An adaptive neural network system for prediction of thermal protective performance of fabrics

Thermal protective performance is very important for heat protective clothing. Based on Matlab neural network toolbox, an adaptive BP neural network with a single hidden layer is constructed to predict thermal protective performance of fabrics. The network consists of nine input nodes, eleven hidden nodes, and one output node. The input variables are fabric weight, thickness, weave, warp density, weft density, warp yarn count, weft yarn count, LOI, and Qmax, while TPP rating is used as output variable. In the training process, the connection weights are modified with gradient-descent algorithm and adaptive learning rate to solve two defects of the BP network. After training, the predicted ability of the proposed neural network is tested. The results show a good correlation between predicted values and experimental values. The adaptive BP neural network can be applied to predict the thermal protective performance of fabric.