/tangibleÎ much

easier to pursue.

New version shows how far from perfect

the system is.

brings the model towards

its aim by making model-

based decisions as sensible

as the modellee-based.

Growing every system tends to get

more and more

functionality and nothing

stops it from growing

infinitely.

Models must remain finite

(according to their

definition!); the size of

compound models (systems

of models) can be huge, but

it still remains a sum of

finite number of

components, each of finite

size.

From Table 4.8 it follows that when (CAx)systems are used as both tools for

modelling

and hosts for the created models, they have a much smaller chance of

becoming perfect than systems of models hosted independently on lean hosts.

Consequently, it is better to focus on the development of the models themselves,

which is one of the main ideas of MCA.

5 Conclusion

Conclusion

A new scientific truth does not triumph by

convincing opponents and making them see the

light, but rather because its opponents

eventually die, and a new generation grows up

that is familiar with it.

Max Planck

A scientific Autobiography and Other Papers, 1949

Whenever science makes a discovery, the devil

grabs it while the angels are debating the best

way to use it.

Alan Valentine

The way to achieving The Perfect Modelling approach can be very long, even

endless. But as the saying puts it, even the longest way starts with the first step.

The Model Centred Approach described in the previous chapter is an attempt not

only to make the first step on this way, but also to achieve some advancement

towards a better – and simultaneously towards The Perfect – modelling approach.

5.1 Based on Highly and Easily Integrable AIEs

One of the main problems of SCA – the integration of model components created

in different CAx-systems – is solved in MCA by unique shifting of the integration

paradigm. In contrast to SCA, which relies on integration of the component

themselves, or even worse – on integration of the authoring tools to achieve simply

integration of the model components – MCA relies on

integration of the

component's functionality

, accomplished by means of cooperative work and

communication (for details on different types of integration we refer to Section

200 5 Conclusion

3.1.3 above). This concept for integration resembles very much how humans

cooperate to integrate their efforts and is based on three pillars: design for

integration, componentization and proper choice of the level of standardization.

Alternatively, backed by Knowledge-based Adaptive Conversion proposed in

Avgoustinov (1997) it can support hierarchically-distributed (

i.e., separately

implemented for each model or component and not as a standalone converter)

offline and online exchange and thus allow the adoption and reuse of models,

created by means of distinct CAx-systems.

5.2 Extremely Flexible and Extendable

Due to its open organization and architecture, the flexibility of MCA remains high

even when the model components are not open themselves. The organization,

having no fixed structure and based on components with clearly defined interfaces,

enables easy exchange of separate components, changes in the interaction of the

model components, exclusion of unneeded or inclusion of additional models. This

allows even benchmarking of components with the same purpose and interfaces,

but from different producers, and to choose/buy the one, which is best suited to the

respective purpose.

The proposed incorporation of intelligence in components at the middle and

upper levels of the model hierarchy (AIEs, AIMs,

etc.; Figure 4.6) increases the

flexibility and efficiency even more, unburdening in addition both model authors

and model users from the frequent need to make routine decisions, calculations or

other recurring routine activities.

The high flexibility and extendibility make MCA universal and useable in a

number of application areas. Combined with its efficiency, this makes it well-

suited also to small and medium-sized enterprises.

5.3 Network and Web-ready Capabilities

Another advantage of MCA is its readiness for use over networks, and possibilities

to support distributed cooperative work. There are two aspects here. On the one

hand, humans are supported in their cooperation according to at least the following

combinations:

• author to author: for cooperative development;

• author to user: for teaching and training;

• user to user: for exchanging experience or cooperative work/use; and

• user to author: for feedback, request of new models/functions or other

customer support.

On the other hand, it is possible to compose sophisticated models, whose

components are distributed in diverse geographical locations but can be used over

the network just as easly and conveniently as the local use of any software system

would be.

5.4 Scalability up to and Beyond Distributed Virtual Enterprises 201

5.4 Scalability up to and Beyond Distributed Virtual

Enterprises

The open organization and the easy integration and flexibility offered by MCA,

satisfy the prerequisites for achieving arbitrary scalability of the models. Their

network-readiness, in addition, provides a way to scale their performance by

distribution of the model parts and the diverse calculations on different computers

and other hardware resources.

5.5 Advantages for Modelling

Summing up, compared to SCA the MCA offers plenty of advantages – among

them higher intuitiveness and ease of use, low (accidental) complexity, leading to

increased efficiency of the modelling, high flexibility, interchangeability and

extendibility of the sub-models, low costs for development, use and maintenance

and reduced volumes of information flow. All these taken together lead to higher

reusability, and eventually to higher economic efficiency.