Research On Intelligent Diagnosis Method Of Transmission Based On Improved CYCBD

The transmission as a key transmission component in the car,its operating condition is directly related to the safety of the vehicle,if the location of the fault can be accurately predicted at an early stage,you can avoid safety accidents caused by the fault,therefore,the transmission early fault for effective diagnosis,to improve the safety and reliability of the transmission is of great significance.However,in practice,due to the complex structure of the transmission and the interaction between the parts,the collected vibration signals are often the result of coupling together different locations,different levels of faults and other vibration signals on the transmission path.Especially under the influence of strong background noise,the early weak fault features are easily filtered out as noise,making it difficult to accurately extract the early generated local weak fault features.In response to the above problems,this paper analyzes and diagnoses early faults in transmissions by blind deconvolution noise reduction method with an automotive single-gear transmission as the research object.The main research contents of the paper are as follows:By analyzing the common failure types,failure mechanisms and characteristics of vibration signals of gears and bearings,which are key components in transmissions,simulation signals were established on this basis.And by building a transmission fault experimental bench,the vibration signal characterization method based on discrete entropy and time-domain eigenvalues of measurement points is studied,and the best measurement points are determined,which provides reliable experimental data for the early fault feature extraction of the transmission.The study understands the theoretical basis of convolution,focusing on the basic theory of multi-point optimized minimum entropy deconvolution(MOMEDA)and maximum second-order cyclic smooth blind deconvolution(CYCBD)and the influence of parameters on the effect of fault feature extraction.On this basis,for the early faults in the transmission,the fault feature information is relatively weak and easily drowned by the strong background noise,which cannot accurately determine the fault features,and the noise reduction effect of MOMEDA will be affected by the noise and parameters.Therefore,the auto regressive model(AR)based and MOMEDA fault characterization method is proposed.Firstly,the auto regressive model is used to noise reduce the vibration signal to improve the signal-to-noise ratio;then the noise reduced signal is used to perform secondary noise reduction using the improved MOMEDA to further enhance the weak shock components in the signal,as a way to extract the early fault features.However,this method cannot effectively extract the early weak faults in the composite faults.In order to accurately extract the early faults in the composite faults,a combination of EEMD and parametric adaptive CYCBD is proposed for early fault feature extraction.Firstly,EEMD is used to decompose the vibration signal and separate the fault features,and then CYCBD is used to extract the periodic shock components from the processed signal,but the noise reduction parameters of CYCBD cannot be selected adaptively,and in order to improve its noise reduction performance,the chimpanzee optimization algorithm based on the discrete entropy of the envelope spectrum is proposed to adaptively select the parameters of the CYCBD filter,and this method effectively realizes the early local fault feature extraction under the transmission composite fault.