基于小波包分解与机器学习的汽车调光电机异音识别

Abnormal Sound Recognition of Automotive Dimming Motor Based on Wavelet Packet Decomposition and Machine Learning

为实现汽车调光电机装置异音检测的自动化,本文采用机器学习的方法开展产品异音识别研究。在分析确定产品异音来源的基础上,采集正常件和异音件的振动信号,利用小波包分解,结合时频域分析,在能量谱和时域特征中提取10个特征向量,基于BP神经网络对200个信号样本进行机器学习分类。并对20个样件进行试验,识别汽车调光电机异音的正确率达到96.7%。研究表明,采用机器学习的方法能够有效地识别电机异音,此研究具有工程应用价值。

In order to realize the automation of abnormal sound detection in automotive dimming motor device,the method of machine learning is used to carry out the research of abnormal sound recognition.On the basis of analyzing and determining the source of distorted sound,vibration signals of normal and distorted parts are collected,and 10 feature vectors are extracted from energy spectrum and time domain features by wavelet packet decomposition combined with time-frequency domain analysis.Finally,BP neural network are used to classify 200 signal samples for machine learning.20 samples are tested,and the correct rate of identifying the abnormal sound of the dimming motor is 96.7%.The research shows that the machine learning method can effectively recognize the motor abnormal sound,and the research has the value in engineering application.