Ultrasonic Assessment of Thickness and Bonding Quality of Coating Layer Based on Short-Time Fourier Transform and Convolutional Neural Networks

Ultrasonic non-destructive analysis is a promising and effective method for the inspection of protective coating materials. Offshore exhibits high attenuation rate ultrasonic energy due to absorption pulse echo testing becomes difficult small amplitude second from back wall layer. In order address these problems, an advanced signal has been proposed. An delay line was applied A short-time Fourier transform (STFT) waveform implemented measure thickness state bonding The material estimated by projection STFT into time-domain. debonding layers were distinguished using ratio magnitude peaks two subsequent wave echoes. addition, advantage STFT-based approach that it can accurately quickly estimate time flight (TOF) even at low signal-to-noise ratios. Finally, convolutional neural network (CNN) automatically determine coatings. time–frequency representation used as input CNN. experimental results demonstrated proposed determines coatings with accuracy. present more efficient compared estimating attenuation.