Study On Debonding Detection Of Metal Rubber Parts Excited By NLFM Ultrasonic

As a secondary suspension device of bogie suspension system of EMU,air spring has functions of supporting body,cushioning impact,etc.Air springs are subject to complex dynamic loads during the operation of trains.The inner metal rubber bonding structure may produce rubber debonding defect,which threatens the safety of trains.Therefore,accurate and effective detection of rubber debonding defect of air spring is of great importance and significance.In order to solve the problem of debonding detection of metal rubber bonded structure of SYS510 E air spring base plate,a non-destructive inspection method based on ultrasonic pulse reflection was adopted in this thesis.In order to improve the accuracy and reliability of rubber debonding defect detection,the frequency modulation excitation technology of ultrasonic testing and the feature extraction method of debonding defect were mainly studied.Firstly,the advantages and disadvantages of traditional single frequency pulse signal,LFM signal,weighted NLFM signal,phase NLFM signal and constructive NLFM signal were analyzed and compared.According to the difference between the energy distribution of signal and the power spectrum distribution of echo,a constructive NLFM signal was proposed to replace the traditional narrow pulse as the ultrasonic excitation signal.It can effectively increase the signal energy and spectrum width of ultrasonic detection,and enhance the ultrasonic detection effect of rubber debonding.Based on the wide-band ultrasonic detection by LFM excitation,several time-frequency feature extraction methods were analyzed.For plane metal rubber bonded structures,the timefrequency energy distribution of echo in different bonding states and frequency range was analyzed by wavelet packet-singular value decomposition method.The frequency domain matrix was established and the identification characteristics of rubber debonding defects with good stability and consistency were extracted.For inclined metal rubber bonded structures,the CEEMDAN-mutual approximate entropy decomposition method was used to adaptively decompose the time domain waveform of the echo at different bonding states and different frequency IMF components.Through the mutual approximate entropy distribution between IMF waveform and original signal,the rubber debonding defect identification characteristics with quite different were extracted.Based on ultrasonic excitation technology and debonding characteristics,a complete ultrasonic detection system for rubber debonding was established based on the technology of FPGA and Lab VIEW.The system application program consists of calibration training,echo detection and database.In the process of calibration training and echo detection,the system firstly used SVD algorithm to de-noise the signal.After feature extraction,several rubber debonding identification characteristics were trained by BP neural network to realize pattern recognition of rubber ultrasonic detection.Finally,according to the trained BP neural network,the tested samples were detected by ultrasonic C-scan for rubber debonding.The effects of various excitation signals on ultrasonic detection of different types of rubber debonding defects in specimens were analyzed and compared.The results show that ultrasonic wave based on NLFM signal excitation by construction method can accurately and effectively identify rubber debonding defects of various sizes,locations and contours.It has high accuracy of ultrasonic detection.Ultrasound C-scan inspection was carried out on SYS510 E air spring with different structures.The results show that the ultrasonic wave excited by NLFM signal can identify the position and contour of rubber debonding defects of different structures stably and reliably.It meets the requirements of ultrasonic debonding detection for SYS510 E air spring.