Research Reports from the Department of Operations

Document Type

Dissertation

Publication Date

6-1-1972

Abstract

A class of industrial inspection problems, important because it often causes production bottlenecks, is posed as a pattern recognition problem. A typical member of this class is the inspection of surface quality in flat-rolled steel. Varying inspection requirements, the limitations of manual visual inspection, and the production potential of modern plants motivate research into automatic inspection. Operations research methodologies are applied to three outstanding problems in applied pattern recognition in the context of the automatic inspection system design problem. The design methodology for the inspection system, as proposed in block diagram form, is based on the solutions to these three problems. The first problem, that of providing a unified approach to pattern recognizer design, was solved by formulating the sensor and feature selection problem as a 0-1 integer linear program. A minimum cost solution is sought, subject to the total feature utility, or usefulness to the inspection problem, exceeding a lower bound. The second problem was that of identifying and evaluating candidate features where the pattern recognizer designer is not an expert in the application area. It was solved by using the DELPHI method to quantify the subjective opinions of steel plant inspection personnel. The third problem, that of real-time adaptation of the pattern classifier parameters, was solved by comparing several approaches and recommending one which is a synthesis of other approaches. The recommended approach requires post-optimality operations in R-1 linear programs for an R-category classifier at each adaptation. Two examples of the design methodology are presented. Both use data obtained from a DELPHI exercise, and one example contains a complete inspection system error rate simulation.

Keywords

Operations research, Pattern recognition systems, Steel industry and trade--Quality control, Automatic data collection systems, Project management--Mathematical models, Delphi method, Machine learning, Linear programming, Quality control--Data processing

Publication Title

Dissertation/Technical Memorandums from the Department of Operations, School of Management, Case Western Reserve University

Issue

Technical memorandum no. 261 ; Submitted in partial fulfillment of the requirements for the Degree of Doctor of Philosophy.

Rights

This work is in the public domain and may be freely downloaded for personal or academic use

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.