Research On The Technology Of Vehicle Fault Diagnosis With Acoustic And Vibration Analysis

During the working process of the vehicle,the vibration signal of the body and the noise signal radiated to the outside can reflect the operating status of the internal components of the vehicle.Analyzing the vibration and noise during the working process of the vehicle can determine whether there is a fault and the location or type of the fault.Among them,tractors that are widely used in the field of agricultural production have exposed parts and strong external sound fields,which are suitable for sound field measurement and analysis.Taking tractor as the research object,this paper puts forward a method of rapid fault detection of tractor based on sound and vibration analysis,which can quickly complete fault detection of tested tractor.In this paper,the current research situation of mechanical fault diagnosis is analyzed and summarized.According to the problems encountered in the research of the tractor’s fault diagnosis,this paper divides the fault detection problem of tractor into two parts:fault diagnosis and fault type identification.The feasibility of the proposed method is verified by the tractor vibration noise signal collected by the vehicle fault diagnosis system.Firstly,the vibration and sound signals of the tractor are filtered,and the improved empirical mode decomposition filtering method is selected as the signal filtering method in this paper.The simulation experiment filtering effect of wavelet threshold filtering,empirical mode decomposition filtering and improved empirical mode decomposition filtering are compared.And the improved empirical mode decomposition filtering method with the best denoising effect is selected as the denoising algorithm in this paper.The improved empirical mode decomposition filtering method is used to filter the actual tractor data.The improved empirical mode decomposition filtering method has stronger denoising ability by comparing the time-frequency domain waveforms before and after filtering.In view of the shortage of normal tractor data samples and the shortage of fault samples in tractor fault detection,a single classification method of support vector data description is proposed for tractor fault judgment.The normal tractor data are train by using support vector data and whale parameter optimization algorithms,and the hypersphere fault diagnosis model is built.Then the state of the sample point is judged by whether the sample point is located inside the hypersphere.The test results show that,using the method of support vector data description to study tractor fault judgment,the fault diagnosis model can be built by using normal tractor data to effectively avoid the lack of fault type samples.Due to the lack of fault samples in the sample bank,the model built directly by support vector machine will misjudge the samples of unknown fault types when the type of tractor failure is studied.And the identification of fault types is poor.Therefore,this paper uses the method of combining support vector data description and support vector machine to identify the fault type.First,determine whether the fault sample to be tested belongs to the known sample in the sample library through support vector data.Then the known fault samples are identified by support vector machine.The test results show that,When the support vector data description is combined with the support vector machine for fault type discrimination,the diagnostic accuracy of known fault types can be improved,and the unknown fault samples outside the sample library can be identified.For unknown fault samples,incremental learning method is added in this paper.For the samples of unknown fault types,the new class identification method is used to judge whether the faults belong to the new type to increase the sample library of fault types and perfect the fault diagnosis model.