Fault Diagnosis Method And Experimental Research Of Ball Screw

Ball screw is widely used in the mechanical equipment fields such as CNC machine tools,industrial robots and aerospace,which has the advantages of high transmission efficiency,high transmission accuracy and long life.Because of extreme operating condition such as high speed and heavy load,the ball screw is vulnerable to be damaged,affecting the accuracy and operation safety of mechanical equipment and leading to serious equipment accidents and huge economic losses.Therefore,fault diagnosis algoruthms on the four typical fault types of ball screw are studied in this dissertation,providing a basis for condition monitoring,equipment maintenance and life prediction of ball screw.The screw pitting fault,the screw wear fault,the screw fixing seat fault and the screw support seat fault were selected as the preset faults for the fault diagnosis test.According to the structural characteristics and motion relationship of the ball screw,the calculation formulas of fault characteristic frequency of screw failure,ball failure and nut failure are derived respectively,and the main characteristic frequency are calculated according to the structural parameters and rotational speed of ball screw used in the test.Analyze the fault vibration signal of the ball screw pair in time domain and frequency domain,and compare the difference between the early fault vibration signal and the fault-free vibration signal.All vibration signals from ball screw fault diagnosis test are preprocessed,including data outliers removal,uniform data extraction,zero mean processing and high frequency noise removal.Aiming at the problem of the early fault features of ball screw pair are difficult to observe and the signal-to-noise ratio is low,an improved adaptive empirical wavelet transform(IAEWT)method based on the scale space theory and Pearson correlation coefficient is proposed.This method can decompose the signals reasonably and effectively,and has better self-adaptive decomposition ability.The improved adaptive empirical wavelet transform method is applied to the fault feature extraction of ball screw,and the time domain feature value,frequency domain feature value and IAEWT decomposition signal component energy value of the ball screw pair are extracted to form the ball screw pair fault feature vector set.The fault feature vector of ball screw pair extracted based on IAEWT method is used as the input parameter of BP neural network,and the fault type of ball screw pair is classified by BP neural network model.A fault pattern classification method based on continuous wavelet transform and two-dimensional convolutional neural network(CWT-2DCNN)is proposed,and it is applied to the fault pattern classification of ball screw pairs.In this method,the frequency spectrum of vibration signal is used as the input parameter of the convolutional neural network model,which can make full use of the fault feature information of ball screw pairs.By comparing the BP neural network model with CWT-2DCNN model,the validity and accuracy of the CWT-2DCNN model proposed in this paper are verified.