Train Model Research Articles

Non-hypothetical projection pursuit regression for the prediction of hydration heat of Portland-cement-based cementitious system

In this study, the non-hypothetical projection pursuit regression (NH-PPR) is proposed. The proposed NH-PPR model can predict the hydration heat based on the four cement phases, FA, SL, cement fineness and hydration time. The NH-PPR model is proposed by using the multiple layer iteration method and the non-hypothetical and non-parametric ridge functions to enhance accuracy and solve the problems caused by the parameter selection and the subjective hypothesis. The modeling data set is applied to train model, the testing data set is regressed and fitted into the model, and then the obtained results are compared with the BP model. To further validate the proposed model, another published data set is used to obtain a higher degree of confidence in the prediction. It is shown that the proposed model obtains the better accuracy, stability and versatility, and avoids the parameter selection and subjective hypothesis.

The Analysis of Utilizing Multiple Fences in High-Speed Tracks on the Aerodynamic Characteristics of a High-Speed Train Model

The Analysis of Utilizing Multiple Fences in High-Speed Tracks on the Aerodynamic Characteristics of a High-Speed Train Model

Train/track coupled dynamics model of long heavy haul train based on substructure and parallel computing

To consider the coupled effect on the running safety between elastic track and longitudinal impulse of Long Heavy Haul Train(LHHT), a train/track coupled dynamics model is established by using connection substructure theory. The ballasted track is divided into several segments called sub-tracks: a sub-track includes rail, sleepers and ballast. In the sub-track model, the sleepers and ballast are modelled as lumped mass. The rail is divided into the contact and connection rail. The contact rail is modelled as an Euler beam to reflect the wheel/rail interaction and the flexible vibration of the rail. The connection rail is modelled as a super element to reflect the interaction between adjacent contact rail. To increase the simulation speed, a new parallel computing method is proposed: a train/track coupled dynamics model is divided into different submodule, a submodule includes a sub-track and a vehicle on the sub-track. A submodule is calculated by a single computer core. The submodule is connected by connection rail, couplers and ballast. The advantage of this parallel method is that the load of each computer core is almost uniform. The simulation speed depends on the number of parallel computing cores instead of one core with a particularly large load. Finally, taking the 10,000-ton train as an example, the distribution of coupler force, the derailment coefficient and wheel unloading rate are given during the train braking on a curve, which shows the application and necessity of the train/track coupled dynamics model based on substructure and parallel computing.

Study of the Effective Torus Exhaust High Vacuum Pumping System Performance in the Inner Tritium Plant Loop of EU-DEMO

The requirement for a reduction of the tritium inventory of the European demonstration fusion reactor (EU-DEMO) has led to the active research and development of a continuously working pumping process termed “KALPUREX.” This process foresees the direct recycling of a large fraction of the unburnt hydrogen isotopologues via superpermeation in metal foil pumps during the burn phase. The remaining exhaust gas mixture is pumped by continuously operating, mercury-driven linear diffusion pumps. Diffusion pumps are kinetic high vacuum pumps whose pumping principle is based on the momentum transfer from a supersonic mercury vapor jet to the pumped gas mixture. Like many high vacuum pumps, they feature species-dependent pumping speeds. In the present work, we develop a simplified hybrid model of the high vacuum pumping train in order to estimate the effective pumping speed of the integrated system. The results of this model and its implications on the further development of the vacuum system are discussed for the burn and dwell phases of EU-DEMO.

Study of performance of low-rank nonnegative tensor factorization methods

Abstract In the present paper we compare two different iterative approaches to constructing nonnegative tensor train and Tucker decompositions. The first approach is based on idea of alternating projections and randomized sketching for factorization of tensors with nonnegative elements. This approach can be useful for both TT and Tucker formats. The second approach consists of two stages. At the first stage we find the unconstrained tensor train decomposition for the target array. At the second stage we use this initial approximation in order to fix it within moderate number of operations and obtain the factorization with nonnegative factors either in tensor train or Tucker model. We study the performance of these methods for both synthetic data and hyper-spectral image and demonstrate the clear advantage of the latter technique in terms of computational time and wider range of possible applications.

Simulation analysis of effect of wheel out-of-roundness on dynamic stress of coil springs in metro vehicles

Many Chinese metros experienced numerous fatigue failures of the coil springs in the suspension of vehicles. The field test results presented in our previous work showed that wheel out-of-roundness (OOR) was responsible for the failure of the coil springs. It is believed that the most effective countermeasure to eliminate wheel OOR is to reprofile wheel. The fatigue life of coil springs is an important reference for determining the reprofiling parameters. A sophisticated vibration model for train and track with flexible coil springs and flexible track was established and validated with field test results. The effects of wheel OOR and P2 resonance on the coil spring can be evaluated with this model. The relationship between wheel OOR and coil spring dynamic stresses was determined using multibody simulation (MBS). The fatigue life of the coil springs was evaluated using the parameters of the amplitude of the wheel OOR and the mileage. Under the premise of finding a balance between improving the service life of the coil springs and reducing the cost of reprofiling the wheel, viable proposals were made to solve the problem of coil spring fracture.

Train operation scheduling optimization based on deep reinforcement learning

The impact of time delays caused by unexpected events on trains spreads along the complex and widely distributed area of the high-speed railway network. Large train arrivals and departures are delayed. Serious social impact and huge loss of economic benefits are incurred. To solve the problems, the dynamic scheduling model of a high-speed train is established with the objective of minimizing the total delay time and energy consumption of the train. The DDPG algorithm is used to solve the model. The effectiveness of the algorithm for the dynamic scheduling of the high-speed train is verified through experiments, which provides a good basis for high-speed train dispatchers to make optimization decisions.

Robust Cruise Control for the Heavy Haul Train Subject to Disturbance and Actuator Saturation

This paper investigates the disturbance observer based robust cruise control problem for the heavy haul train with input saturation and disturbances. Both uncertain parameters and actuator saturation are taken into account in the dynamic model of the heavy haul train. To reject the influence of disturbances from the input channel, a linear disturbance observer is proposed to approximate the unknown disturbance, and an augmented system is constructed by combining the train state and the disturbance estimation error. According to the Lyapunov stability analysis method and the guaranteed cost control theory, a sufficient condition for the existence of the composite state-feedback control law and the disturbance observer parameter matrix are obtained, the coupler deviation and the disturbance estimation error are stable at the equilibrium point, and meanwhile the minimization of a given train performance index is ensured. Numerical experiments are provided to illustrate the effectiveness of the proposed approach.

The Node Selection Strategy for Federated Learning in UAV-Assisted Edge Computing Environment

Recently, edge computing plays the important role in internet of things (IoT). However, for the emergency scenarios, the capacity of edge computing is limited. Thus, the UAV-assisted edge computing is introduced for IoT. Meanwhile, it faces privacy disclosure issue that the the data generated on UTs is transmitted to edge/cloud servers for processing. So, the federated learning (FL) is adopted to train model. But due to the difference between UTs or data quality, and the dynamic wireless network, the FL faces the bottleneck of training efficiency and accuracy in UAV-assisted edge computing environment. In this paper, a three-layer FL architecture is proposed. the FL training process is optimized by jointly optimizing the UT selection, the UAV selection, the data set selection and the wait delay selection of one-time training. The convergence gap optimization problem is built by analyzing model training convergence, and the the node selection algorithm is designed. Finally, the extensive simulation experiments are conducted to verify the feasibility and efficiency of proposed algorithm for FL.

Information Leakage in Deep Learning-Based Hyperspectral Image Classification: A Survey

In deep learning-based hyperspectral remote sensing image classification tasks, random sampling strategies are typically used to train model parameters for testing and evaluation. However, this approach leads to strong spatial autocorrelation between the training set samples and the surrounding test set samples, and some unlabeled test set data directly participate in the training of the network. This leaked information makes the model overly optimistic. Models trained under these conditions tend to overfit to a single dataset, which limits the range of practical applications. This paper analyzes the causes and effects of information leakage and summarizes the methods from existing models to mitigate the effects of information leakage. Specifically, this paper states the main issues in this area, where the issue of information leakage is addressed in detail. Second, some algorithms and related models used to mitigate information leakage are categorized, including reducing the number of training samples, using spatially disjoint sampling strategies, few-shot learning, and unsupervised learning. These models and methods are classified according to the sample-related phase and the feature extraction phase. Finally, several representative hyperspectral image classification models experiments are conducted on the common datasets and their effectiveness in mitigating information leakage is analyzed.

Numerical Study on the Influence of Coupling Beam Modeling on Structural Accelerations during High-Speed Train Crossings

The applied mechanical modeling of train and bridge structures is essential in accurately predicting structural vibrations caused by high-speed trains, particularly for the often design-relevant structural accelerations. Considering the interaction effects between the train, the superstructure, and the supporting structure yields more realistic and lower acceleration results compared to simplistic reference models disregarding interaction dynamics. The research presented in this article focuses on modeling single-span girder bridges with a ballasted superstructure as coupling beams. In this approach, the bridge is represented as two vertically coupled beams, with one representing the track (rails and sleepers) and the other representing the supporting structure. Their connection incorporates the stiffness and damping properties of the ballasted superstructure, reproducing its load distribution capacity. A numerical study encompassing a wide range of bridge parameters is conducted, focusing on the calculation of maximum structural accelerations. The results from modeling the bridge as a coupling beam model are compared to those from a simply supported Bernoulli–Euler beam. The excitation is applied as either a moving load or a multi-body model of an Austrian Railjet to evaluate the interdependencies of interaction effects between the vehicle and track and between the track and bridge structure. The comprehensive analysis considers varying bridge span, fundamental bending frequency, mass distribution, structural damping, and coupling stiffness and damping properties to identify correlations with the achievable acceleration reduction using the coupling beam model. Compared to the reference model, the coupling beam model can significantly reduce structural acceleration by up to 80%. Reduction levels primarily depend on the coupling stiffness and critical train speed relative to the bridge’s fundamental frequency, with higher fundamental frequency bridges benefiting the most. These findings provide valuable insights for future structure-specific recommendations for adopting the coupling beam and multi-body train models.

Wear rate and profile prediction of Cu-impregnated carbon strip for high-speed pantograph

The wear status of pantograph carbon strips has a crucial impact on the receiving quality of high-speed trains, and accurate wear rate prediction is essential for timely maintenance and replacement. In this study, the wear mechanism of Cu-impregnated carbon strip is analyzed through micro-analysis, and an improved method for calculating wear rate is proposed by enhancing the original Holm-Archard adhesive wear model. The proposed approach considers the collecting current of the pantograph, burning arc, and contact loss rate of the pantograph, thereby improving the accuracy of wear rate calculation. Moreover, the study validates the relationship between contact resistance and contact force by comparing the model's results with experimental data. Furthermore, the wear profile of MCS is predicted and calculated by considering the wear position of the zigzag catenary on the sliding plate, which provides a more comprehensive analysis of the wear behavior of MCS. To validate the proposed model, actual wear and operation data of an eight-unit high-speed train's pantograph are employed, and the results indicate an average absolute error of 0.19 mm at the deepest wear depth (−16 mm–16 mm) of MCS, which demonstrates the feasibility and effectiveness of the proposed model.

An approach to railway network sections modeling based on queuing networks

The paper is devoted to the study and long-term forecasting of railway network sections operating. We call a railway network section a set of several railway stations interacting with each other and lines between them. We propose a methodology for mathematical modeling of train traffic on the railway network based on the queuing theory. The resulting models are a set of mathematical descriptions of the incoming train traffic and the train running within the system. Several Markovian Arrival Process are used to describe train arrivals from different directions. Such a description makes it possible to take into account the parameters of separate train flows, which depend on the category of trains and their directions. A queuing network in which all nodes have a finite capacity simulates the movement of trains through the system. This mathematical apparatus allows us to consider the nonlinear structure of the railway network, the operation features of stations, the capacity of railway lines, and the influence of random factors. To apply the methodology, we have chosen two objects that differ in infrastructure and properties of train flows. The first is located in the east of Russia and focused on servicing freight trains. The second is located in Germany, close to Belgium and The Netherlands borders. It is characterized by the predominance of passenger traffic. We construct mathematical models and perform numerical simulations. Based on the results obtained, the maximum allowable load is estimated, and bottlenecks in the structure of transport systems are found. Besides, we give some recommendations on how to increase capacity in the long term.

Research on coupled vibrations between the low-to-medium speed maglev train and a long-span continuous beam bridge

This study investigates the strong coupling vibration between low-to-medium speed maglev trains and the track beam, which results from unique transmission principle of the trains. The research focuses on the impact of train parameters on the dynamic response of the car-bridge system, using a long-span continuous beam bridge as the background. Initially, a multi-body dynamics model of maglev trains was established based on the theory of maglev force equivalence as stiffness and damping. Subsequently, the solid element available in ANSYS was employed to construct a finite element model of the large-span bridge. In SIMPACK, the train-bridge coupled system was established to evaluate the dynamic response of the coupling system under various working conditions. The results indicate that the equivalent models of maglev train demonstrates excellent control performance. The vehicle parameters of weight and speed do not significantly affect the bridge vibration characteristics. However, the lateral dynamic behavior of the bridge increases significantly when the train velocity reaches 160 km/h. The operation of double-line trains in opposing directions can result in more complex bridge vibrations, with a greater dynamic response observed. Therefore, the analysis of the coupled vibration of the car-bridge system under complex operating conditions is required in the design phase. Analyzing the coupled vibration of train-bridge systems during the design phase is crucial for ensuring the safe and efficient of maglev trains.

A framework for simulating snow accumulation and ice accretion on high-speed trains

A simulation framework for three-dimensional particle trajectory tracking, snow accumulation and ice accretion modelling on solid bodies in high-speed air flow is presented. The framework, named HADICE, solves the aerodynamic flow field as Reynolds averaged Navier-Stokes (RANS), with optional hybrid RANS/LES capabilities for modelling complex turbulent flows. Particle trajectory tracking is performed in an Eulerian-Lagrangian approach and wall collision is modelled using a hard collision model with reflect condition and a momentum based trap condition. A novel accurate and robust iterative evaluation method for the local collection efficiency is proposed for arbitrary 3D geometry and flow. Ice accretion is modelled in a multi-step approach and iced geometry is updated through mesh deformation using a radial basis function (RBF) interpolation method. Snow accumulation and ice accretion predictions based on the framework are validated against climatic wind tunnel experimental measurements. A full 3D simulation is demonstrated for snow accumulation and ice accretion on a 1:8 scaled high-speed train model. Using the presented framework, snow accumulation and icing simulation for high-speed trains can be conducted in an accurate and efficient manner, which is of great importance for physical investigations and the design of anti-snow and ice protection systems.

Measurement of vehicle speed based on the GCC algorithm and its application in anti-slip control

Measuring the speed of locomotives is important for train operation, especially in anti-slip control. A speed measurement method is proposed based on the correlation of vibration signals and the core process is to estimate the time shift of the windowed signals using the generalized cross-correlation (GCC) algorithm. A sensitivity analysis of the algorithm and suggested reasonable ranges for critical parameters are conducted. The effectiveness of the proposed method is demonstrated using both simulated and tested vibration signals, showing that it performs well under both constant and variable speed conditions. The proposed method had a maximum improvement in root-mean-square error (RMSE) of 2.72% compared to the measurement of the existing cross-correlation (CC) algorithm. The proposed method is integrated into a heavy-haul train model with an anti-slip controller. Simulation results indicate that the method can accurately measure the locomotive speed when partial or all wheels slip.

Investigating transfer learning to improve the deep-learning-based segmentation of organs at risk among different medical centers for nasopharyngeal carcinoma

Purpose: Convolutional neural networks (CNNs) offer a promising approach to automating organ segmentation in radiotherapy. However, variations of segmentation protocols made by different medical centers may induce a well-trained CNN model in one center and may not perform well in other centers. In this study, we proposed a transfer learning method to improve the performance of deep-learning based segmentation models among different medical centers using nasopharyngeal cancer (NPC) data.Methods: The NPC data included 300 cases (S_Train) from one institution (the source center) and 60 cases from another (the target center), divided into a training set of 50 cases (T_Train) and a test set of 10 target cases (T_Test). A ResNet CNN architecture was developed with 103 layers. We first trained Model_S and Model_T from scratch with the datasets S_Train and T_train, respectively. Transfer learning was then used to train Model_ST by fine-tuning the last 10 layers of Model_S with images from T_Train. We also investigated the effect of the numbers of re-trained layers on the performance. The performance of each model was evaluated using the dice similarity coefficient, and it was used as the evaluation metrics. We compared the dice similarity coefficient value using the three different models (Model_S, Model_T, and Model_ST).Results: When Model_S, Model_T, and Model_ST were applied to the T_Test dataset, the transfer learning (Model_ST) had the best performance. Compared with Model_S, the p-values of all organs at risk were less than 0.05. Compared with Model_T, the p-values of most organs at risk were less than 0.05, but there was no significant statistical difference in Model_ST for the brain stem (p = 0.071), mandible (p = 0.177), left temporal lobes (p = 0.084), and right temporal lobes (p = 0.068). Although there was no statistical difference for these organs, the mean accuracy of Model_ST was higher than that of Model_T. The proposed transfer learning can reduce the training time by up to 33%.Conclusion: Transfer learning can improve organ segmentation for NPC by adapting a previously trained CNN model to a new image domain, reducing the training time and saving physicians from labeling a large number of contours.

Retracted: Analysis of Feasibility Modeling of Multimode Physical Fitness Training for Long-Distance Runners

Retracted: Analysis of Feasibility Modeling of Multimode Physical Fitness Training for Long-Distance Runners

Reinforcement learning method for the multi-objective speed trajectory optimization of a freight train

With the rising urge to mitigate the green house effect, reducing the energy consumption of the freight train attracts much attention. Multiple constraints should be taken into account to solve the energy-efficient control problem, which can be reformulated as the multi-objective optimization. This paper proposes a Reinforcement Learning (RL) method for the multi-objective speed trajectory optimization with the goal of the energy-efficiency, punctuality and accurate parking simultaneously. Since the solution space for the optimization problem in this paper is large and discrete, a Gated Recurrent Unit (GRU)-based network is proposed to achieve the fast approximation of the optimal value function instead of the lookup Q-table. Meanwhile, a new architecture, including the embedding matrix, is used to model the control sequence that generates the speed trajectory. Besides, this paper constructs a Deep Q-Network (DQN) framework to train the GRU network without relying on the prior knowledge of the freight train model. Finally, the Intelligent Train Operation (ITO) algorithm is proposed and verified, using the data of Beijing–Guangzhou Railway Line and HXD1B electric locomotive. The case studies indicate that the reward function for the ITO algorithm converges rapidly and the energy consumption monotonically decreases with the trip time, which satisfies the multiple optimization objectives. In terms of saving the energy consumption, the ITO algorithm performs better than Fuzzy Predictive Control (FPC), Genetic Algorithm (GA) and the field test data. The computation time of different speed trajectories demonstrates that the ITO algorithm is applicable to generating the optimal speed trajectory off-line.

Study on characteristics and community structure of high-speed rail network centered on Shanghai

The high-speed railway is a typical representative of China 's high-speed transportation infrastructure, with typical complex network characteristics. Based on the complex network theory, this paper conducts an empirical study on a Space-P network model for high-speed train originating from Shanghai. Then the network is divided into communities by louvain algorithm. The average distance of the network is 1.76, the modularity of community division is 0.3683, and the clustering coefficient is 0.6859, which is typical of small-world network. Due to Space-P model, the network is highly connected, resulting in a relatively low modularity. To a certain extent, the network reflects the developed high-speed rail level and wide direct range of Shanghai, and also reflects the convenience of China 's high-speed rail.