The Experimental Study On Abnormal Sound Fault Diagnosis And On-line Test For Automobile Headlamp Leveler

With the development of electronic technology,the automotive micro-motor products have been widely used as actuators in automotive electrical equipment.The automobile manufacturers have very stringent quality requirements to automotive micro-motor products.However,as a labor-intensive industry,the automotive micro-motor industry still stays in the stage that sorting products' sound quality by human,which not only increases labor costs but also seriously reduces efficiency.It is even more difficult to adapt to the increasingly competitive market.Therefore,this paper takes a headlamp leveler as the research object,carries out the experimental research on the fault diagnosis and on-line detection of the abnormal sound of the automobile headlamp leveler,so as to improve the science and automation of the detection of the abnormal sound of headlamp leveler.The specific research content is as follows:(1)A data acquisition platform based on vibration signal was set up,and the collected signals were analyzed in a traditional analysis of time domain and frequency domain,then the wavelet transform was used to analyze and process the signal;(2)The vibration signals of headlamp levelers are classified into four types of sound quality based on human listening: 1 normal type and 3 abnormal sound types.On the basis of signal analysis,the feature parameters of these types of signals were determined and extracted.There are 3 kinds of statistical parameters in each time domain and frequency domain,and 16 kinds of energy spectrum parameters by wavelet packet decompositon in time-frequency domain.The dimension of this 22 extracted feature parameters was reduced through Principal Component Analysis(PCA),and 90% of the cumulative component of the principal component is reserved for following classification study.(3)A BP artificial neural network of abnormal sound identification project was constructed for the headlamp leveler.The feature set after dimension reducing was selected to form a standard training sample set.The final established neural network model could identify the four types of product sound quality types and the prediction accuracy rate was more than 90%.(4)The specific causes of various types of abnormal sounds were analyzed by combining the vibration failures of typical mechanical equipment of spectrum analysis with finite element modal analysis techniques and crossover test.Finally,the mechanism of different abnormal sound faults was found out,which could provide a theoretical advice for designers to improve the products.(5)Based on Visual C++ programming and MATLAB software,an abnormal sound diagnosis and recognition system was developed.It both has on-line detection and off-line diagnostic functions.When placed the system on the production line,the recognaition rate of abnormal sound was up to 90%,which realized the on-line abnormal sound detection and identification requirements.