Research On Typical Fault Diagnosis Of Rolling Bearing Based On DSCNN-LSTM

Due to the continuous advancement of science and technology,the further development of industry has been promoted,and the degree of automation of modern equipment has become higher and higher.Industrial Internet of Things technology is widely used.The era of big data brings opportunities and challenges to the industry.How to make scientific decisions with massive data is an urgent problem to be solved at present.Predictive and health management is a key technology to ensure the safety and reliability of equipment.Predictive maintenance is added on the basis of preventive maintenance to analyze and judge the real-time running state of equipment and analyze the types of failures.The indispensable part of rotating machinery is the rolling bearing,but it is very easy to be worn out,which can lead to equipment failure.Therefore,the health of their operation plays a decisive role in the safe and reliable operation of the whole equipment.Once a failure occurs,from economic losses to devastating human casualties,the damage can be severe.Aiming at this problem,this paper takes rolling bearings,a common part in mechanical equipment,as the research object,and proposes a data-driven fault diagnosis method for rolling bearings.At present,many methods need to use signal processing technology for preliminary processing of vibration signals,but this method does not require signal processing and manual identification.It can automatically complete feature identification and fault diagnosis.Due to its direct application in vibration signals,in order to improve the ability of feature extraction and classification accuracy,this paper proposes a hybrid rolling fault diagnosis method based on deep separable convolutional network and long and short time memory network(DSCNN-LSTM).The main work includes:Firstly,research and analyze the research background in the field of fault diagnosis and the research results of scholars at home and abroad.In order to improve the shortcomings of the existing methods,the research plan of this article is proposed.Secondly,according to the existing theory foundation,a new rolling bearing typical fault diagnosis algorithm,DSCNN-LSTM,is proposed based on the combination of deep separable convolutional network and long and short-term memory network.This algorithm adopts double DSCNN and double LSTM parallel structure,extracted directly by the LSTM hidden features in the original sensor signal.There is no correlation between two paths,but the output of the two paths will affect the learning result of data,and then classify the fault.Then,experiment on the CWRU authoritative public data set provided by Case Western Reserve University.Apply a variety of algorithms to compare the contents of the data set and the data with added noise,and then the accuracy,confusion matrix and other indicators are used to evaluate the performance of the algorithm.The experimental results show that compared with the common fault diagnosis methods,the presented method can extract the characteristics of the deeper,the classification accuracy rate reached more than 99%.Finally,according to the proposed fault diagnosis algorithm,the proposed algorithm is applied to the fault diagnosis system,and a prototype system of rolling bearing typical fault diagnosis is designed and implemented.