Ultrasonic flaw detection using split-spectrum processing combined with adaptive-network-based fuzzy inference system

In ultrasonic nondestructive evaluation, in order to successfully detect flaw echoes corrupted by scattered random echoes, a robust and efficient method is required. In this paper, a method utilizing split-spectrum processing (SSP) combined with an adaptive-network-based fuzzy inference system (ANFIS) has been developed and applied to ultrasonic signals to perform the signal classification task. SSP can display signal diversity and is therefore able to provide the signal feature vectors for signal classification. ANFIS maps signal feature vectors to outputs according to an adaptive learning process and fuzzy If-Then rules. The combination of SSP and ANFIS can perform both ultrasonic flaw detection and signal classification. The SSP-ANFIS method has been tested using both simulated and experimental ultrasonic signals, and the results show that SSP-ANFIS has good sensitivity in detecting ultrasonic flaw echoes in the presence of strong clutter when the signal-to-noise ratio is about zero dB.