For bearing fault diagnosis,in addition to considering the relationship between spatial structure,the differences of data in different time dimensions should also be analyzed,and deep convolutional recurrent neural network(DCLSTM)should be established by comprehensive use of convolutional neural network(CNN)and long short-term memory network(LSTM)to achieve excellent temporal and spatial characteristics.The sensor sequence features can be fully utilized.Using the adaptive learning method of the initial signal to determine the feature parameters,it can effectively avoid the influence of subjective factors in the process of feature selection and classification.The results show that compared with the traditional method,the accuracy of DCLSTM model is more than 99％,and the fluctuation range is small,showing high stability.DCLSTM network has the highest stability,the corresponding standard deviation is 0.59％,and the fault diagnosis effect is better than the traditional method.DCLSTM diagnostic accuracy reached the highest level for each motor load,with a fluctuation range of only 0.64％.Support vector machine(SVM)and CNN fluctuated by 2.8％and 5.99％respectively,indicating that DCLSTM network can realize the requirement of high precision diagnosis for faults in various working conditions.
NETL
NSTL
万方数据
