Ultrasonic Material Crack Detection With Adaptive LMS-Based Wavelet Filter

Ultrasonic non-destructive testing has been widely used in assessment of the integrity of engineering materials and structures such as pipelines, high-temperature alloys, bridges, and other load-bearing structures. The ultrasonic echoes from these structures are usually noisy. Effective noise reduction techniques are needed in order to obtain accurate assessment of the health condition of the inspected materials. In this paper, a new noise reduction method for ultrasonic signal processing is proposed, based on wavelet analysis and the least mean squares (LMS) filter. It is designed for effective identification of the ultrasonic signal reflected from the tip of a crack. It is especially useful when the properties of the ultrasonic crack signal are unknown and the noise is heavy. The performance of this technique for SNR enhancement is evaluated using simulated ultrasonic signals. Experimental data obtained from a steel plate with a crack produced by an electrical-discharge-machine (EDM) are used to validate the proposed method. The results show that this method is suitable for processing noisy ultrasonic signals for crack detection in steel plates.