Automated Classification of Ultrasonic Signal via a Convolutional Neural Network

Ultrasonic signal classification in nondestructive testing is of great significance for the detection defects. The current methods have mainly utilized low-level handcrafted features based on traditional processing approaches, such as Fourier transform, wavelet transform and like, to interpret information carried by signals classification. This paper proposes an automatic method via a convolutional neural network (CNN) which can automatically extract from raw data classify ultrasonic collected circumferential weld composed austenitic martensitic stainless steel with internal slots. Experiments demonstrate that our outperforms classifier manually extracted features, achieving accuracy rate up 0.982. Furthermore, we visualize shape, location orientation defects C-scan imaging process results, validating effectiveness results.