Study Of Automobile Alternator Noise Online Detection And Noise Source Indentification System

As people’s living standards improve, people have more expectations on automobile’s performances, especially on its Noise Vibration and Harshness performance. Along with the main noise such as the engine noise, the tire noise and the aerodynamic noise are controlled, the original inconspicuous noise of automobile alternator become more and more prominent. Currently, in aspect of noise online detection, certain automobile alternator manufacturer’s system can only record the sound pressure level versus time, without considering the influence of ambient noise, and the system can’t identify the source of excessive noise.This paper aim at the existing shortcomings of the present noise online detection, using wavelet threshold denoising method, mathematical morphology denoising method, noise and vibration experimental analysis of alternator method and the BP artificial neural network pattern recognition method, develop an automobile alternator noise online detection and noise source identification system. This paper includes the following main contents:Analyze the characteristic of the automobile alternator online detection noise, and select the appropriate pre-processing method according to the characteristic of the noise, simulation analysis shows the importance of the pre-processing method parameters. And apply this method to the measured signal, and achieve good de-noising performance.Comparative Analysis of automobile alternator’s SPL in cold-load condition, hot-load condition and no-load condition; Using step-wise operation method, order analysis result shows the electromagnetic noise, the aerodynamic noise, and the mechanical noise frequency distribution; analyzing the vibration and acoustic signal around the peak speeds, point out the nature root cause of the peak value of the SPL curve.Considering the background noise is different between the testing room and the semi-anechoic room, study the environment correction coefficient of the SPL. BP neural network, which can indicate the type of excessive noise is constructed. Its input is obtained by the top seven order IMF’s normalized energy feature vectors, and its output is the type of the excessive noise.LabVIEW software platform is used to develop automobile alternator noise online detection and noise source identification system. Engineering verification is performed by using the measured data.This paper develop an automobile alternator noise online detection and noise source identification system, make the alternator noise online detection more accurate and achieve the initial identification of noise sources. This paper has some practical value in engineering.