Modeling the sound absorption behavior of carpets using artificial intelligence

Abstract

The ability to provide some degree of noise attenuation is one of the most important properties of carpet. This is achieved by making the room in which the carpet is being installed less reverberant or minimizing the transmission of footstep noise through floors. This study aims to predict the sound absorption coefficient of acrylic carpet at different frequencies using three computational intelligence techniques viz., Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Interface System (ANFIS), and Genetic Algorithm (GA). To this end, carpets with different pile height and densities were produced. In order to simulate walking traffic, the carpets were exposed to 50, 100, 150, and 200 dynamic cycles. The sound absorption coefficient (SAC) of carpets was experimentally measured using a two-microphone impedance tube based on the transfer-function method. The effect of input parameters on SAC was statistically investigated, and the results showed that all parameters have a significant impact on SAC at a 95% confidence interval. To improve the prediction accuracy of the model, GA was implemented for the optimization of ANN and ANFIS parameters. The prediction accuracy of hybrid models ANN-GA and ANFIS-GA was compared with the traditional regression model by the mean absolute percentage error (MAPE). The results indicated that the prediction accuracy is considerably enhanced by using an optimized ANN and ANFIS structure. The MAPE for ANN-GA, ANFIS-GA, and regression models was found to be 11.85%, 17.68%, and 61.82%, respectively. The results demonstrated the applicability and performance of the hybrid ANN-GA model for the prediction of SAC of carpet.