Computer Graphics & Geometry

Mismatch Detection in Distributed Computer Aided Geometric Design

Victor Taratoukhine
Department of Information Systems,
Ulyanovsk State Technical University, Ulyanovsk, Russia, and
Researcher, Department of Computer & Information Sciences,
De Montfort University, Milton Keynes,U.K.
Vtaratoukhine@dmu.ac.uk

Kamal Bechkoum
De Montfort University,
Department of Computer & Information Sciences,
Milton Keynes, U.K.
kbechkoum@dmu.ac.uk

Contents

• Abstracts
• 1. Introduction
• 2. Multi-agent systems in design
• 3. Problem
• 4. A taxonomy of mismatches in design
• 5. Formal description of the multi-agent framework
• 6. The distributed knowledge- base
• 7. A multi-agent Framework: Cooperation and negotiation
• 8. Conclusion and future work
• References

1. Introduction
Research into the use of knowledge engineering in design has become widely accepted as a fast growing subfield of Artificial Intelligence (AI). Increasing numbers of researchers, and research groups, are active within this emerging subfield. From advocates of “knowledge intensive” CAD/CAM/CAE (eg. [1][2]) to promoters of broader “intelligent CAD frameworks” (eg. [3][4]) the common thread is the use of AI tools and techniques to provide automatic and semi-automatic solutions to the problem. These solutions aim at increasing the “intelligence” of existing CAD/CAM/CAE systems [9].
The AI technologies used are varied and include expert systems [5], genetic algorithms and evolution programming [6], neural networks [7], fuzzy logic [8] and multi-agent systems [4][10][11][12][13]. Hybrid methods combining more than one technology have also been used [14].
It is fair to say though, that design engineers are still skeptical about the ability (or inability) of current intelligent design-support systems. For example, even when endowed with some sort of intelligent behaviour, existing CAD/CAE systems cannot handle several types of inconsistencies that may occur during the design phase.
Mainly due to the complexity of the design process, existing solutions tend to approach the problem from a very specific angle. For example, commercial systems such as CATIA (Dassault Systems) and I-DEAS (SDRC) do provide assembly mismatch analysis, but their approaches are more focussed on the analysis of tolerances. Other contributions [5] are constrained by the number and types of mismatches considered. Often, attention is given to a few geometric mismatches only, with very little concern about (say) material or cost considerations.
Moreover, even when the proposed approach is successful in detecting a design anomaly rarely does it suggest a satisfactory way to resolve the problem. In most of the previous work all the design knowledge is centralised into one unit: the knowledge base [3][5][6]. The centralisation of knowledge coupled with the absence of a negotiation mechanism (between all parties involved in the design) makes the process of predicting the impact of any modification an (almost) impossible task.
We re-inforce here the view that a multi-agent approach can tackle many of the problems posed by the centralisation of knowledge into a single Knowledge Base. A brief overview of related work is presented in Section 2. In Section 3, we describe the problem in the light of currently proposed solutions. A key issue raised in this paper is that any solution is as good as the knowledge about the problem itself. In other words, if a particular type of mismatch is ignored the system cannot be expected to deal with that sort of mismatch when it occurs. This is why we attempted to define a basis for a taxonomy of design mismatches which is described in Section 4. The conceptual framework, based on a multi-agent architecture is presented in Section 5. A proposal for the organisation of the distributed knowledge base, knowledge representation paradigms, types and classes of knowledge is presented in Section 6. Cooperation and negotiation strategies presented in Section 7. In Section 8 are described concluding notes and future work.
2. Multi-agent systems in design
Due to the above mentioned difficulties, it seems reasonable to believe that a Distributed Artificial Intelligence framework, using a multi-agent architecture, is a natural alternative in dealing with the problem at hand. Work in this area is already gaining an increasing popularity.
J. D’Ambrosio et. al. [10] have presented also an agent-based framework to support hierarchical concurrent engineering. Within the framework preferences and constraints of a design supervisor are distributed to design subordinates, who are expected to exploit their local expertise within the context provided by global design information.
M. Hale and J. Graid [11] have developed a distributed intelligent system for aircraft design based on conception of a design integration framework. An Intelligent Multi-disciplinary Aircraft Generation Environment (IMAGE) is discribed which uses state-of-the-art computing technologies.
R. Subbu et. al. [12] have described an information architecture called Virtual Design Environment. The approach utilises evoluntionary intelligent agents as program entities which generate and execute queries among distributed computing applications and design databases.
H. Frost and M. Cutkosky [13] have developed a multi-agent architecture, which makes the capability of manufacturing processes manifest to designers starting with earliest stages of geometry specification. The approach is being implemented using agents, written in object-oriented language, which exchange feature-based capability models.
None of the aforementioned applications of multi-agent systems is orientated towards dealing with the problem of detecting mismatches that may occur during the integration phase of distributed design.
K. Bechkoum [5] describes a Intelligent Mismatch Control System (IMCS) which has the potential to detect some types of mismatches. The IMCS implementation is an important step towards a more comprehensive solution but is far from being defects free. For example, the number and types of mismatches handled by the system is narrowed down to a few geometric mismatches.
The work presented here takes the IMCS’ development one step forward. A new multi-agent architecture is proposed which gives the IMCS the ability to handle issues peculiar to the nature of distributed design. This multi-agent architecture will be at the heart of an intelligent distributed mismatch control system (IDMCS) that aims at ensuring that the overall design is consistent and acceptable to all.
3. Problem
The problem becomes that of creating a conceptual framework for building a knowledge-based mismatch control system. The framework is based on a community of agents which are capable of learning and/or adapting to changes in the environment.
In order to contribute towards a solution to the above problem the following steps were taken:
-define a taxonomy of distributed design mismatches.
-develop a Conceptual Framework for a multi-agent system that handles these mismatches. This should take into account:
-the design knowledge needed to be considered whithin each agent.
-the knowledge representation paradigm.
-communication and negotiation issues, including conflict resolution.
4.A taxonomy of mismatches in design
The problem of devising a fully-fledged taxonomy for design mismatches is a very complex one. This is because design is a multi-disciplinary task that involves several stages. These stages include input data analysis, conceptual design, basic structural design, detail design, production design, manufacturing processes analysis, and documentation (see [9]).
A broad classification based on geometrical mismatches is represented in [5]. Some of the important parameters to consider in the case of Design for Assembly (DFA), are presented in [15].
Our taxonomy uses some of these known parameters, but is especially oriented for implementation for mismatch detection during the integration phase of mechanical engineering design. The proposed taxonomy of mismatches is presented in Fig. 1. The types of possible connections are presented in Fig.2.
It can be seen from Fig. 2, for example, weld connections require accordance between types of materials and material thickness, as well as, geometric parameters of material.



The bolted connection (see Fig. 3) requires parameters such as thread major diameter, minor diameter and pitch to be in accordance.
For the correct mismatch detection process we need to represent into our knowledge-base geometric information and information about materials from which parts are prepared.
It is making the capacity for new a class of design systems to impoving assembly mismatches at the earliest stage of the design.
5. Formal description of the multi-agent framework
The conceptual framework of the IDMCS is shown in Fig.4. The architecture assumes that the design knowledge is encapsulated within the different members of agent community. Conceptual framework (CF) may be formally presented formally as follows:
CF = {AP₁, ... , AP_t, ... , AP_n},
where
AP_t is the t^th Assembly Part, t = 1,2, … , n.
AP={DA₁, … , DA_i, ... , DA_m, CA₁, ... , CA_j, … ,CA_k},
DA_i is the i^th Design Agent (D-agent), i = 1, 2, …, m,
CA_j is the j^th Control Agent (C-agent), j = 1, 2, …, k.
Each DA_i consists of eight elements: FB - facts base, which including information about geometric characteristics of the part and material type. KB - knowledge-base. K - corrector block - which is adapted knowledge-base, as the result of communications with any other agents. I - interference engine. LI - local interface mechanism.
Each CA_j consists: MB - metaknowledge base, knowledge-base of control agent, interference engine, corrector block, local and global interface mechanism (GI).
Communications is key ability of multi-agents systems. The proposed communication protocol (CP) for the above agents is as follows:
CP={L_c1, ... , L_cm; L_i1, ... ,L_im; L_c^c₁, ... , L_c^c_k; L_i^c₁, ... , L_i^c_k},
where: L_i - information language, which describes current situation into multi-agent system, L_c - control language, which includes imperative commands about adaptation fact base of design agent for mismatch improvement, adaptation and modification D-agent’s knowledge-base. L_i^c - information language, which describes current situation for control agents, L^c_c - control language, which adapt meta- and knowledge-base of C-agents.
6. The distributed knowledge- base
The organisation of the distributed knowledge-base is presented as follows:
FB of D-agent including frame facts containing information about parts geometric and material consistency.
Rules in KB and MB of D- and C-agents provide analysis of design situation using experts knowledge. Each rule M_p, p = 1, ... , r, is characterised by premise part, comprising the IF preconditional statements, and the consequent part (THEN part), comprising the inferred outputs.
KB of D-agents is represented by the following rules of productions: rules called Receptors for making an analysis of the design situation, rules called Classifiers for making classification of situations according to the necessity of control actions.
KB of C-agents consists: Receptors, Classifiers and Effectors.
Receptors are represented as:
IF < situation = ST > THEN < start C >,
where ST - set of mismatch situations,
ST Î {ST₁, ST₂, ... , ST_p},
C - classifier. Classifiers are devided into three types:
IF < ST > THEN < estimation >,
IF < ST > THEN < estimation and recommendation>,
IF < ST > THEN < start E >,
where E - is effectors. Effectors are divided into 2 sets:
E = < E_ext, E_int >,
E_ext is a set of rules for modification and adaptation of the FB and KB of D-agents and E_int - self-adaptation MB and KB of C-agents.
7. A multi-agent Framework: Cooperation and negotiation
We represent conceptual framework as community of calendar [16] and reactive [2] agents. In our case D-agent is a reactive agent, which negotiate with other D-agents using design’ schedule (assembly sequence) generating C-agent.
Each D-agent is operated as independent entity and interactes asynchronously with other D-agents on peer-to-peer level and client server architecture (under the supervision of the C-agent).

D-agents are endowed with the capacity of negotiation with one another to ensure that any mismatches are detected and that a solution is proposed, but do not modify each other, because that is responsibility of C-agents.
C-agent receives the new information from D-agents using syntax of L_i, negotiate with other C-agents, using L_i^c and L_c^c , and updates the D-agents fact- and knowledge base using L_c, if mismatches occur.
8. Conclusion and future work
In this paper we have introduced a taxonomy of mismatches in design, have described a conceptual framework for building knowledge-based mismatch distributed control system which would be capable of detecting and resolving mismatches.
A proposal for negotiation protocol, which described the mechanisms of communication between agents, is also outlined.
The next step is a creation of initial prototype of IDMCS, which will liable us to evaluate the effectiveness of the proposed framework.
References
[1] T. Tomiyama, “Towards knowledge intensive intelligent CAD,” JSME-ASME Workshop on Design, pp. 46-51, 1993.
[2] T. Tomiyama, T. Kiriyama, and Y. Umeda, “Towards knowledge intensive engineering,” in Knowledge Building and Knowledge Sharing, IOS Press, pp. 308-316, 1994.
[3] V. Akman, P.J. ten Hagen, and T. Tomiyama, “A Fundamental and Theoretical Framework for an Intelligent CAD System,” Computer Aided Design Journal, Vol. 22, pp. 352-367, 1990.
[4] J. Bento, and B. Feijo, “An Agent Based Paradigm for Building Intelligent CAD Systems,” Artificial Intelligence in Engineering Journal, Vol. 11, pp. 231-244, 1997.
[5] K. Bechkoum, “Intelligent Electronic Mock-up for Concurrent Design,” Expert Systems with Applications Journal, Vol. 12, pp. 21-36, 1997.
[6] J. S. Gero, “Adaptive Systems in Designing: New Analogies from Genetics and Developmental Biology,” in Adaptive Computing in Design and Manufacture, I. Parmee (ed.), Springer, London, pp. 3-12, 1998.
[7] D. D. Daberkow, and D. N. Marvis, “New Approaches to Conceptual and Preminary Aircraft Design: A Comparative Assessment of a Neural Network Formulation and a Response Surface Methodology,” Proceedings of World Aviation Congress and Exposition, Anaheim, CA, USA, Paper N. SAE-985509, 1998.
[8] I. V. Semoushin, V.V. Shishkin, and V. V. Taratoukhine, “Knowledge-based Network Simulation System,” Proceedings of the 7^th International Fuzzy Systems Association Congress, Czech Republic, Prague, pp. 532 – 537, 1997.
[9] C. McMahon, J. Browne, Computer-aided Design and Manufacture, Addison-Wesley Press, 1993.
[10] J. G. D'Ambrosio, T. Darr, and W.P. Birmingham, “Hierarchical Concurrent Engineering in a Multiagent Framework,” Concurrent Engineering: Research and Applications Journal, Vol. 4, pp. 47-57, 1996.
[11] M. A. Hale, and J. I. Craig, “Preliminary Development of Agent Technologies for a Design Integration Framework,” Proceedings of 5^th Symposium on Multi-disciplinary Analysis and Optimisation, Panama City, FL, USA, September 7-9, 1994.
[12] R. Subbu, C. Hocaoglu, and A. Sanderson, “A Virtual Design Environment using Evolutionary Agents,” Proceedings of the 1998 IEEE International Conference on Robotics & Automation, Belgium, pp. 247-253, 1998.
[13] H. R. Frost, and M. R. Cutkosky, “Design for Manufacturability via Agent Interaction,” Proceedings of ASME Design for Manufacturing Conference, Irvine, CA, August 18-22, Paper N 96-DETC/DFM-1302, 1996.
[14] S. H. Kim, “An Automata-Theoretic Framework for Intelligent Systems,” Robotics and Computer Integrated Manufacturing Journal, Vol. 5, pp. 43-51, 1989.
[15] S. Lee, G. Kim, and G. Bekey, “Combining Assembly Planning with Redesign: An Approach for More Effective DFA,” Proceedings of the 1993 IEEE International Conference on Robotics and Automation, Vol. 3, pp. 319-325, 1993.
[16] J. Liu, and K. Sycara, “Distributed Meeting Scheduling,” Proceedings of Sixteenth Annual Conference of the Cognitive Science Society, 13-16 August, Atlanta, USA, 1994.

---

Computer Graphics & Geometry