Research On Intelligent Fault Diagnosis Method For Industrial Big Data

With the speedy technology development and increasing pace of modern life,mechanical equipment and production systems have been playing an increasingly important role in national economic and social development.On the one hand,these advances have indeed met the objective requirements for the development of modern large-scale industries,such as increasing production efficiency,reducing production costs,and saving energy,and have achieved tremendous social and economic benefits.However,on the other hand,the stability and robustness are also more important.Due to a variety of inevitable factors,equipment and systems may suffer from a variety of failures,which may cause malfunction of the system,reduce the work efficiency at times,and cause production to be suspended,resulting in undeniable property damage and even casualties.In response to this situation,all countries in the world have paid great attention to the fact that people are more and more urgent about the requirements for system safety,stability,long-term,and full-load operation.Through monitoring and predicting the state trend,they hope to know the status of the system in a timely manner and achieve the purpose of early detection and early detection of faults,and eliminate the occurrence of major accidents.This will also benefit the growth of the system's service life and reduce the number of system downtime.Therefore,fault diagnosis of equipment is considered an important application and it is widely used to evaluate the reliability and safety of various mechanical equipment.With the rise of artificial intelligence since the 1980s,intelligence has become the mainstream and direction of research in the field of fault diagnosis.The traditional diagnostic process with signal detection and processing as the core has been replaced by a diagnostic process centered on knowledge processing.Given the uncertainties in the types of equipment and the variety of failure modes,it is obviously impractical to use only one specific method to perform fault diagnosis of all equipment.For this reason,this paper first proposed three unique intelligent fault diagnosis algorithms based on BP neural network,support vector machine and deep learning,and applied them to the fault diagnosis of rolling bearings,rocket engines and an industrial motor pump respectively.By comparing and studying the experimental results of three single fault diagnosis algorithms,the paper finally presented a hybrid two-stage pipeline framework.In our method,we assemble support vector machines(SVM)and deep neural networks(DNN)in cascade manner with dynamic weighted features.In the first stage,SVM is used to classify different operating conditions.In the second stage,for degraded and ambiguous examples,DNN is adopted to identify the specific fault.Our adaptive framework is evaluated using real machine data.The framework aims at the specific diagnostic problems of equipment,and through reasonable selection of fault features and network models,quickly build a nonlinear mapping relationship between the symptoms of the fault and the cause of the fault,so as to achieve a high level of intelligent fault diagnosis.The experimental results show that the performance of the proposed WSVM-DNN algorithm is significantly improved compared with the previous three simple algorithms.Due to the constraints of various factors,there is still a long way to go before the full realization of fault diagnosis intelligence,but applying artificial intelligence to fault diagnosis undoubtedly injects unlimited prospects and possibilities into this field.It will be the main direction of future development in this field.