Research On The Application Of Grating In Bearing Acoustic Emission Signal Measurement

Compared with nuclear power generation and thermal power generation,wind power generation is more green environmental protection,and it has more adequate resources.With the rapid development of wind power industry,the installed capacity of wind turbines increases year by year.It has become the world various countries' main research problems that how to discover the wind turbines failure in time,judge the fault types accurately and ensure the wind turbine generator system runs safely and reliably.In this paper,the wind turbines bearing is taken as the research object,wind turbines bearing fault monitoring system is set up,and the system is used for the research which monitors bearing running status in real time and does fault diagnosis.In this system,a grating sensor is used to extract bearing acoustic emission signals.When bearing is occurring failure,the acoustic emission phenomenon causes the displacement changes of the bearing surface;grating sensor turns the displacement variation quantity into electrical signals;AD9467 collects sensor's output signals;STM32F429 filters the signal,subdivides the signal,and does fault diagnosis for the signal.When failure is occurring,the system has alarm tips function.Furthermore,the system also has data storage function and network transmission function.The resolution of the grating sensor determines the recognition ability of the bearing fault acoustic emission signal.In this paper,in order to improve the accuracy of the system fault diagnosis,the moiré fringe signal needs be high multiple subdivided.Firstly,signal wavelet decomposition and noise reduction are carried out by intelligent wavelet threshold de-noising method.Secondly,for the phenomena that the moiré fringe signal has direct-current level,unequal amplitude,non-orthogonal phase and so on,doing moiré fringes signal compensation can effectively improve the accuracy of subdivision.Thirdly,in this paper,the moiré fringe signal subdivision method based on L-M BP neural network is researched deeply.By adding new judgment conditions to improve the algorithm of L-M,according relationship of the error of the this time training results and the error of the last training results,new weight can be derived.This method can improve the neural network training speed and precision of the results.Comparing the results of it with the results of RBF neural network moiré fringe signal subdivision method,the experimental results show that the moiré fringe signal subdivision method based on improved L-M BP neural network is faster and the error fluctuation range is smaller.Then,the corresponding amplitude in different frequencies can be got through doing spectrum analysis of the displacement values obtained from the subdivision.Finally,comparing the amplitude with the bearing critical failure amplitude,it is possible to judge whether the bearing is faulty.The simulation results and experimental results show that the moiré fringe signal subdivision method based on improved L-M BP neural network can achieve 20000 subdivision,the resolution can reach 1nm,and the acoustic emission signal of the nanometer scale bearing crack failure can be identified.It is shown that the grating sensor which is processed by moiré fringe signal subdivision can be used to extract the acoustic emission signal of the bearing fault,and the bearing crack failure can be diagnosed by wavelet neural network fault diagnosis method.