Nonlinear modeling and prediction of forklift acoustic annoyance based on the improved neural networks

Abstract

Aiming at the characteristics of high decibels and multiple samples for forklift noise, a subjective evaluation method of rank score comparison (RSC) based on annoyance is presented. After pre-evaluation, comprehensive evaluation and data tests on collected 50 noise samples, the annoyance grades of all noise samples were obtained, and seven psycho-acoustic parameters including linear sound pressure level (LSPL), A-weighted sound pressure level (ASPL), loudness, sharpness, roughness, impulsiveness and articulation index (AI) were determined by correlation calculation. Considering the nonlinear characteristics of human ear subjective perception, objective parameters, and annoyance were used as input and output variables correspondingly and then three nonlinear mathematical models of forklift acoustic annoyance were established using traditional artificial neural network (ANN), genetic-algorithm neural network (GANN), and particle-swarm-optimization neural network (PSONN). Moreover, the prediction accuracy of the three models was tested and compared by sample data. The results indicate that the average relative error (ARE) between the experimental and predicted values of acoustic annoyance based on PSONN model is 3.893%, which provides an effective technical support for further optimization and subjective evaluation.