Research On Grey Prediction Method Of Rolling Bearing Fault Based On Empirical Mode Decomposition

The development of industry is inseparable from the development of science and technology.Machinery and equipment follow the pace of science and technology in the direction of intelligence,complication and automation.It is an important research content to ensure the safe and stable operation of the equipment.Rolling bearing is a special part which is widely used in rotating machinery.Its basic structure consists of two internal and external raceways and a group of rollers.The structure seems simple,the actual structure is precise and the manufacturing process is complex.Because the rolling bearing also has the support function,therefore has the remarkable influence to the entire rotating machinery equipment running state.Therefore,the fault prediction of rolling bearing is carried out.The method has important academic significance and engineering application value,and can effectively avoid the occurrence of serious accidents in the process of equipment operation.The contents of this paper are as follows:First of all,the development history and research status of fault diagnosis and fault prediction technology are reviewed,and the causes and types of rolling bearing faults are analyzed.The calculation method and vibration characteristic of fault characteristic frequency in different states are explained theoretically.Secondly,the theory of empirical mode decomposition(Empirical Mode Decomposition,EMD)is introduced,and the phenomenon of modal aliasing in the components obtained from the decomposition of signals is studied.The phenomenon of modal aliasing shows that the component components after decomposition contain different time scales,which makes it difficult to distinguish the physical meaning of each modal component.In order to provide accurate feature vectors of physical meaning for subsequent fault diagnosis and prediction,this paper presents a method of modal aliasing elimination based on independent component analysis(ICA),which provides a good input feature for fault prediction.This method Firstly,the(MCKD)algorithm of maximum correlation kurtosis deconvolution is used to reduce the noise of the early fault signal to reduce the modal aliasing caused by noise interference.Then the signal is decomposed by EMD method and the component is obtained.Finally,the phase space reconstruction and the negative entropy FastICA algorithm are used to separate the aliasing components of the existing modal aliasing components,and the components with a single physical meaning are obtained.In the end,the theory of grey model GM(1,1)is introduced,but there is residual interference in its prediction value,which leads to its low prediction accuracy.Therefore,the residual error correction is carried out by using extreme learning machine(Extreme Learning Machine,ELM)to improve its prediction accuracy.The ELM-GM(1,1)model was established.On this basis,a grey prediction method of rolling bearing fault based on EMD is proposed.The method uses the improved EMD method to decompose the vibration signals of rolling bearings into a series of intrinsic modes with fault characteristics in time and frequency domain.The function IMF(Intrinsic Mode Function),calculates the RMS value and kurtosis value of the IMF where the fault feature frequency is located as the characteristic index parameter and takes it as the performance degradation quantity to describe the health state of the rolling bearing and forms the series of the performance degradation characteristic quantity.The change trend of fault characteristic quantity of rolling bearing is predicted by training and forecasting model of degraded performance characteristic quantity.The reliability and practicability of the method are proved by experiments.