L. P. Khoo

Framework of a fuzzy quality function deployment system

SUMMARY Quality function deployment (QFD) is a planning methodology for translating customer needs into appropriate product features. In QFD operation, relationship matrices are used to describe the relations between different customer needs, between different design requirements and between customer need and design requirement. These relations are expressed using linguistic or crisp variables. Linguistic variables are characterized by ambiguity and multiplicity of meaning. Current research in QFD focuses on two areas: simplification of the documentation process and computerization of QFD. Effort to address the semantic in the linguistic variables, however, has been neglected. To fully automate the laborious manual QFD operation, the ability to interpret the semantic of the linguistic variables has become necessary. In this work, an approach centred on the application of possibility theory and fuzzy arithmetic has been developed to address the ambiguity. Details of the approach and the framework of a fuzz...

Feature extraction using rough set theory and genetic algorithms--an application for the simplification of product quality evaluation

Feature extraction is an important aspect in data mining and knowledge discovery. In this paper an integrated feature extraction approach, which is based on rough set theory and genetic algorithms (GAs), is proposed. Based on this approach, a prototype feature extraction system has been established and illustrated in an application for the simplification of product quality evaluation. The prototype system successfully integrates the capability of rough set theory in handling uncertainty with a robust search engine, which is based on a GA. The results show that it can remarkably reduce the cost and time consumed on product quality evaluation without compromising the overall specifications of the acceptance tests.

Solving the assembly configuration problem for modular products using an immune algorithm approach

Designing for modularity enhances the agility of a manufacturing system. It allows manufacturing systems to be built under high product customization requirements and at the same time keeping product development time low. Modular products imply in themselves a complex multiproduct assembly line design challenge. The design of such assembly systems comprises some of the most challenging computations in engineering and mathematics. More often than not, the optimal solution to the problem could not be found. This paper describes an approach based on the principles of natural immune systems for solving the design of assembly system for modular products. The approach, which was based on immune algorithm, was illustrated using an assembly combinatorial problem and the results obtained were compared with those derived by an established heuristic algorithm, the genetic algorithm. The results show that the immune algorithm outperforms the genetic algorithm in terms of convergence trends, distribution of near-optimal solutions and quality of solutions.

Information deduction approach through quality function deployment for the quantification of the dependency between design tasks

The design structure matrix (DSM) has been widely known as an effective approach for the modelling and analysis of a design process for the manufacture of a product from the perspective of information flow. It provides a formal method to capture and manage the interactions and interdependences among design tasks. Nonetheless, the difficulties in constructing a reliable numerical DSM prevent wider applications of DSM and its derived techniques. In this work, an approach to quantify systematically the dependency between design tasks in a DSM is proposed. The proposed approach aims at estimating the coupling strength of tasks in a DSM by making full use of the information contained in quality function deployment matrices. To realize this, a house of quality matrix is first constructed, followed by transferring the information in the house of quality to an extended design deployment matrix, and then the dependence strength of each task–attribute pair in the extended design deployment matrix is analysed. The details of the proposed approach are presented. The performance of the proposed prototype is illustrated by using a case study on a burn-in system. The results obtained from the case study are discussed.

A fuzzy-based genetic approach to the diagnosis of manufacturing systems

Abstract This paper describes the development of a hybrid approach that integrates graph theory, fuzzy sets and genetic algorithms for the diagnosis of manufacturing systems. The approach enables the modelling of causal relations of system components in manufacturing systems. Based on the model thus established, a worst-first search technique has been proposed and developed for the identification of probable fault-propagation paths. As manufacturing diagnosis often involves the interpretation of uncertainty, fuzzy-set theory is employed for this purpose. Unlike conventional diagnostic systems which assume that all the system components or nodes of a manufacturing system model are measurable, the genetic-algorithm-based search engine developed in this work is able to deal with nodes that cannot be, or are not, measured. Details of the hybrid approach, the worst-first search technique and the genetic-algorithms-based search engine are discussed. The framework of a prototype fuzzy-based genetic diagnostic system is described. Details of the system validation are also presented.

PCB assembly planning using genetic algorithms

Printed circuit boards (PCB) are used extensively in industry for the manufacture of electronic and electromechanical products. One of the primary concerns in the manufacture of PCBs is the determination of the optimal assembly plan. This paper presents work that leads to the development of an approach to PCB assembly planning using genetic algorithms (GAs). The approach takes into consideration component insertion priority and sequencing decision rules. A polygamy reproduction mechanism with dual mutation has been proposed and implemented. Details of the approach are described. A PCB model extracted from the literature was used for performance evaluation. Details of the evaluation are presented.

Agent-based multiple shop floor manufacturing scheduler

This paper describes an agent-based architecture for scheduling multiple shop ̄ oors using a genetic algorithm-enhanced scheduling engine. The architecture consists of two main modules, the manufacturing scheduling server (MSS) and shop scheduling client system (SSCS). Two types of software agents, namely a supervisory agent and a shop ̄ oor agent, undertake the scheduling of individual shop as well as the global schedule. The supervisory agent coordinates the negotiations among the shop ̄ oor agents as they attempt to arrive at a global near optimal schedule, while, at the same time, resolving con ̄ icts in the local shop ̄ oor schedules. The prototype agent-based multi-shop scheduling system was validated by a hypothetical six products £ three shop ̄ oors and a plastic injection moulding company. In both cases, the prototype agent-based multi-shop scheduling system generated a feasible and near optimal schedule for the entire manufacturing system.

A prototype fuzzy resistance spot welding system

In assembly automation, the need to perform on-line quality monitoring and to provide near real-lime inspection of resistance spot welds, has attained unsurpassed importance in industries. This paper describes the development of a prototype fuzzy control system for a two-stage resistance spot welding machine with potential of being incorporated into an automation line. It encompasses a new approach aimed at providing simultaneous control of more than two major welding parameters for achieving consistent quality spot welds by assessing the condition of the heat affected zone created. The prototype system, developed in C language, comes with a decision making mechanism and a quick search mechanism. Its rule-base contains a total of 125 heuristic control rules derived from literature reviews and interviews with experts and through experiments. The various factors affecting the quality of spot welds are outlined in the paper. The correlation between spot weld quality and welding parameters were examined using...

Knowledge Acquisition and Uncertainty in Fault Diagnosis: A Rough Sets Perspective

The ability to acquire knowledge from empirical data or the environment is an important requirement in better understanding many natural and artificial organisms. This ability relies heavily on the quality of the raw information available about the target system. In reality, these raw information/data may contain uncertainty and fuzziness, that is, it may be imprecise or incomplete. A number of techniques, such as the Dempster-Shafer theory of belief functions and fuzzy set theory, have been developed to handle knowledge acquisition in environments that exhibit uncertainty and fuzziness. However, the advent of the rough set theory in the early 80’s provides a novel and promising way of dealing with vagueness and uncertainty. This chapter will address the issue systematically by covering a broad area including knowledge acquisition / extraction, uncertainty in general, and techniques for handling uncertainty. The basic notions of rough set theory as well as some recent applications are also included. Two simple case studies related to fault diagnosis in manufacturing systems a reused to illustrate the concepts presented in this chapter.

Adapting IDEF0 modelling to perform manufacturing diagnosis

This paper presents the work done to adapt a system modelling methodology, ICAM DEFinition Zero (IDEF0), to perform manufacturing diagnosis. It describes the basic notions of IDEF0 modelling and the underlying principle of a novel reasoning technique, the “worst-first” search, developed for manufacturing diagnosis. The reasoning technique which was originally based on graph theory and possibility theory, has been adapted to access the information stored in an IDEF0 model. Details of a prototype IDEF0-based system for manufacturing diagnosis are presented. The results of system validation based on a manufacturing system for the production of mechanical components are reported.