Loader Working Pump Noise Control Based On Independent Component Analysis

With the rapid development of China’s industrialization and urbanization, construction machinery is playing a predominant role in railway construction, infrastructure construction and industrial and agricultural production. Although domestic construction machinery products has developed a lot in the last few decades, gaps still remain compared to developed countries in terms of performance and service life of the products. As noise level is one of the most important criterions that used to evaluate the performance of construction machinery products, many countries and regions have established related standards, to serve as market entry thresholds. Thereby, in order to improve the performance and expand the market of construction machinery products, more noise control research should be done. As we know wheel loader is one of the most widely used construction machineries, it has many noise source components including but not limited to engine, transmission box and hydraulic pump, and the noise transfer paths are very complex. In this case, we must find the primary noise source(s) and suppress it/them to lower the overall noise level of the machine. Thereby, identifying the main noise source correctly is the premise and key to carry out targeted and efficient noise control.This paper took a 5t wheel loader as the research object, and complete machine noise test was carried out under hydraulic working condition. It turned out that the field sound power level of the loader under hydraulic working condition is 113.34dB(A), a bit higher than the limitation 113dB(A) required by Chinese GB16710 standard, and well above euro EN474 limitation 107dB(A). Moreover, hydraulic working (gear) pump was identified as the primary noise source by using 1/3 octave analysis method, spectral analysis method and frequency identification method.This paper pointed out that separating/extracting a specified noise source from an "unknown" mixed noise model is a blind source separation (BSS) problem, and independent component analysis (ICA) is known as one of the most successful algorithms used to solve the problem. Currently, it is very easy to separate a linear instantaneous mixing model (also called instantaneous ICA model) and get highly accurate estimations of independent sources. However, in most practical engineering situations, real signals have time-delay effect and multi-path effect during transmitting (especially acoustical signals), which fits convolution ICA model rather than instantaneous ICA model, and frequency-domain FastICA algorithm was proposed to solve the convolution ICA model. After being verified by simulation analysis for feasibility, the frequency-domain FastICA algorithm was used to separate/extract gear pump noise (9 teeth) from the data recorded in bench test, three noise sources, namely gear pump noise, electric motor noise and noise of other accessories (such as relief valve) were separated successfully, and it turned out that average sound pressure level of the 9-tooth gear pump is 87.92dB(A).This paper briefly studied forming mechanism of gear pump noise, and two strategies, gear teeth number increasing and gear profile modification were adopted to control the noise, then tests were performed on the same test bench and the same ICA algorithm was used to extract the pump noise, to verify the noise control effect. It turned out that after increasing teeth number of each gear to 12, average sound pressure level of the pump dropped 0.81dB(A); after longitudinal (crowned teeth) gear profile modification and improving the gear precision grade from 9 to 7, average sound pressure level of the pump dropped 1.53dB(A) further, and 2.34dB(A) in total compared to the original 9-tooth pump.