Research On Gearbox Fault Diagnosis Method Based On Random Forest

Gearbox is a power transmission component widely used in mechanical transmission,which plays an important role in the whole mechanical industry.The running state of gear box largely reflects the overall working state of mechanical equipment.Once the gearbox fails,it usually causes the whole mechanical equipment to stop working,resulting in economic losses,safety risks and even casualties.In the running process of mechanical equipment,its vibration and noise signal is the carrier of gearbox fault characteristic information,so it is of great significance to carry out early fault diagnosis analysis and predict fault development trend through the vibration signal of gearbox.In this paper,a gearbox fault diagnosis method based on random forest(RF)is proposed.In view of the nonlinear and nonstationary characteristics of gearbox fault vibration signal and the scarcity of labeled fault samples,a new adaptive signal decomposition method,variational mode decomposition(VMD),is used to extract the features of fault signals.Firstly,the basic principle and steps of VMD algorithm are introduced in detail.Aiming at the problem that two important parameters of VMD algorithm,preset scale K and penalty factor,have great influence on decomposition results,swarm intelligence optimization algorithm is used to adaptively optimize the two key parameters in VMD algorithm.Combined with the experimental data,the results show that the method can extract fault features effectively and accurately.Then,after the initial feature extraction of the original signal,the feature importance score of random forest algorithm is used to further select the fault features with sensitive information.Aiming at the problem that the labeled fault samples of gearbox fault vibration signals are scarce,the sensitive feature data are divided into marked samples and unlabeled samples with different proportions as the input of fault diagnosis classifier model,and the gearbox fault diagnosis classifier model based on semi supervised random forest classification algorithm is established.Finally,the gearbox vibration signal is taken as the research object.According to the data sets of different proportion labeled samples,the improved semi supervised learning self-training algorithm and RF method are compared and analyzed,The improved semi supervised random forest classification algorithm model has a low demand for labeled data,and the proportion of unlabeled data has little impact on its classification accuracy.To a certain extent,it solves the problem of error accumulation in semi supervised learning and the scarcity of labeled samples in gear fault vibration signal.It can effectively use a large number of unlabeled data and achieve high fault Diagnostic accuracy.