Wen‐Hai Chih and Dwayne A. Rollier
Prototype patterns and pattern diagnostic characteristics have beenproposed in a previous article. Simulation results based on theprototypes and the diagnostic characteristics…
Abstract
Prototype patterns and pattern diagnostic characteristics have been proposed in a previous article. Simulation results based on the prototypes and the diagnostic characteristics have also been presented as a justification for the study. Outlines a methodology, with three major components, which designates eight processors to identify the unnatural pattern. The working memory (blackboard) characterizes the symbolic structure of the transformed data and stores the intermediate results from each processor. Eight processors are the kernel of the knowledge base used to classify the pattern of the observations. The comparison of intermediate results is executed in the inference engine, which makes the preliminary decision. The implementation of the processors is coded in Turbo C and runs on an IBM PC/XT/AT or compatible PC. The results of the implementation and validation demonstrate that the methodology does a good job for two‐variable pattern recognition.
Details
Keywords
Wenhai W. Chih and Dwayne A. Rollier
Statistical quality control charts cannot indicate explicitly whetherthere is any special disturbance in the manufacturing process. patternrecognition scheme can solve this…
Abstract
Statistical quality control charts cannot indicate explicitly whether there is any special disturbance in the manufacturing process. pattern recognition scheme can solve this problem. The simultaneous control of two or more variables is necessary when the quality of the product depends on the joint effect of these variables. Studies the combination patterns of random and random, shift and cycle, trend and cycle, and trend and shift for two variables. Proposes a T2 control chart and uses simulation to determine pattern diagnostic characteristics for these combinations. The pattern diagnostic characteristics studied are window size, zone boundary, and zone representation. The results indicate that window size 20 is appropriate for these particular parameters, equal probability and the highest percentage alternative are adopted as the zone boundary and the zone representation, respectively. The sensitivity analysis of the pattern parameters indicates the pattern diagnostic is robust for changes in the parameter values.