Search

Aiming at the issue that most of the existing knowledge graph-based methods for personalized learning resource recommendations do not take full advantage of collaborative signals from learner interaction data, we introduce a novel model named Collaborative Knowledge Graph Attention Network-based Learning Resource Recommendation Model (CKALR). Firstly, instructional resources attribute is utilized to structure a knowledge graph, then naturally combines the explicit collaborative signals from learner-learning resource interactions with the auxiliary knowledge associations in the graph. At the same time, an attention method is employed to accurately obtain the individual preference implied in the learner’s past interaction information, thereby further enriching the feature representations of both learner and the learning resources. Finally, we compute the inner product of the representations to estimate the user’s preference for a given learning resource. The experiments are performed using the publicly available learning resource datasets, MOOCCube and Book-Crossing, and evaluated using metrics AUC, F1 score and Top-K evaluation metrics. The results show notable enhancements in both accuracy and interpretability when compared to other benchmark algorithms.

Accurate classification of moss species is essential for progress in ecology and biology. However, traditional methods for classifying moss require significant expertise, and current deep learning techniques struggle due to limited dataset diversity and poor performance in multi-class classification tasks. To overcome these challenges, we proposed the Swin Routiformer, a new algorithm for moss image classification that enhances the Swin Transformer with bi-level routing attention. Addressing the issue of limited data, we constructed a dataset with images of 110 different moss types. Additionally, we propose the Crop-Similar data augmentation algorithm, specifically designed for moss images, to reduce background noise interference and prevent information loss due to feature scaling. Adopting the Swin Transformer model with its multi-level hierarchical architecture for visual feature extraction, we introduce the Swin Routiformer Block, which enhances the network’s feature interaction capabilities, reduces computational complexity, and improves classification accuracy and image processing speed for moss species. Our experimental results show that the Swin Routiformer achieves a top-1 accuracy of 82.19% and an f1-score of 82.79% on the test set, outperforming most mainstream models by 4.53% and 1.81% respectively compared to the baseline Swin Transformer model. These findings establish the Swin Routiformer as a valuable tool for the precise identification of moss species, offering significant contributions to the related fields.

The box-counting dimension, which can effectively reflect the complexity and self-similarity of models, is an important method for calculating the fractal dimension of models. To improve the precision of the box-counting dimension algorithm, in this study, we propose an improved 3D box-counting dimension algorithm for models composed of triangular meshes. Through linear coding and optimization of intersection detection between triangles and octree cells, we can accurately calculate the box dimension of models with less memory expense. Through the results of measuring regular geometric bodies and spatial Voronoi bodies, it can be seen that the method has good performance, and the ability of filling space of fractal bodies can be calculated accurately. In the circumstance of 3D box-counting dimension algorithm cannot work well for comparing curvature changes of non-fractal surfaces, the weighted box-counting dimension algorithm can be used to quantitatively analyze the complexity of surface curvature.

Train network control management system (TCMS) is an important part of the High-speed rail train. Because of the TCMS's complex and redundant structure, long-term operation environment, etc., breakdowns inevitably in the long-time running. Based on the historical fault data of the TCMS accumulated during their online service, the working principles, failure modes, and effects analysis of TCMS are researched and the dynamic fault tree (DFT) model of TCMS failure is built. Then, the dynamic fault tree model is transformed into the Bayesian network (BN) model, which can model the reliability of such types of systems. Finally, combining DFT with BN is used for fault probability estimation and reliability assessment. The results present that increasing the reliability of key modules for the TCMS would be of great help to High-speed rail train engineers in the fault diagnosis field.

A reverse-blocking AlGaN/GaN metal-insulator-semiconductor high electron mobility transistor (RB-MISHEMT) is proposed and fabricated. Compared with the conventional MISHEMT with ohmic drain, the proposed device features a hybrid Schottky-ohmic drain with a low work function Tungsten (W), based on which the state-of-the-art ultralow turn-on voltage ( ${V} _{\mathrm{ on}}$ ) of 0.25 V could be realized without degradation in on-state characteristics. In addition, the fabricated RB-MISHEMT exhibits the excellent reverse blocking voltage of −1332 V (at ${V} _{\mathrm{ GS}}= 0$ V) and forward blocking voltage of 1315 V (at ${V} _{\mathrm{ GS}} = -15$ V) with a specific on-resistance ( ${R} _{\mathrm{ on,sp}}$ ) of 3.5 $\text{m}\Omega $ cm2, leading in the highest power figure-of-merit (FOM) of > 494 MW/cm2. The good thermal stability could also be observed in fabricated RB-MISHEMT. The corresponding operation mechanism of RB-MISHEMT are also revealed by Silvaco ATLAS simulations. These results demonstrate the great potential in power electronics applications.

For a high-speed train, the braking system plays an essential role in safe transportation. Efficient state monitoring and anomaly detection may provide useful information for real-time decisions. With large amount of monitoring data, data-driven methods, especially deep learning methods are widely adopted for anomaly detection in various industrial applications. Although deep learning methods have advantages in discovering non-linear relations among complex and high-dimensional data, the large amount of hyperparameters can be hardly well tuned with a high computational burden for onboard computers. Therefore, in this work an efficient online anomaly detection model based on Broad Learning System (BLS) is established for detecting anomalies in the braking system. Furthermore, considering the intrinsic imbalanced data size on anomaly and normal states, the cost-sensitive learning method is integrated in the BLS model, for the first time. The proposed model is evaluated on real data collected from a high-speed train operating for one year, with respect to two performance metrics, i.e. G-mean and F1-score. Comparisons with benchmark neural networks and the combinations of sampling methods and BLS are also considered in this work.

In the process of 3D modeling of geological body with fault, it is the primary problem to calculate the intersecting line between fault plane and various geological models. To quickly locate geological models intersecting fault planes, this paper adopts the algorithm of intersection detection of hybrid bounding boxes. Firstly, the principal component analysis is used to generate the bounding box of the tightly bounding geological model. Then, based on the separation axis theorem, an improved algorithm is proposed to realize the intersection detection of hybrid bounding boxes, which effectively reduces the number of detection and improves the efficiency. Experimental results show that this algorithm can greatly shorten the time of intersection test, and improve the efficiency of intersection calculation. This paper provides an algorithm basis for quickly locating the geological model cut by fault plane.

This paper presents a novel variable speed limit control system under the vehicle to infrastructure environment to optimize the freeway traffic mobility and safety. The control system is a multiagent system consists of several traffic control agents. The agents work cooperatively using the proposed distributed reinforcement learning approach to maximize the freeway traffic mobility and safety benefits. The traffic mobility objective is to maintain freeway traffic density slightly under the critical point to produce the maximum traffic volume, while the traffic safety objective is to reduce the speed difference between adjacent segments. The merits of distributed reinforcement learning are its model-free nature, and it can improve its performance continually as time goes on. The control system is developed on an open source traffic simulation software. Results revealed that compared with no control cases, the proposed system can noticeably decrease the total travel time and increase the bottleneck outflow. Moreover, the speed difference between freeway segments indicating the potential rear-end collision risk is significantly reduced. We also found that there could be more than one optimal traffic equilibrium according to different control objectives, which inspire us to design more optimal strategies in the future.

Increased wind power penetration and load demands lead to an increase in the peak-valley difference, which challenges system stability. This paper presents a hierarchical control strategy to coordinate multiple units of aggregated air-conditioning (AC) loads for power regulation to help reduce the peak-valley difference. First, the independent system operator (ISO) determines the purchased reserve capacity according to the roughly expected regulation effect, and the range of adjustable instructions is derived based on the purchased reserve capacity to help reduce the computation burden. Second, the economy optimizations between the ISO and load aggregators (LAs) are carried out to determine the distribution of purchased control units for AC. Third, after implementing unified planning by the ISO, the LAs deliver the scheduling instructions, including the triggering times and unit numbers, to the AC units to achieve short-term power regulation. Additionally, the scheduling accuracy can be set by changing the scale of the AC control units according to the requirements of the instructions. The simulation results show the effectiveness of the proposed control framework in Lianyungang, with 1000 control units participating. The method proposed in this paper could relieve the burden of the peak-valley difference effectively with less economic expenditure.

It is always difficult to evacuate crowds in public places like subway stations. The traditional crowd behavior simulation models often ignore two important issues in crowd evacuation: pedestrian tracking and individual differences. To solve the problem, this paper combines social force model (SFM) with deep learning into a novel pedestrian detection method. Firstly, several deep learning algorithms for pedestrian detection were compared, and the best ones for sparse and dense crowds were determined. Next, the pedestrian positions in a real video were acquired by the selected algorithms, and converted into actual coordinates in the scene. Then, the evacuation process was simulated with our method and the SFM based on these coordinates. The results show that our model output closer-to-reality results than the SFM. The research findings shed important new light on evacuation in crowded areas.

This paper proposes a self-learning enhanced energy management for plug-in hybrid electric bus (PHEB). Two innovations are made to distinguish our contributions from others. First, a target preview-based reference SOC plan method is proposed, where a linear reference SOC trajectory will be re-planed at fixed time steps (every 60 s) by taking the expected SOC at the destination as the target, and a feasible zone is exclusively defined to provide a margin for the fuel economy improvement. Second, a reinforcement learning-Pontryagin's minimum principle (RL-PMP)-based energy management constituted by the RL and PMP algorithms is proposed, where the average velocity with time moving, the SOC, and the normalized trip distance are taken as states and the co-state is taken as the action. Different from the conventional RL methods, the co-sate is only updated at the fixed time step, and the reward will be evaluated in the next 60 s. This can greatly accelerate the training process of the RL-PMP. In addition, the driving conditions in the next 60 s are predicted by a designed Markov chain. The simulation results demonstrate that the fuel economy of the RL-PMP is close to the dynamic programming (DP) and can be averagely improved by 14.155% compared to the rule-based control strategy.