Development of an Empirical Model for Noise Prediction in Typical Industrial Workrooms Based on Artificial Neural Networks

Abstract

Noise prediction techniques can be employed as practical tools for evaluating the cost-effectiveness of acoustic treatments and consequently, prevent blind treatments by experts so that more acceptable conditions are obtained. One of the most important issues in this regard is the development of accurate models for analysis of the complex relationships among acoustic features affecting noise level in workrooms. In this study, artificial neural networks were employed to develop a relatively accurate model for noise prediction in noisy industrial workrooms. The data from nine acoustic, structural and embroidery process features influencing the noise in 60 embroidery workrooms was used to develop the noise prediction techniques. Multilayer feed forward neural networks with different structures were developed by using MATLAB. The best neural networks could accurately predict the noise level (RMSE=0.69 dB and R2=0.88). Although networks are empirical in nature, the results confirmed the potential of this approach for minimizing the uncertainties in acoustics modeling. This model gives professionals the opportunity to make an optimum decision about the effectiveness of acoustic treatment scenarios in workrooms.