Title :
Automatic Defect Identification of Eddy Current Pulsed Thermography Using Single Channel Blind Source Separation
Author :
Bin Gao ; Libing Bai ; Wai Lok Woo ; Gui Yun Tian ; Yuhua Cheng
Author_Institution :
Sch. of Autom. Eng., Univ. of Electron. Sci. & Technol. of China, Chengdu, China
Abstract :
Eddy current pulsed thermography (ECPT) is an emerging nondestructive testing and evaluation (NDT&E) technique and has been applied for a wide range of conductive materials. In this paper, a single-channel blind source separation is proposed to process the ECPT image sequences. The proposed method enables: 1) automatically extract valuable spatial and time patterns according to the whole transient response behavior without any training knowledge, 2) automatically identify defect patterns and quantify the defects, and 3) to provide guidelines of choosing the optimal contrast functions that can improve the separation results. In this paper, both mathematical and physical models are discussed and linked. The basis of the selection of separated spatial and time patterns is also presented. In addition, an artificial slot and a thermal fatigue natural crack are applied to validate the proposed method.
Keywords :
automatic optical inspection; blind source separation; computerised instrumentation; crack detection; eddy current testing; fatigue cracks; feature extraction; image sequences; infrared imaging; materials science computing; thermal stress cracking; transient response; ECPT image sequences; NDT&E; artificial slot; automatic defect identification; automatic time pattern extraction; automatic valuable spatial pattern extraction; conductive materials; eddy current pulsed thermography; nondestructive testing and evaluation; optimal contrast function; single channel blind source separation; thermal fatigue natural crack; transient response behavior; Cameras; Coils; Eddy currents; Heating; Image sequences; Transient response; Vectors; Blind source separation (BSS); eddy current pulsed thermography (ECPT); mathematical modeling; nondestructive evaluation (NDT&E); pattern extraction;
Journal_Title :
Instrumentation and Measurement, IEEE Transactions on
DOI :
10.1109/TIM.2013.2285789