An improved u-net method for denoising ultrasonic echo signals in carbon fiber composites

At present, in the echo signals generated by ultrasonic non-destructive testing of carbon fiber composite materials, there inevitably exist a large number of structural noises, which cause great interference to the identification of defect signals and the detection of defects. To solve this problem, an improved U-Net-based denoising method is proposed to achieve effective noise suppression in ultrasonic signals. When applied to the echo signals of carbon fiber composite materials, the proposed method demonstrates superior noise reduction performance compared to both the conventional U-Net and wavelet denoising algorithms. The signal-to-noise ratio (SNR) is employed as the evaluation parameter for denoising effectiveness, and the results indicate that the proposed approach achieves a higher SNR, confirming its improved denoising capability over conventional U-Net and wavelet denoising algorithms. Furthermore, the denoised signals are stitched in a two-dimensional space to realize the planar reconstruction of defects. The reconstructed defect image is compared with the original un-denoised image and the image denoised by the traditional wavelet method, which demonstrates that proposed algorithm in this paper has stronger denoising effect and robustness.