Intelligent Curing of Thick Composites Using a Knowledge-Based System

Abstract

An expert-system-based tool was developed in order to operate the autoclave cure of a thick fiber reinforced thermosetting matrix composite laminate in an optimal manner. The best temperature profile is obtained using a novel model-based optimization technique. It incorporates known heuristics about the problem to prune the search space and reduce the dimensionality of the problem so as to converge to a near-optimum in a reasonable amount of time. The objective function used has components for minimizing the total cure time, reducing the temperature excursions during an exotherm and minimization of process-induced residual stresses in the laminate. The inevitable divergence between the process and the simulation precludes the direct application of such optimal profiles. We have devised a general methodology to address this issue of process-model mismatch, which allows for the implementation of model-based optimal profiles in real situations. This methodology is directed at ensuring that the actual cure follows the same trends as predicted by the optimal cure, rather than trying to achieve an exact match. Trend analysis is used to access the underlying process trends in the profiles of different variables of the optimal solutions. Data from the autoclave is also analyzed on-line to determine the trends in real-time. A correlation between the control decisions and the trend analysis of a number of profiles from the optimization module is used to determine rules to translate the `simulation' optimal profile to one that will be used on the process. The results from application of this overall strategy for the curing of glass-polyester composites are presented. The expert system achieved the desired objectives of minimizing the cure time, while still producing a high quality part.