Advancing One-Sided Non-Destructive Testing for Denoising: Integrating Deep Learning for Enhanced Defect Detection and Quality Assurance

This paper presents a comprehensive study on the enhancement of one-sided Non-Destructive Testing (NDT) methods—Ultrasonic Testing (UT), Eddy Current Testing (ECT), and Radiographic Testing (RT)—through the application of deep learning models. The primary focus is on improving defect detection accuracy, signal-to-noise ratio (SNR), and reducing the minimum detectable defect size. For UT, a Convolutional Neural Network (CNN) was employed to process deconvolved ultrasonic signals, achieving a defect size detection of 0.8 mm and improving SNR to 35 dB with an accuracy rate of 98%. ECT was enhanced using a Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) layers, reducing the defect size detection to 0.3 mm, increasing SNR to 45 dB, and achieving a detection accuracy of 99%. In RT, a Deep Convolutional Generative Adversarial Network (DCGAN) was used to enhance X-ray image quality, enabling the detection of defects as small as 0.7 mm with an SNR of 40 dB and a detection probability of 97%. These results highlight the potential of deep learning models to significantly improve the efficiency, accuracy, and reliability of NDT processes. Despite these advancements, challenges such as the need for extensive datasets and high computational power remain, signaling areas for future research. Keywords: Non-Destructive Testing (NDT), Ultrasonic Testing (UT), Eddy Current Testing (ECT), Radiographic Testing (RT), Deep Learning, Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Generative Adversarial Networks (GAN), Defect Detection, Signal Processing, Quality Assurance