Railway wagon brake shoe fault detection method based on deep learning

Abstract

The invention provides a rail wagon brake shoe fault detection method based on deep learning. The rail wagon brake shoe fault detection method comprises the following steps: acquiring a train image to be detected; constructing a brake shoe fault data set, wherein the fault data set is divided into a training set, a verification set and a test set; constructing a target detection and fault identification network by using a deep learning technology; pre-training the target detection and fault identification network on a public data set, and optimizing network parameters as a final algorithm model; and applying the algorithm model to the railway wagon to detect the brake shoe fault. Compared with a traditional railway wagon brake shoe fault detection algorithm, the method can overcome the problems of picture illumination conditions, shooting angles and the like, and has high robustness and accuracy; meanwhile, the fault can be accurately identified and positioned only by using a single-stage target detection method through a confrontation training method and the improvement of a Fusion structure, the real-time performance and the accuracy of detection are considered, and the labor intensity of detection personnel is reduced.

Description

technical field [0001] The invention relates to the technical field of fault detection, in particular to a fault detection method for railway freight car brake shoes based on deep learning. Background technique [0002] Traditional railway freight car brake shoe fault detection, most of them rely on the method of manually viewing the image of passing trains for fault detection, which has the problems of low detection efficiency and high labor cost. On the one hand, the inspection will lead to the fatigue of the inspectors, on the other hand, there are hidden dangers of missed inspections and false inspections. The use of computer-aided fault automatic detection method can reduce the workload of vehicle inspectors and improve the efficiency and accuracy of detection. [0003] Traditional computer-aided railway wagon fault detection methods rely on classic machine learning and image processing methods. For example, use edge extraction, Hough transform and other image process...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: Aiming at the problems that the current railway freight car brake shoe fault detection relies on labor, high cost, and the traditional computer-aided detection method has low efficiency a

Images