Study on the effects of denier and shapes of polyester fibres on acoustic performance of needle-punched nonwovens with air-gap: comparison of artificial neural network and regression modelling approaches to predict the sound absorption coefficient of nonwovens

In this article, a comparative analysis of artificial neural network (ANN) and regression modelling approaches has been carried out to predict the sound absorption coefficient (SAC) of nonwovens plus air-gap at wide range of frequencies (50-6300 Hz). Needle-punched nonwoven fabrics were produced with different denier and cross-sectional shapes of polyester fibres to study their combined effect on acoustic performance of nonwovens. The surface area of fibres, specific airflow resistance and mean flow pore size of nonwovens were analysed to explain their sound absorption behaviour. Finer fibre nonwovens perform better than the coarser fibre nonwoven as sound absorber. The effective surface areas of fibres in the nonwoven structure greatly affects the SAC. Finer fibres will get aligned easily in z-direction compared to coarser fibres, facilitating formation of more tortuous channels in the fabric structure contributing damping of sound waves. It has been observed that ANN model predicts the SAC with high degree of accuracy than the regression model. The ranking of input parameters in predicting SAC of nonwovens was analysed. Both the models ranked frequency of sound is the major determinant for predicting SAC followed by specific airflow resistance of nonwoven fabric.