Measurement Science and Technology, Год журнала: 2024, Номер 35(11), С. 116115 - 116115

Опубликована: Авг. 12, 2024

Abstract Metal couplers are susceptible to unpredictable failure and fracture under long-term high-load conditions in heavy-haul railway transportation. The current mainstream manual inspection method has the disadvantages of high subjectivity a priori knowledge requirements, thus not meeting rapid analysis requirements production companies. Therefore, this study, an automated is proposed for coupler fractures. First, novel image segmentation (PermuteNet) combining visual multilayer perceptron convolutional neural network designed segment different patterns surfaces. uses two newly modules—permute attention module context module—to improve network’s ability perceive weakly differentiated objects, thereby improving recognition model patterns. In addition, deep supervisory function adopted accelerate convergence speed network. Finally, deployed on computer conjunction with developed client application implement single-click detection pattern analysis. Experiments performed using dataset established on-site data; achieves mean intersection over union 77.8%, which considerably higher than that other existing methods. By software, area realized. Thus, provides more convenient accurate identification solution factory inspectors broad prospects.

Язык: Английский