Predicting thermal behaviour of multilayered fabric assemblies using artificial neural networks

Abstract

The study aims to predict the thermal resistance of a multilayered fabric assembly comprising an inner layer of interlock fabric, an outer layer of PU-coated nylon, and a middle layer of spacer fabric, hollow polyester wadding, or micro polyester wadding, using Artificial Neural Networks (ANN) in MATLAB. Two neural networks are developed to predict the thermal resistance. Network one (N1) consists of three layers with four neurons in the hidden layer, and network two (N2) comprises three layers with three neurons in the hidden layer. In N1, four input parameters—thermal resistance of individual layer (inner, middle, and outer) and the thickness of the multilayered assembly—are employed. In N2, only the thermal resistance of the individual layers is used as input. The predictive performance of both models is evaluated using four statistical parameters: root mean square error (RMSE), mean bias error (MBE), mean absolute error (MAE), and coefficient of determination (R²).  The results indicate that even without incorporating the thickness of the multilayered assembly, the ANN model can accurately predict the overall thermal resistance based solely on the thermal resistance of the individual layers.