Search

Rail defect detection is crucial to rail operations safety. Addressing the problem of high false alarm rates and missed detection rates in rail defect detection, this paper proposes a deep learning method using B-scan image recognition of rail defects with an improved YOLO (you only look once) V3 algorithm. Specifically, the developed model can automatically position a box in B-scan images and recognize EFBWs (electric flash butt welds), normal bolt holes, BHBs (bolt hole breaks), and SSCs (shells, spalling, or corrugation). First, the network structure of the YOLO V3 model is modified to enlarge the receptive field of the model, thus improving the detection accuracy of the model for small-scale objects. Second, B-scan image data are analyzed and standardized. Third, the initial training parameters of the improved YOLO V3 model are adjusted. Finally, the experiments are performed on 453 B-scan images as the test data set. Results show that the B-scan image recognition model based on the improved YOLO V3 algorithm reached high performance in its precision. Additionally, the detection accuracy and efficiency are improved compared with the original model and the final mean average precision can reach 87.41%.

should be removed

This paper presents a data and mechanics integrated approach (DMIA) to analyze small-radius rail life caused by both severe rail defects and excessive rail wear. First, the Archard wear model, as a mechanics-based model, is adopted to calculate the wear growth chart under different radii. Second, based on the growth chart, the observed over-threshold rail lifetime is adjusted to estimate the lifetime when it reaches the wear threshold. Third, considering the influence of different curve radii and the segment length, an accelerated failure time (AFT) model, as a data-driven model, is used to analyze the heterogeneity of the corrected measurement data. A survival probability is obtained with increasing gross tonnage for different curve radii. Finally, the expected MGT is calculated for rail defect & rail wear-based renewal for curves with different radii, and the obtained values are compared with the corresponding values in the existing standard.

Automatic air brake systems are widely used in freight train braking to ensure railway operation safety. Various types faults pose an enormous threat to freight operations. Existing algorithms lack a unified framework for generating key features. In this research, we propose a novel feature engineering framework for the fault diagnosis of freight train air brakes. First, experimental data are collected through a three-car in-lab experimental platform. Second, a peak detection method combined with first-order difference function to partition and classify the air pressure time series into the braking phase and releasing phase. Third, a divided-and-integrated framework is designed for feature engineering. Feature selection is carried out via a modified reinforcement learning method. Finally, multiple machine learning algorithms are explored and the results indicate that random forest method shows the best performance. The proposed model achieves about 99% accuracy for car-level fault detection and over 94% accuracy for component-level fault diagnosis.

This study developed a multi-classification model for vehicle interior noise from the subway system, collected on smartphones. The proposed model has the potential to be used to analyze the causes of abnormal noise using statistical methods and evaluate the effect of rail maintenance work. To this end, first, we developed a multi-source data (audio, acceleration, and angle rate) collection framework via smartphone built-in sensors. Then, considering the Shannon entropy, a 1-second window was selected to segment the time-series signals. This study extracted 45 features from the time- and frequency-domains to establish the classifier. Next, we investigated the effects of balancing the training dataset with the Synthetic Minority Oversampling Technique (SMOTE). By comparing and analyzing the classification results of importance-based and mutual information-based feature selection methods, the study employed a feature set consisting of the top 10 features by importance score. Comparisons with other classifiers indicated that the proposed XGBoost-based classifier runs fast while maintaining good accuracy. Finally, case studies were provided to extend the applications of this classifier to the analysis of abnormal vehicle interior noise events and evaluate the effects of rail grinding.