NEURAL NETWORK ANALYSIS APPLICATION TO PERMEABILITY DETERMINATION OF FIBERGLASS AND CARBON PREFORMS

Preform permeability is an important process parameter in liquid injection molding of composite parts.This parameter is currently determined with time consuming and expensive experimental procedures.This paper presents the application of a back-propagation neural network to predicting fiber bed permeability of three types of reinforcement mats. Resin flow experiments were performed to simulate the injection cycle of a resin transfer molding process.The results of these experiments were used to prepare a training set for the back propagation neural network program.The reinforcements consisted of plain-weave carbon,plain-weave fiberglass,and chopped fiberglass mats.The effects of reinforcement type, porosity and injection pressure on fiber bed permeability in the preform principal directions were investigated.Therefore,in the training of the neural network reinforcement type,these process parameters were used as the input data.Fiber bed permeability values were the specified output of the program.As a result of the specified parameters,the program was able to estimate fiber bed permeability in the preform principal directions for any given processing condition.The results indicate that neural network may be used to predict preform permeability.