Research On Fault Diagnosis Of Motor Bearing Based On Parameter Optimization VMD And Convolution Neural Network

As the most common rotating mechanical equipment,electric motor plays an important role in the modern power system.Restricte d by harsh working environment,the motor is prone to failure,which will challenge the stability of the whole system.The rolling bearing fault is one of the important reasons for the failure of the motor.It is of great practical significance to carry out fault diagnosis research on it for maintaining the healthy operation of the motor equipment,ensuring the smooth operation of the system and improving the economic benefits.This paper takes the motor rolling bearing as the research object,decomposes the fault signal using the optimized variational mode decomposition(VMD)method,and carries out the secondary extraction of the fault characteristics,establishes the fault diagnosis model of convolution neural network combined with bidirectional long and short time memory network,sets two conditions of noise interference and variable load,and verifies the effectiveness of the proposed method in practical application compared with other methods.The main research contents are as follows:Firstly,in response to the difficulty in extracting features from bearing vibration fault signals,this paper introduces the VMD algorithm with strong noise resistance and high decomposition accuracy.In view of the shortcomings of VMD algorithm itself,the influence of the different selection of decomposition modulus and penalty factor on the decomposition results is verified through simulation,and the envelope entropy is used as the fitness function of the optimization algorithm.Finally,two optimization algorithms,Hayeis Hawks Optimizer(HHO)and Sparrow Search Algorithm(SSA),were introduced separately.Through comparative analysis of optimization simulation using actual data,the SSA-VMD algorithm was determined as the main decomposition algorithm in this paper,and the optimal parameter values were obtained through optimization.Secondly,in response to the problem of how to classify and diagnose the extracted fault feature components,this article conducts secondary feature extraction on the optimized VMD decomposed modal components and visualizes the feature parameters.Based on data characteristics,a DCNN BiLSTM fault diagnosis model is constructed by combining Convolutional Neural Networks(CNN)with Bidirectional Long Short term Memory(BiLSTM).Finally,in order to prove the availability of the method proposed in this paper in practical applications,the bearing fault diagnosis experiments of four different methods are carried out respectively in the simulated setting of variable noise and variable load conditions.The experimental results show that the VMD-SSA-DCNN-BiLSTM bearing fault diagnosis method proposed in this paper has higher diagnostic accuracy,better anti-noise ability and stronger adaptability to variable load compared with other methods.