Research On Helical Fault Diagnosis Model Based On Data-driven Incremental Mergence

In recent years,the rapid development of information technology and industrial Internet of things has promoted revolutionary innovation and breakthroughs in manufacturing.Taking the “industrial 4.0” of German as the representative,many countries have successively introduced various measures to attract the backflows of manufacturing and improve the level of manufacturing intelligent.As a global manufacturing center,China has put forward the action plan of first decade,"Made in China 2025",to implement the strategy of manufacturing power.Accelerating the intelligent manufacturing is the strategic measure to implement the deep integration of industrialization and informatization and the creation of manufacturing power.It is also the key to China's manufacturing industry keeping pace with the world's development trends and achieving transformation and upgrading.It can be seen that the use of big data and other advanced technology concepts to promote the development of intelligent manufacturing has become a general trend.With the deep integration of information technology and intelligent technology,it has become easier to acquire the massive data constantly generated during the running process of large mechanical equipment,which makes the efficient diagnosis and prediction of fault type by data analysis method have become an interesting and demanding research topic in the field of intelligent manufacturing.However,in the face of massive new running data,the traditional machine learning method cannot meet the needs of real-time processing.More importantly,the states and properties of the equipment will change over time,and the potential information of newly generated data has more important value to the current state and future trend of equipment.Although incremental learning could constantly learn new knowledge from newly generated data while keeping most of the learned knowledge,it will affect the diagnosis effect seriously when the data stream of incremental generation with the characteristics of massive volume,imbalanced,strong noise and strong causality isn't well treated in the field of fault diagnosis.In view of the above problems,a fault diagnosis method with incremental learning mechanism is proposed based on the characteristics of mechanical equipment fault data.In the proposed method,a dynamic deep learning model based on incremental mergence(IMDDL)is proposed firstly by using deep learning to solve the problem of fault modes extraction and classification for massive new data;Second,the imbalanced data processing and effective example selection model are introduced on the basis of this technology to solve the commonly imbalanced problem of fault data and the strong causal correlation problem of fault;Finally,a helical fault diagnosis method based on data-driven incremental mergence(IMH)is formed,which could effectively overcome the problems caused by the characteristics of equipment fault data with massive volume,imbalance,strong noise,non-stationary,strong causality.The experimental results of bearing running state data demonstrate that the proposed method could not only realize the extraction and transmission of effective information for fault data and guarantee the real-time processing of massive data generated during the running process of equipment by incremental learning,but also solve the problem of significant imbalanced for fault data and eliminate the negative influence of equipment running state data on the accuracy of fault diagnosis by imbalanced data processing.Therefore,the proposed method significantly improves the accuracy of diagnosis and saves the time cost,contributing to meet the requirements of equipment fault data.