Research On Fault Diagnosis Of Walking Gearbox Of Combine Harvester Based On VMD And KELM

The walking gearbox is an important transmission component of the combine harvester,which is responsible for transmitting the power of the engine to the walking wheels to drive the whole machine forward.The working conditions of the combine harvester in the field are harsh and changeable,and the load on the walking gearbox is complicated and prone to failure.The failure of the walking gearbox will seriously affect the harvesting efficiency of the crops during the busy agricultural season,causing economic losses.Therefore,it is of great significance to carry out condition monitoring and fault diagnosis of the walking gearbox of the combine harvester.This article mainly takes the walking gearbox of the combine harvester as the research object,observes the damage of the gearbox according to the previous fatigue test,analyzes the common failure mode and vibration mechanism of the gearbox,and provides a reference for the subsequent failure test.A feature extraction method based on genetic algorithm(GA)optimized variational mode decomposition(VMD)and a kernel extreme learning machine(Kernel Extreme Learning)optimized based on Whale Optimization Algorithm(WOA)are proposed Machine,KELM)Fault identification method.The main work and conclusions of the specific research are as follows:(1)Carry out common failure analysis of combine harvester gearbox and analysis of typical gear failure vibration signal.In order to understand the possible failures of the walking gearbox of the combine harvester,fatigue tests were conducted to observe the damage in the gearbox after actual work,and the common failure types and gear failure forms of the walking gearbox were analyzed according to the test results.In order to obtain the vibration signal of the faulty gearbox,the gear failure types such as pitting corrosion,chipping,and broken teeth were manually set.On the spectrogram,the fault signals of tooth collapse and breakage are analyzed to compare with the subsequent fault diagnosis methods.(2)Carry out research on feature extraction methods based on GA-VMD.Aiming at the problem of modal aliasing caused by parameter selection in the VMD decomposition method,a genetic algorithm with sample entropy as fitness function is proposed to solve the optimal parameter combination in VMD.Experimental results show that the VMD decomposition parameters obtained by this method can effectively avoid the problem of modal aliasing.After obtaining the optimal decomposition layer number K = 6 and the penalty factor C = 1814,the optimal parameter combination is used to perform VMD decomposition on the vibration signal.Further obtain the sample entropy of each eigenmode component(IMF).The test compares the VMD sample entropy,EMD sample entropy,and wavelet packet energy ratio as the fault signal feature quantity,and the recognition effect in the same classifier.The test shows that using VMD sample entropy as the fault signal feature,the average recognition rate is excellent.Based on EMD sample entropy and wavelet packet energy ratio characteristics.(3)Fault identification method based on WOA-KELM.Aiming at the problem that the KELM kernel parameter ? and the penalty factor C are selected incorrectly,which leads to the low accuracy of KELM,a WOA algorithm is proposed to optimize the two parameters and determine the input parameters in the model.Obtain the sample entropy of each eigenmode component(IMF),together with the time domain and frequency domain features to form a fault feature set,and carry out a comparison test of WOA-KELM and ELM,KELM and other classification recognition models.The model has the best recognition rate,reaching 96.4%.