Output details

Signal denoising and ultrasonic flaw detection via overcomplete and sparse representations

Ultrasonic flaw detection enhancement is a very important issue in ultrasonic non-destructive evaluation (NDE), particularly for highly-scattering materials. This paper enhances the most recent signal processing techniques of sparse signal representations to improve ultrasonic flaw detection. A new highly effective method for noise suppression and flaw extraction from background noise is introduced. Experimental verification shows superior performance for flaw echo detection when ultrasound signals are contaminated with white noise and/or correlated noise by comparison to existing techniques such as wavelet transform methods. This laid the foundation for EPSRC grant EP/G004358/1 and influenced industry practice at Delphi Electronics and Safety.