Rolling Element Bearing Fault Prediction Based On Grey Sequential Extreme Learning Machine

With the rotating machinery tending to be large scale,the requirement of safety and reliability of machinery is increasing.Rolling bearing is the crucial parts of rotating machinery,but it is easy damaged.So,running status of rolling bearing influences the operation state of mechanical equipment directly.Fault prediction of rolling element bearing is significant to avoid serious accidents and huge economic losses effectively.Depth research on theory and key technology of the method used in fault prediction of rolling bearing is proposed,the main work is as follows:Through the system analysis of rolling bearing fault prediction method and the research status at home and abroad in this field,it can be found that rapid and accurate bearing fault prediction technology has become a research focus.In the numerous prediction methods,the methods based on data has been widely adopted.This paper mainly introduces the data based fault prediction technology?The vibration signal of rolling bearing has the characteristics of non-stationary and nonliterary,as well as the shortcomings of the classical methods,so the method of grey theory is introduced to predict the development of bearing fault and the modeling mechanism and application scope of grey model were studied.Aiming at the deficiency of traditional grey GM(1,1)model,a series of improved grey models are discussed.In this paper,the grey forecasting model MGM(1,n)model is used to describe the multiple diagnostic indexes from the system point of view.There are many problems in the traditional neural network gradient learning algorithm,such as the long training time,the various training sample.ELM neural network is introduced in this paper,ELM(Extreme learning machine)has the advantages of short learning time,simple and easy to implement,good generalization performance and the ability to avoid being trapped in local optimal.It has been successfully applied to the function fitting and forecasting applications.The basic theory of extreme learning machine,the learning algorithm of feed forward neural network and the learning algorithm of extreme learning machine are successively introduced in the paper.The vibration signals of rolling bearings in practical environment are not easy to be extracted in the noise background,and have both nonlinear and non-stationary characteristics.Through theory analysis and simulation of rolling bearing.we found that EEMD has better anti aliasing effect than EMD in nonlinear signal processing and owns a very big advantage in fault feature extraction.Grey model has advantage on predicting the development sequence,ELM neural network has high non-linear mapping characteristics.This paper puts forward for calculating the combined weight coefficient and combine the two into the grey ELM prediction model,which can describe the complex trend with both deterministic and volatility.For bearing fault prediction,the combination model has the advantages of full use of information,and high accuracy.The RMS containing the fault frequency is used as the fault characteristic vector of the bearing and the input parameters of the predictive model.The results show that the prediction accuracy is higher than that of the single model.