Problems in Control Chart Pattern Recognition Systems

Abstract

Introduction Artificial Intelligence has been applied to a variety of areas in the fields of statistics and process control. Expert systems have ranged from consultants to robotic vision systems. Hand outlined the development of a based on studying the way a consultant approached a problem and addressed problems that could be encountered in the development of the system[1]. Hahn defined a statistical expert as an automated consultant and data analyst that guides the analyst directly to a solution of a general problem[2]. But, Hahn concluded, more research is needed before a general expert system is successful and expert systems which are more specialized may be more feasible. There are specialized systems in quality and process control which exist in the manufacturing area. Giovannone discussed the integration of robots in locations around Italy[3]. These robots take measurements but most of them only have an interface to a program that controls the sampling and records the information. Haton also describes the application of artificial intelligence to sample inspection[4]. The integration of vision systems used for quality inspection will expand during the next decade, according to Haton. An expert system to monitor the vision system is a possibility. Ackerman, Plsek and Surette describe some systems developed by AT&T for internal corporate use[5]. Most systems mentioned by them operate in real time and cover aspects of process control such as product yields, reliability analysis and quality cost planning. A few commercial packages implementing artificial intelligence are available for quality control. A survey of packages a few years ago indicated that a few packages by A. William Wortham and Associates analyse charts using artificial intelligence[6]. The analyses in these packages consisted of several tests, all combined through artificial intelligence to give a considered decision. There are many more software packages that perform chart analysis but do not mention artificial intelligence applications. Most of the packages only test for a small number of patterns, like shifts and trends, or use runs rules for detection. There is no mention of accuracy of these packages or the false alarm rate for the tests performed.

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 — 7 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.