ABSTRACT This study tackles the challenge of managing midpoint potential imbalance and suppressing current zero‐crossing distortion in Vienna rectifiers under conditions unbalanced load. To address this, an adaptive weight model predictive control strategy is proposed. The research begins by analyzing interaction between maintaining balance mitigating distortion, taking into account influences, such as ripple, sampling inaccuracies, load imbalance. A prediction formulated to predict behavior...

The health management of railway vehicles is crucial to secure safety and efficiency in the long-term operation high-speed trains. Meanwhile, complex components put forward a higher requirement for robustness condition monitoring systems, especially abilities identify unexpected faults. misidentification infrequent faults could lead unpredictable consequences vehicle's safety. This paper proposes novel method detecting train bogie based on Bayesian deep learning. First, Monte Carlo-Based...

The unmanned aerial vehicle (UAV) technique combined with the cellular network is a promising and reliable to provide low-cost, flexible, easy-to-expandable network. In this article, we propose UAV-assisted cooperative transmission (UCTN), where UAVs base station (BS) transmit data jointly through cooodination of software-defined (SDN). As UAV has limited on-board energy, aim maximize energy efficiency (EE) by optimizing UAV–user association, location, BS resource allocation, load allocation...

Feature-based (FB) algorithms for automatic modulation recognition of radar signals have received much attention since they are usually simple to realize. However, existing FB approaches focus on several specific modulations and fail when applied various modulations. To overcome this issue, we propose a effective algorithm based Manhattan distance-based features (MDBFs) in paper. MDBFs new that can be different The main contributions paper as follows. First, represented wavelet ridges, which...

Health monitoring and fault diagnosis of a high‐speed train is an important research area in guaranteeing the safe long‐term operation railway. For multichannel health system, major technical challenge to extract information from different channels with divergence patterns as result distinct types layout sensors. To this end, paper proposes novel group convolutional network based on synchrony information. The proposed method able gather signals similar process these specific groups neurons...

This paper addresses the problem of performance degradation estimation high-speed train lateral damper based on SDS-CNN. The proposed SDS-CNN consists two types convolution modules, i.e., DA-Module and FE-Module, where is used to adjust data dimension map original vibration signals into high dimensional space, while FE-Module employed extract features different frequencies from scales adaptively. Experimental results CRH380A speed validate superiority structure over FCN, MCNN, Time-CNN,...

High-speed train bogies are essential for the safety and comfort of operation. The performance bogie usually degrades before it fails, so is necessary to detect degradation a high-speed in advance. In this paper, with two key dampers on taken as experimental objects (lateral damper yaw damper), novel 1D-ConvLSTM time-distributed convolutional neural network (CLTD-CNN) proposed estimate bogie. CLTD-CNN an encoder-decoder structure. Specifically, encoder part structure consists 1D-CNN module...

The safe and comfortable operation of high-speed trains has attracted extensive attention. With the train, performance train bogie components inevitably degrades eventually leads to failures. At present, it is a common method achieve degradation estimation by processing vibration signals analyzing information contained in signals. In face complex signals, usage theory, such as entropy, estimations not satisfactory, recent studies have more often used deep learning methods instead traditional...

Convolutional neural networks (CNNs) have achieved beyond human-level accuracy in the image classification task and are widely deployed real-world environments. However, CNNs show vulnerability to adversarial perturbations that well-designed noises aiming mislead models. In order defend against perturbations, adversarially trained GAN (ATGAN) is proposed improve robustness generalization of state-of-the-art by training. ATGAN incorporates training into standard procedure remove obfuscated...

High-speed train bogie, the only component connecting body and track, its degradation fault would directly threaten safety of vehicle. However, learning-based diagnosis methods are faced with imbalanced between normal samples samples, which lead to poor performance. This paper provides a architecture for high-speed based on convolutional neural network, critical comparison three representative class balancing techniques, including weighted loss, focal synthetic minority over-sampling...

In the intelligent high-speed railway system, security of deep neural networks-based train bogie fault diagnosis methods is challenged by adversarial attacks, which can mislead model predictions with maliciously designed examples. However, existing do not consider robustness against attacks. To address aforementioned challenge, we propose a novel method called AdvSifter to perform robust for leverages training (AT) guarantee fundamental robustness. Besides, defense algorithm residual...

This paper investigates the data collection in an unmanned aerial vehicle (UAV)-aided Internet of Things (IoT) network, where a UAV is dispatched to collect from ground sensors practical and accurate probabilistic line-of-sight (LoS) channel. Especially, access points (APs) are introduced some unlicensed band improve efficiency. We formulate mixed-integer non-convex optimization problem minimize flight time by jointly designing 3D trajectory sensors' scheduling, while ensuring required...

Ensuring the safe and stable operation of high-speed trains necessitates real-time monitoring diagnostics their suspension systems. While machine learning technology is widely employed for industrial equipment fault diagnosis, its effective application relies on availability a large dataset with annotated data model training. However, in practice, informational samples often insufficient, most them being unlabeled. The challenge arises when traditional methods encounter scarcity training...

To address the challenges in establishing state transfer matrix and complexity of eigenvalue calculation determining multi-parameter stability boundaries high-order nonlinear Vienna rectifiers, a novel numerical computation method is proposed this paper. This leverages criterion grid variable step search to efficiently calculate these boundaries. The small-signal model rectifier derived by constructing time-varying using periodic solution harmonic balance method. Eigenvalues are rapidly...

The fault detection and isolation system is the key element for safe long-term operation of high-speed train. multi-channel signals provided by parallel monitoring are usually closely coupled highly uncertain, which difficult to analyze. This paper proposed a depth-wise convolution modular structure diagnosis with signal address complex dynamic operating conditions trains. A scalable designed provide low coupling high transparency, could easily configurable function-level according...

Deep neural networks have been found to be easily misled by adversarial examples that are maliciously crafted adding small perturbations. A variety of methods proposed generate examples, but more efforts needed them with high perceptual quality and low computation costs. In this paper, we propose an attack method uses a conditional encoder-decoder network named Image-To-Perturbation perturbations in residual learning fashion. can learn the mapping from clean images according perturbations,...

Deep neural networks (DNNs) have been known to be vulnerable adversarial attacks. Adversarial training (AT) is, so far, the only method that can guarantee robustness of DNNs However, generalization accuracy gain AT is still far lower than standard an undefended model, and there a trade-off between adversarially trained model. In order improve performance AT, we propose novel defense algorithm called Between-Class Training (BCAT) combines learning (BC-learning) with AT. Specifically, BCAT...

As early clinical signs of diabetic retinopathy, hard exudates were the most characteristic symptom and essential for detection retinopathy. However, even experienced ophthalmologists, it would be a time-consuming task to identify through digital color fundus photographs retina. Alternatively, automatic diagnostic systems provided efficient services but raised concerns about reliability safety. This paper focuses on developing high-reliability scheme detecting evaluating in under...

Regulating the output current and reducing harmonic distortion are key control problems in airport navigation dimming power supply (DPS). In this study, a new method which could dynamically adjust P I gains is presented for DPS based on radial basis function (RBF) neural networks combined with gradient descent method. To implement proposed method, factors influence performance of revealed by analyzing dynamic response system. The controller design process detail along system stability...