Operation optimization for modern subway trains usually requires the speed curve optimization and speed curve tracking simultaneously. For the speed curve optimization, a multi-objective seeking issue should be addressed by considering the requirements of energy saving, punctuality, accurate parking, and comfortableness at the same time. But most traditional searching methods lack in efficiency or tend to fall into the local optimum. For the speed curve tracking, the widely applied proportional integral differential (PID) and fuzzy controllers rely on complicated parameter tuning, whereas robust adaptive methods can hardly ensure the finite-time convergence strictly, and thus are not suitable for applications in fixed time intervals of trains. To address the above-mentioned two problems, this paper presents a novel approach for speed curve seeking and tracking control. First, we present the random reinforcement genetic algorithm (GA) algorithm to avoid the local optimum efficiently. Then, a sliding mode controller is developed for speed curve tracking with bounded disturbance. The Lyapunov theory is adopted to prove that the system can be stabilized in the finite time. Finally, using the real datasets of Yizhuang Line in Beijing Subway, the proposed approach is validated, demonstrating its effectiveness and superiorities for the operation optimization.

The intrusion recognition along the railway is still a challenging problem in railway safety monitoring based on the phase-sensitive optical-time-domain reflectometer (Φ-OTDR), because the access scope is expanding between operation and external environment, and there are many unknown vibration sources interfering at different fiber locations, which are unpredictable and changing from time to time. To reduce the nuisance alarm rate (NAR) of the system, a novel data driven identification method based on XGBoost model is proposed in this article. We have completed a real implementation from the sensor to the final output, including data collection, signal processing (framing, denoising), feature extraction, model designing and evaluation, which is a successful case of distributed optical fiber sensing in the railway sector. The results indicate that the average recognition accuracy of this article is as high as 98.5% in frequently external intrusion events. All the related performance metrics of confusion matrix are better than other popular methods, such as random forest, support vector machine and multilayer perceptron.

Recently, deep learning methods are becomingincreasingly popular in the field of fault diagnosis and achieve great success. However, since the rotation speeds and load conditions of rotating machines are subject to change during operations, the distribution of labeled training dataset for intelligent fault diagnosis model is different from the distribution of unlabeled testing dataset, where domain shift occurs. The performance of the fault diagnosis may significantly degrade due to this domain shift problem. Unsupervised domain adaptation has been proposed to alleviate this problem by aligning the distribution between labeled source domain and unlabeled target domain. In this paper, we propose triplet loss guided adversarial domain adaptation method (TLADA) for bearing fault diagnosis by jointly aligning the data-level and class-level distribution. Data-level alignment is achieved using Wasserstein distance-based adversarial approach, and the discrepancy of distributions in feature space is further minimized at class level by the triplet loss. Unlike other center loss-based class-level alignment approaches, which hasto compute the class centers for each class and minimize the distance of same class center from different domain, the proposed TLADA method concatenates 2 mini-batches from source and target domain into a single mini-batch and imposes triplet loss to the whole mini-batch ignoring the domains. Therefore, the overhead of updating the class center is eliminated. The effectiveness of the proposed method is validated on CWRU dataset and Paderborn dataset through extensive transfer fault diagnosis experiments.

It is a practical challenge to provide reliable and efficient communication for data transmission in wireless sensor networks. To recover from packet losses, conventional approaches tried to use retransmission or FEC mechanisms. However, these mechanisms may introduce excessive energy overhead for reliability guarantee. By exploiting the wireless broadcast nature and the node overhearing capability, we propose a novel cooperative communication scheme EECC to improve data transmission performance for wireless sensor networks. In this scheme, cooperative reply is performed at each hop by the best-suited node elected from those that have successfully overheard the transmitted packet. EECC is not a routing protocol but rather works as an augment to minimize the impact of packet losses on network performance. Extensive analytical and experimental results confirm that our scheme is very effective in improving both energy efficiency and end-to-end delay for data transmission in lossy networks.

Electronically controlled pneumatic(ECP) brake is widely used in heavy-haul train. Although the latest data-driven fault diagnosis can exploit the collection data from the braking system, it still has challenges for effective fault diagnosis model because of industrial data noise and insufficient fault samples. This paper proposes a fault diagnosis model based on multi-dimensional feature fusion and Gradient Boosting Decision Tree(GBDT) enhanced classification (MFF-GBFD). Firstly, the multi-dimensional features are extracted. By principal component analysis and feature fusion, the redundant features are eliminated. Then the model is trained under ensemble learning framework with boosting strategy. Experiments are carried out on the data from the ECP braking system of DK-2 locomotive. The efforts show that the proposed MFF-GBFD model presents better performances as a result from the early-stage feature extraction, feature selection and feature fusion. It also has higher accuracy and F1 values compared with the traditional classification algorithms.