Abstract
Manufacturing of quality products via liquid molding processes such as Resin Transfer Molding (RTM), calls for a precise control of resin progression through fibrous preforms during mold fill. Lack of an effective process control leads to formation of dry spots and voids that are detrimental to product quality. This study presents the use of physics-based process simulations in real-time, towards a generalized process control. The implementation of process simulations for on-line model-predictive control requires that the simulation time scales be less than the time scales of the process. An artificial neural network trained using data from numerical process models is used to provide rapid, real-time process simulations for the model-based control. A simulated annealing algorithm, working interactively with the neural network process model, is used to derive optimal control decisions rapidly and on-the-fly. The controller performance is systematically demonstrated for several processing scenarios.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have