Inspection of weld penetration depth in thin structures using laser-generated lamb waves and emat receiver

Weld penetration depth;Laser/EMAT ultrasonic;Lamb waves;Transmission coefficient;Continuous wavelet transform;Neural network

Yang, Lei ; Ume, I. Charles Mechanical Engineering Melkote, Shreyes N. Sabra, Karim Michaels, Jennifer E. Zhang, Chun (Chuck) ; Ume, I. Charles

University:Georgia Institute of Technology

Department:Mechanical Engineering

关键词: Weld penetration depth;    Laser/EMAT ultrasonic;    Lamb waves;    Transmission coefficient;    Continuous wavelet transform;    Neural network;   

Others  :  https://smartech.gatech.edu/bitstream/1853/55503/1/YANG-DISSERTATION-2015.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

Using laser/EMAT ultrasonic technique to inspect penetration depths of welds in thin structures is limited by lack of effective signal interpretation methods. The laser-generated ultrasounds in thin structures are very complicated. This research conducts a systematic study to investigate how to effectively apply laser/EMAT ultrasonic technique to measure penetration depths of welds in thin structures. First, the signals acquired during inspections are characterized using 2-D Fourier Transform and wavelet analysis to establish a good understanding of laser-generated Lamb waves in thin structures. Secondly, interactions of laser-generated Lamb waves with welds are investigated by measuring full wave field signals and 3-D Fourier Transform analysis. Thirdly, feasibility of using transmission coefficients of Lamb waves to predict weld penetration depths is studied both numerically and experimentally. A continuous wavelet transform-based algorithm proves effective to fast calculate transmission coefficients of Lamb waves from single acquired signals. Finally, welds of varying penetration depths are manufactured to test the transmission coefficient-based method. Lamb waves sensitive to weld penetration depths are selected and their transmission coefficients are used to develop a neural network to predict weld penetration depth in thin structures accurately. Some considerations regarding the implementation of the developed method are also discussed.