Noise Source Identification and Experimental Research of Engine Compartment of a Forklift Based on Fast Independent Component Analysis and Scan and Paint

Abstract

The noise signals from the engine compartment of a forklift are wideband non-stationary random signals as their structure and working process are complex. In order to separate and identify the noise sources of the engine compartment, blind source identification analysis was carried out based on fast independent component analysis (FastICA) algorithm and experimental research on noise sources localization were done by using Microflown’s Scan & Paint system. Firstly, a numerical analysis method for effectively achieving noise source identification was proposed. Secondly, the feasibility of FastICA algorithm and the efficiency of the proposed method were verified through simulation. Thirdly, the statistical independence and Gaussian of noise signals were analyzed. The results show that noise signals meet the preconditions of independent component analysis (ICA). Then, noise signals were separated by the proposed method. The corresponding relation between independent components (ICs) and different noise sources was obtained. And the accuracy of the identification results was validated with Scan & Paint sound source localization system. The differences between experimental and numerical analysis results are less than 5%. Finally, de-noising methods are devised based on sound source characteristics.