Research On Adaptive Decomposition Of Multi-source Impact Signaland Fault Diagnosis Method For Diesel Engine

Diesel engine is a kind of internal combustion machinery widely used in ships,vehicles,electric power,construction machinery,and other fields.Due to the complex structure,variable operating conditions,and harsh working environment,various mechanical faults occur frequently,including bearing bush wear,valve wear,cylinder liner wear,and connecting rod bolt fracture,etc.The research on diesel engine fault monitoring and diagnosis technology has achieved a lot at home and abroad,but there are still some deficiencies in fault feature extraction,especially in the early weak feature extraction of mechanical faults.The adaptive decomposition and feature extraction of the impact signal of the complex compact shell under the condition of multisource strong interference has been a difficult problem in the field of diesel engine fault monitoring and diagnosis.In the thesis,the diesel engine is taken as the research object,and the aim is to improve the technical level of diesel engine fault early feature extraction,and the research is carried out around the adaptive decomposition and fault diagnosis methods of shell vibration signal in the frequency domain,time domain,and time domain.Thus,operating condition recognition and signal sparse representation is completed,fault impact sensitive features are extracted,fault intelligent diagnosis models are built,and fault data of connecting rod bearing bush wear,abnormal valve clearance,and misfire are used for method application research.The main research work and achievements of the thesis can be summarized as follows:Firstly,the thesis proposes an improved variational mode decomposition(VMD)method to extract the sensitive frequency band of fault impact of diesel engine multi-source impact signals.For VMD method parameter setting,two parameter optimization schemes are designed.Further,the improved VMD method is applied to the simulated signal and the experimental signal of connecting rod bearing wear fault.The results show that improved VMD can effectively decompose and extract the early impact frequency band signal of the fault.Secondly,aiming at the difficulty of multi-source impact signal identification and location in time domain,based on the geometric characteristics of local high energy density distribution of impact signal in time domain,this thesis proposes a variational time domain decomposition(VTDD)method.The VTDD method can complete the adaptive decomposition of multiple impact signal in the time domain,and adaptively identify the number of impacts in the complex signal,the time domain center and boundary of the impact waveform.For the parameter setting problem,three adjustment parameters are designed and the parameter optimization selection methods are proposed.The VTDD method is applied to the simulation signal,abnormal valve clearance fault and misfire fault data,and the results show that the VTDD method can accurately locate and separate the time-domain impacts and fault weak impact components.Then,aiming at the actual demand of the time-frequency domain adaptive decomposition of multi-source impact signals of diesel engines,combined with the characteristics of energy concentration and rapid attenuation of time-frequency amplitude of impact components in the timefrequency domain,this thesis proposes a variational time-frequency decomposition(VTFD)method.The VTFD method performs the adaptive decomposition of multiple impact signals directly in the time-frequency domain.The VTFD method is applied to the simulation signal and the connecting rod bearing bush wear fault signal.The results show that the VTFD method can effectively decompose the multi-source impact signals from the time-frequency domain,and locate and extract the fault impact components at the same time,which verifies the adaptability and accuracy of the VTFD method.Finally,based on the research results of the signal decomposition methods mentioned above,the sparse representation and diagnosis methods for typical faults of diesel engines under variable operating conditions are proposed to address the problem of difficult fault diagnosis under different operating conditions.This method combines the firing impact signals of each cylinder to train and test the diesel engine operating condition recognition model,and the robustness of the operating condition recognition model is tested using misfire fault data.Subsequently,a DS dictionary is constructed based on the decomposition signal(DS)obtained by the VTDD or VTFD method,and the signal is sparsely represented using the orthogonal matching pursuit algorithm.The sparse coefficient is used as the feature value for diagnosing abnormal valve clearance and connecting rod bearing wear faults.This thesis presents an in-depth research on the feature extraction of diesel engine faults and variable operating condition fault diagnosis,with a clear engineering application background and practical demand.The research results have formed new methods of adaptive decomposition of multi-source impact signals in time,frequency,and time-frequency domains,and new models for condition identification and fault diagnosis.The application of simulation and experiment has achieved good results.It provides support for the development and application of diesel engine fault monitoring and diagnosis technology.