Use of simulations in optimization of, and design of a knowledge-based control system for, a composite manufacturing process

The objective of this work is to exploit mathematical simulations for developing Knowledge Based Systems (KBSs) for intelligent design and control. A generic framework for such KBSs is proposed, leading to a system with a hierarchical and modular architecture. Off-line optimization using the simulation and real-time control issues are treated separately.;We have developed a new method for determining optimal trajectories for batch processes, called Local Criterion Optimization (LCO), which uses available heuristics to reduce the difficult functional optimization to a low-order discrete optimization. For a test (the brachistochrone) problem, the technique was found to converge to the optimum very quickly and to be very tolerant about the mathematical form in which the heuristics are input. The method requires minimal interaction in that application requires only a process simulation and a mathematical specification of the heuristics. For the autoclave molding process, the slow heat transfer in the system typically leads to an exotherm and non-uniform curing. LCO was used to determine an optimal cure cycle where not only are the gradients during the cure minimized but the processing time is reduced substantially.;Model-based optimal profiles cannot be applied directly in practice because of the inevitable mismatch between the model predictions and process behavior. A strategy based on the trend analysis of process variables was devised to overcome this difficulty. The optimal operation is defined in terms of a specific temporal correlation of the control decisions in the process to the episode end-points of the (simulated) profiles of the variables that will be used to monitor the process. The optimal profile is implemented using feedback from the process and modifying the control decisions based on the changes in trends of the measured variables. This forces the process to follow the same "trajectory" as the optimal profile.;The overall strategy was implemented in a real-time KBS environment for the autoclave cure process. The output of the model-based optimization forms the core of the knowledge. The trend analysis-based modification was executed using rules written in a consistent manner, to obtain a transparent and easily modifiable knowledge base. Results from computer-controlled cures show that the KBS was successful in delivering high-quality parts on a consistent basis. (Abstract shortened by UMI.).