Noise Source Identification Technology Research Of Loader Based On The EEMD-ICA-CWT Approach

Wheel loaders are widely used in construction area. It has mounts of advantages, such as a fast operation speed, high efficiency, good mobility. It has a broad market prospect at the same time. However, compared with foreign products, domestic wheel loaders generally have the problem of excessive noise. With the implementation of the newest national standards about the noise limits of earth moving machinery, the noise level will be more strictly. To effectively control the noise of their products, should be the key of the products upgrading and the competitiveness enhancement for the companies. On the other hand, sound source identification technique based on signal processing had been carried out in terms of noise source identification of mechanical equipment. But the signal processing methods had their own problems, need to be improved immediately. To carry out the noise source identification research of wheel loaders based on modern signal processing methods has very important theoretical significances and application values.The basic algorithms of empirical mode decomposition(EMD), independent component analysis(ICA) and continuous wavelet transform(CWT) were studied. For the modal aliasing phenomenon of EMD method which leading to physical meanings loss of the components and the obvious defect when dealing with a single-channel signal of ICA, the author proposed a method based on the combination of ensemble empirical mode decomposition(EEMD) and ICA. EEMD and ICA methods both had their advantages. EEMD method was adopted to obtain a series of intrinsic mode functions, which weakened the model aliasing problem. At the same time, it increased the dimensions of a single-channel signal which could overcome the problem when dealing with a single channel signal of ICA. Then, the correlation coefficients between the independent components and simulated signals were introduced as evaluation parameters. A group of simulation signals similar to loader noise signals were used to evaluate the recovery performances of EMD-ICA and EEMD-ICA methods based on the proposed evaluation parameters. It verifies the EEMD-ICA method had better recovery characteristics of the source signals, which was more suitable for the noise source identification of loaders.In order to separate noise sources of LG 953 loader, the EEMD-ICA-CWT method was used to study the blind source separation and noise source identification based on the radiation noise of loader. The CWT method was used for its better time-frequency localization features to analysis time-frequency characteristics of the ICA results. Combining the results with different noise sources frequency spectrums, the physical sources as the fundamental crux of loader were accurately identified.In a word, the EEMD-ICA-CWT method for blind source separation based on a single channel signal was proposed. It overcame the problem that the number of sensors must be larger than or equal to the number of separated components of ICA method. At the same time, it weakened the model mix superposition problem which usually came out in processing signals with the algorithm of EMD. Finally, combining the CWT method which used for its better time-frequency localization features, the characteristics of the noise sources were identification.