Date of Award

Fall 1992

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Mechanical Engineering - (Ph.D.)

Department

Mechanical and Industrial Engineering

First Advisor

Keith T. O'Brien

Second Advisor

Rong-Yaw Chen

Third Advisor

E. S. Geskin

Fourth Advisor

Nouri Levy

Abstract

Injection molding of engineering thermoplastics is the most widely used manufacturing method in industry. It is a priority to maintain a deviation-free operating environment to ensure high quality, low cost manufacture. An expert system for the injection molding of engineering thermoplastics has been investigated. The system can be used to attenuate the deviations experienced during the injection molding of engineering thermoplastics. The system is coded in C programming language.

The resolution procedures of this system include two stages such as the definition of declarative knowledge and the procedure of corrective actions. In the definition of declarative knowledge, all of the necessary information is collected for firing the inference engine. This information includes the material type, the material manufacturer, the material grade, the recommended operating conditions, the operating conditions, the deviation type, and the correlative weighting factors.

The, procedure of corrective action is classified by fishbone diagram into four different levels. These levels include and are ranked as method corrective actions, operating variable corrective actions, mold corrective actions, and material corrective actions.

The rule values of the corrective action in each level are assigned to determine the rank for employing these corrective actions. Among, those rule values, the rule values of the method corrective actions, of the mold corrective actions, and of the material corrective actions are determined by the degree of difficulty required to eliminate the deviation and the input of the molding experts. A decision algorithm is developed to calculate the priority weighting, factors, rule values, of each operating variable corrective action. Furthermore, the Pareto principle is introduced to analyze the control parameters of the decision algorithm.

During the interactive procedures of eliminating the deviation, the system provides an explanation function for each step. It allows the system to illustrate the reason for each action to the user. A self-learning mechanism is also developed in this study. This self-learning mechanism based on the response of the resolution results modifies the parameters which influence the sequence of the corrective actions.

The system has been examined by experts in the field of injection molding. It is recognized that the system not only provides reasonable resolution sequences for eliminating the deviation, but also, accurate suggested actions for the user. Furthermore, the resolution actions have been simulated in the injection molding filling package MOLDFLOW. This confirms that the resolution actions can actually influence the parameters which can eliminate or reduce the deviations.

Share

COinS