Date of Award

1994

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Chemical Engineering

First Advisor

Armando B. Corripio

Abstract

Reaction Injection Molding produces parts that must meet property specifications to be useful. For example a car bumper must have a high impact strength to prevent the car from being damaged during a collision. This research presents a control algorithm that can be implemented to control the part properties of reaction injection molded parts. The part property need not be measurable on line if the property is affected by the reaction conditions within the mold. This is an important feature of the algorithm since many important properties, such as yield strength, cannot be measured without destroying the part. The algorithm will of course work with on-line property measurements. The control algorithm implements two empirical models of the process. One model relates the variables that can be manipulated by the operator, such as reactant temperature and reactant ratio, to the desired part property and the other model relates model reaction conditions to the part property. The second model uses coefficients of a simplified reaction temperature profile model as the independent variables. A statistical process control filter developed by R. Rhinehart (Rhinehart, 1991) reduces the variation in the property estimates that are sent to the controller. The controller minimizes the moves in the manipulated variables subject to the estimate of the part property being equal to its set point for the part property. The control algorithm was tested on a RIM unit located in the Chemical Engineering Department at Louisiana State University. The algorithm when subjected to a set point change and a disturbance in the chain extender was able to follow the set point in only eight shots and to compensate for the disturbance by bringing the process back to set point within five shots.

Pages

264

Share

COinS