Neural Based Optimization of Composite Curing Process

This paper addresses the application of nature inspired methods to the optimization of resin cure in thick or thin composite laminates. The methodology presented in this work consists in the coupling of artificial neural network with optimization algorithms, focusing on the thermal cycle evaluation that further the progress of the degree of cure (DoC) within the material. Simultaneously, inhibition of temperature overshooting and excessive through-thickness gradient of temperature and cure degree is provided. Genetic algorithms and fuzzy logic controller were applied to identify the optimal processing parameters. In the former strategy, the optimization algorithms iteratively refine the thermal cycle, whose fitness score is provided by the neural model. In the latter, the neural network delivers a precise prediction regarding composite material behavior over the course of resin cure, while the fuzzy logic controller continuously applies the correct alterations to the thermal cycle. In both cases, a significant reduction of the thermal cycle, with respect to the recommended one, is demonstrated.