Group technology and expert systems concepts applied to statistical process control in small-batch manufacturing

Abstract

This research addresses the issues of small lot sizes and manufacturing complexity in the implementation of statistical process control (SPC) in a small-batch, flexible manufacturing environment. The efforts of this research are divided into two major phases. In the first phase, a general approach for implementing SPC in a small-batch, flexible manufacturing environment is proposed to overcome the difficulties encountered in this type of manufacturing. The proposed methodology combines analysis of variance with group technology (GT) part family concepts. Data collected from industry is used to illustrate the proposed methodology. The results indicate that the proposed methodology works successfully. In the second phase, an automatic process deviation reasoning system is proposed as an alternative to the traditional human-directed problem-solving approach. The proposed system consists of a deviation recognition algorithm and a knowledge-based expert system. The functions of this system include recognition of process deviation, pointing out the sources of variation, and suggesting the direction of action. The recognition algorithm combines a decision-theoretic approach with a syntactic approach appropriate for recognizing nonrandom patterns on a control chart. Primitives, grammar, a distance metric and a decision-making criteria are defined. The algorithm is presented in a step-by-step procedure as well as in a dynamic programming formulation. The algorithm performance is reported for numerous patterns including cycles, trends, systematic patterns, stratifications, shifts, and mixtures using simulated data. The conclusion is that the algorithm performs very well on nearly all of the nonrandom patterns and would be useful in an automated or unmanned environment. The knowledge-based expert system is written in Turbo Prolog and runs on an IBM PC/XT/AT or PC compatible. The system consists of four major components: interpretation, diagnosis, action planning, and statistical consulting. In the proposed construct, the identified pattern together with process and product characteristics serves as the evidences for determining the control state of the process and the direction of diagnosis. A set of sample sessions are given to illustrate the execution of the proposed construct. The proposed construct provides a groundwork for a full-scale, real world application. The research begun with this dissertation may be broadened in several future directions. The ideas of combining part family concepts and ANOVA could be extended to the multivariate cases. In the deviation recognition algorithm, research efforts could focus on the primitive extraction and the extension to non-Shewhart control schemes. Further efforts in development of the knowledge-based expert system may concentrate on the investigation of an inexact reasoning mechanism.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 23 citations exist upstream.