The manufacturing processes of casing rings are prone to multi-type defects such as holes, cracks, and porosity, so ultrasonic testing is vital for the quality of aeroengine. Conventional ultrasonic testing requires manual analysis, which is susceptible to human omission, inconsistent results, and time-consumption. In this paper, a method for automated detection of defects is proposed for the ultrasonic Total Focusing Method (TFM) inspection of casing rings based on deep learning. First, the original datasets of defect images are established, and the Mask R-CNN is used to increase the number of defects in a single image. Then, the YOLOX-S-improved lightweight model is proposed, and the feature extraction network is replaced by FasterNet to reduce redundant computations. The Super-Resolution Generative Adversarial Network (SRGAN) and Convolutional Block Attention Module (CBAM) are integrated to improve the identification precision. Finally, a new test dataset is created by ultrasonic TFM inspection of an aeroengine casing ring. The results show that the mean of Average Precision (mAP) of the YOLOX-S-improved model reaches 99.17%, and the corresponding speed reaches 77.6 FPS. This study indicates that the YOLOX-S-improved model performs better than conventional object detection models. And the generalization ability of the proposed model is verified by ultrasonic B-scan images.

The aeroengine casing ring forgings have complex cross-section shapes, when the conventional ultrasonic or phased array is applied to detect such curved surfaces, the inspection images always have low resolution and even artifacts due to the distortion of the wave beam. In this article, taking a type of aeroengine casing ring forging as an example, the Total Focusing Method (TFM) algorithms for curved surfaces are investigated. First, the Acoustic Field Threshold Segmentation (AFTS) algorithm is proposed to reduce background noise and data calculation. Furthermore, the Vector Coherence Factor (VCF) is adopted to improve the lateral resolution of the TFM imaging. Finally, a series of 0.8 mm diameter Side-Drilled Holes (SDHs) are machined below convex and concave surfaces of the specimen. The quantitative comparison of the detection images using the conventional TFM, AFTS-TFM, VCF-TFM, and AFTS-VCF-TFM is implemented in terms of data volume, imaging Signal-to-Noise Ratio (SNR), and defect echo width. The results show that compared with conventional TFM, the data volume of AFTS-VCF-TFM algorithm for convex and concave is decreased by 32.39% and 73.40%, respectively. Moreover, the average SNR of the AFTS-VCF-TFM is gained up to 40.0 dB, while the average 6 dB-drop echo width of defects is reduced to 0.74 mm.