Application Of CEEMDAN Method In In-vehicle Noise Source Identification Technology

Transfer path analysis is an important research method for identifying noise source technology in vehicles,but the test vehicle is vulnerable to the interference of external factors in each path in its input side under working conditions,and the collected working condition data have certain noise,which in turn has an impact on the accuracy of the transfer path model.In order to xccchave a more accurate representation of the noise source signal in the later analysis of the contribution to the target point,the CEEMDANWT combined noise reduction method is used to reduce the noise of the transfer path condition data and improve the accuracy of the model.The CEEMDAN method decomposes the noisy signal into multiple intrinsic modal components with frequencies distributed from high to low,and the adaptive noise addition feature can avoid the occurrence of modal mixing in traditional methods.Meanwhile,in order to obtain more accurate noise,the sample entropy is introduced to quantize the components,and the noise reduction is supplemented by wavelet method.By comparing the images of the two threshold functions,the soft threshold function is smoother and more continuous,and has better noise reduction effect.By comparing the two thresholding functions,the soft threshold function is smoother and continuous,and has a better noise reduction effect.The different methods are used to reduce the noise of the simulated signal separately,and the SNR of the CEEMDAN-WT combined noise reduction method is 29.6798 and the RMSE is 0.1540,which is the best noise evaluation performance index,indicating the best noise reduction effect.A domestic pure electric vehicle is used as the test vehicle to build the OTPA model of motor mounts,motor,gearbox,and driver’s right ear side.The collected time-domain operating data are decomposed by singular value decomposition technique to eliminate the crosstalk between the inputs,and the model is verified by the measured response and synthetic response at the target point and the contribution amount analysis is done.The noise source signals were all noise-reduced,and the comparison of the image results showed that the measured response of the driver’s right ear side after noise reduction was in better agreement with the fitted response OTPA model,with SNR of 16.514 and RMSE of 0.0126.In addition,the contributions obtained from the original OTPA model were in three main frequency bands: 34.56Hz-43.25 Hz,73.59Hz-83.42 Hz,73.59Hz-83.42 Hz,131.71Hz-148.36Hz;the main frequency ranges after the noise reduction method are:36.78Hz-40.95 Hz,75.66Hz-80.29 Hz,135.95Hz-141.25 Hz,and the main frequency interval range is obviously reduced.It indicates that the noise reduction method is better applied in the noise source identification,and has a reference significance for the subsequent vehicle NVH optimization.