Research And Application Of The Engine’s Online Abnormal Sound Detection Based On Wavelet Filter And Polar Mirror

Because of the characteristics that most engine’s abnormal sound is detected by the method of human auscultation, which will result in error detection, undetecting, low accuracy and impacting human health if someone working long time. So this paper presents a non-contact method which will use the abnormal sound to analysis the engine. Firstly, filter the signal’s noise which is acquired from the engine’s production line, so as to restore the engine’s inherent characteristics which are masked by noise. Then, mirror the signal in the polar coordinates, to form a symmetrical image; and then average a large number of the same kinds images of engines, to form all kinds benchmark image of engines; finally, after processing the undetecting signal in the methods of filtering and mirroring in polar coordinates, find out the type of the undetecting engine through the maximum value from matching all kinds of benchmark of engines. So as to achieve the purpose of engines’ abnormal sound detection in line.This paper introduces the field of engines’ abnormal sound briefly, to know that the fault diagnosis is moving to the direction to intelligent, convenient and systematic, and if we make use of the characteristics of current computer technology and interdiscipline, the detection of engines’ abnormal sound will have much room for development, which lays the right direction for following study.Then, through the analysis of wavelet theory, we know that the background noise of the engine’s sound signal will be removed effectively after wavelet filtering, and the transient pulse information of the signal will be retained which is consistent with the expectations of engines’ abnormal sound detection. And through the method of mirroring in polar coordinates, the restored time-domain acoustic signal will be symmetrized in polar coordinates which will demonstrate the inherent characteristics by another view, and very easy to identify.Use the noised ‘bumps’ signal to simulate the engine sound signal, in order to test the effect of the wavelet correlate filter method. By comparison with a conventional wavelet filter method, and using SNR and MSE arguments as the evaluation index, it knows that wavelet correlation filter can restore mutation process of the engine effectively, and the result of filtering is good.In the confirmatory analysis of measured engine’s sound signal, we find that the mirroring image in polar coordinates can also detect the engine’s sound signal in some degree without the help of the wavelet correlation filtering, but it can’t recognize the difference between the normal engine and the crankcase with abnormal sound engine, existing the rate of error. After the combine of the mirroring in polar coordinates method and the wavelet correlate filter method, the accuracy of the engines’ abnormal sound detection is high.Finally, an engines’ abnormal sound detecting system based on NI data acquisition and the software of Lab VIEW, MATLAB is applied to the practice.