Research On Fault Diagnosis Of Gearbox Based On VMD And Extreme Learning Machine

Gearbox is an important mechanical transmission device,and it works in complex working environment such as large load,large torque and high temperature,which leads to frequent failure of gearbox.As the main load-bearing part of the gearbox,the fault of the gear is also the main reason of the gearbox failure,so the fault diagnosis of the gear has great significance.The fault diagnosis of the gearbox in this article mainly focuses on the gears in the gearbox.The specific research contents are as follows:(1)FOA-VMD(Fruit Fly Optimization Algorithm-Variational Mode Decomposition)method was used to reduce the noise of the original signal and enhance the fault characteristics,aiming at the problem that the fault characteristics were not obvious at the low signal-to-noise ratio of the fault signal.Firstly,the K and α parameters of VMD(Variational Mode Decomposition)are optimized by the Drosophila algorithm.Then,the original fault signal is decomposed,and the components with more fault excitation components are selected from the components according to the principle of envelope entropy for reconstruction,and the noise components with less fault correlation and irrelevant are eliminated,so as to achieve the purpose of reducing noise and enhancing fault characteristics.Experiments show that FOA-VMD method can effectively eliminate the noise components in the original signal and enhance the fault characteristics,and the processing effect is better than EMD(Empirical Mode Decomposition)method,which lays a foundation for the subsequent gearbox fault diagnosis.(2)Extract 16 time-frequency index parameters from the noise-reduced signal as the original fault features,and then the feature selection method based on the improved distance method was used to select the best feature in 7 dimensions.Subset,excluding redundant information and fault-independent information in the original fault features.The experimental results show that the training and testing of the learning machine based on the selected fault features can provide the diagnostic accuracy of the model,and the selected fault features are more classified than the original fault features.(3)In this paper,extreme learning machine(Extreme Learning Machine,ELM)is selected as the classifier.Firstly,the activation function and the number of hidden layer nodes are selected through experiments.Then,aiming at the problem that the input weights of elm are determined randomly and no longer adjusted,which leads to low classification accuracy,an improved limit learning machine model based on the maximum entropy principle is proposed for the inverse optimization of input and output weights.The experiment shows that the improved model improves the accuracy of fault classification to a certain extent.In order to improve the robustness of the diagnosis model and further improve the diagnosis accuracy,this paper proposes a gearbox fault diagnosis model with output weighting of multiple limit learning machines on the basis of improving the limit learning machine.Firstly,the output of each limit learning machine model is weighted according to the loss function when the limit learning machine training is completed,and finally the output of each limit learning machine is added The weight is added and the result after adding is regarded as the final diagnosis result.Experimental results show that the diagnostic accuracy of the combined model is significantly higher than that of the single improved model and BP neural network model.