Signal processing methods to improve the Signal-to-noise ratio (SNR) in ultrasonic non-destructive testing of wind turbine blade

Abstract Ultrasonic non-destructive testing (NDT) methods are being used quite effectively nowadays, but the multilayered structure of composite materials results in the serious problem in the detection of defects/flaws. The resulting ultrasonic signal is often noisy and denoising of this signal is necessary in order to extract useful information so that faults can be detected, located and sized. Currently, there is a high demand for automatic ultrasonic signal processing techniques to not only remove the need for manual flaw detection and assessment, but also increase the accuracy, reliability and repeatability of the non-destructive evaluation. There are various signal processing techniques which can be used in ultrasonic measurements and selection of appropriate method is one of the major key factors in the field of ultrasonic testing of composite materials. In the presented work, the sample of wind turbine blade (WTB) manufactured using glass fiber reinforced plastic (GFRP) was investigated using ultrasonic NDT in order to estimate the artificially made disbond type defects of 15 mm and 25 mm diameter on the trailing edge. The transmitting and receiving transducers were fixed on movable panel at distance of 50 mm and guided waves (GW) were received at each one millimeter step along the scanning distance of 500 mm. The measurement is performed using low-frequency (LF) ultrasonic system which was developed by Ultrasound Institute of Kaunas University of Technology. Various signal processing techniques were applied to overcome the structural noise and/or extract the information about the defects. The three most promising signal processing techniques: cross-correlation methods, wavelet transform (WT) and Hilbert-Huang (HHT) transform were discussed and compared in the process of defects estimation.