Reduction of a structural noise by application of the wavelet transform with level-dependent thresholds

Abstract

In this study de-noising results of the ultrasonic echo-signals using the discrete wavelet transform with level-dependent thresholds is presented. The proposed technique has been applied for simulated and measured ultrasonic signals, obtained from austenitic steel sample and glass fiber reinforced plastic composite pipe. It was obtained, that “Symlet-10” mother wavelet using “Minimax” threshold selection rule provides significant improvement for ultrasonic grain noise reduction in the case of detection of intergranular stress corrosion cracks in austenitic steel samples (in the case of numerical simulation). However, the detection of the defects in composite plastic pipe using the proposed technique for de-noising experimentally obtained signals is enough complicated. The signals reflected by regular discontinuities, like interfaces between adjacent layers of the multilayered structure covers the useful signals from the internal defects (like side drilled holes (SDH)). In order solve these problems, the different type of signal processing technique should be used, for example like the proposed by authors in [17].