Thesis Details


Thesis Title: A methodology using expert systems techniques to automate the process of diagnosing faulty plant conditions as identified by existing monitoring techniques
Thesis Author: Joe Long
Abstract: Computer monitoring systems being used in textile manufacturing are capable of collecting greater amounts of descriptive information than ever before. By itself, this information is difficult to manage and is not in a proper form to effectively diagnose manufacturing problems met in day to day operations. However, with some facility for manipulating this information, effective problem diagnosis can be achieved. Therefore, a methodology was developed in this thesis which suggested a way of using data collected from monitoring systems to automate problem diagnosis. The methodology specifies three. . necessary stages for automated probLem diagnosis. The first stage, called actuation, identifies the existence of a problem. Next, localization, the second stage, determines where the problem is located. Resolution, the final stage, then generates a plan of attack by which the problem can most effectively be attacked. To demonstrate an !lPplication of this methodology, a prototypical expert system was constructed. This system was developed for the problem domain of Sulzer projectile weaving machines, and demonstrated that data produced by monitoring systems could be used to automate the diagnosis of manufacturing problems encountered in day to day textile operations. The methodology developed in this thesis will provide manufacturing organizations with several benefits. First, the expertise of key personnel, one of a company's most valuable resources, will now be captured and made available with no time or location restrictions. In addition to this, management will now be freed from the mundane task of monitoring data analysis. As a result, they will have more time to spend doing the job they were hired to do -- use their brains and make the company more profitable.