Search

Accurately understanding the temperature field distribution and hot-spot temperature of oil-immersed transformers is of great significance for ensuring the safe and stable operation of the transformer. A magnetic-thermal-flow coupling simulation model for oil-immersed transformer is established in this paper. In order to better simulate the heat dissipation process of the transformer, the winding uses a pie shaped structure with interval arrangement instead of the traditional cylindrical structure. The losses of iron core and winding are calculated through magnetic field simulation. Then, the losses are used as heat source for coupling simulation of thermal field and laminar flow to obtain the oil flow distribution, 3-D temperature field distribution, and temperature and location distribution of hot-spot. Finally, coupling simulations are conducted under different load currents to analyze the impact of load currents on the temperature and location of transformer hotspots. The simulation results show that as the current increases, the hot-spot temperature increases. When the load current changes, the hot-spot fluctuates within the range of 85% - 90% of the total height of the low-voltage winding, providing a methodological basis for the maintenance of transformers and monitoring of hot-spot temperatures.

Few-shot image classification remains a persistent challenge due to the intrinsic difficulty faced by visual recognition models in achieving generalization with limited training data. Existing methods primarily focus on exploiting marginal distributions and overlook the disparity between the product of marginals and the joint characteristic functions. This can lead to less robust feature representations. In this paper, we introduce DBDC-SSL, a method that aims to improve few-shot visual recognition models by learning a feature extractor that produces image representations that are more robust. To improve the robustness of the model, we integrate DeepBDC (DBDC) during the training process to learn better feature embeddings by effectively computing the disparity between product of the marginals and joint characteristic functions of the features. To reduce overfitting and improve the generalization of the model, we utilize an auxiliary rotation loss for self-supervised learning (SSL) in the training of the feature extractor. The auxiliary rotation loss is derived from a pretext task, where input images undergo rotation by predefined angles, and the model classifies the rotation angle based on the features it generates. Experimental results demonstrate that DBDC-SSL is able to outperform current state-of-the-art methods on 4 common few-shot image classification benchmark, which are miniImageNet, tieredImageNet, CUB and CIFAR-FS. For 5-way 1-shot and 5-way 5-shot tasks respectively, the proposed DBDC-SSL achieved the accuracy of 68.64±0.43 and 86.02±0.28 on miniImageNet, 73.88±0.48 and 89.03±0.29 on tieredImageNet, 84.67±0.39 and 94.76±0.16 on CUB, and 75.60±0.44 and 88.49±0.31 on CIFAR-FS.

Additive manufacturing is capable of forming part solids from a digital model of the workpiece, which often needs to be sliced due to the nature of the process of stacking materials layer by layer. Previous algorithms require more time to order the intersection points during slicing and only consider a fixed axis vector as the slicing direction. With the increasing complexity of the workpiece model, it is necessary to improve the efficiency of the slicing algorithm, and the given slicing direction should be flexible and convenient. Therefore, this paper proposes an efficient algorithm for slicing in any direction using the shared edges of adjacent triangles. Firstly, the model can be efficiently sliced in any direction by custom vectors, which is convenient for process validation; secondly, the shared edges of neighboring triangular facets intersecting the plane need to be calculated only once to reduce redundant calculations; finally, the ordered intersections of the constructed inner and outer contours are directly outputted without any additional sorting to reduce the running time. The experimental results show that compared with other algorithms, the proposed algorithm can improve the execution efficiency and eliminate the constraint that the slicing direction is fixed on the Z-axis, which is of application value and reference significance for additive.

Few-shot learning continues to pose a challenge as it is inherently difficult for visual recognition models to generalize with limited labeled examples. When the training data is limited, the process of training and fine-tuning the model will be unstable and inefficient due to overfitting. In this paper, we introduce NegCosIC: Negative Cosine Similarity-Invariance-Covariance Regularization, a method that aims to improve the mean accuracy from the perspective of stabilizing the fine-tuning process and regularizing variance. NegCosIC incorporates a negative simple cosine similarity loss to stabilize the parameters of the feature extractor during fine-tuning. In addition, NegCosIC integrates invariance loss and covariance loss to regularize the embeddings in order to reduce overfitting. Experimental results demonstrate that NegCosIC is able to bring substantial improvements over the current state-of-the-art methods. An in-depth worse case analysis is also conducted and shows that NegCosIC is able to outperform state-of-the-art methods on worst case accuracy. The proposed NegCosIC achieved 2.15% and 2.13% higher accuracy on miniImageNet 1-shot and 5-shot tasks, 3.22% and 2.67% higher accuracy on CUB 1-shot and 5-shot tasks, and 2.13% and 7.74% higher accuracy on CIFAR-FS 1-shot and 5-shot tasks in terms of worst-case accuracies.

Non terrestrial network (NTN) has been regarded as an important component of beyond 5G communication. The 3rd Generation Partnership Project (3GPP) has officially started studies on NTN for the purpose of integrating satellites with New Radio (NR) terrestrial networks. However, the NR defined orthogonal frequency division multiplexing (OFDM) waveform is susceptible to doubly selective fading and nonlinear distortions in the NTN uplink, thus leading to severe performance loss. In this paper, a type of block scalable OFDM (BS-OFDM) waveform is proposed, which has modified subcarrier and sub-symbol segmentation when compared to the conventional NR defined OFDM. With our proposed framework, block scalable discrete Fourier transform spread OFDM (BS-DFT-s-OFDM) and block scalable orthogonal time frequency space (BS-OTFS) can also be implemented. To obtain a comprehensive understanding of different waveforms, performance metrics concerning peak to average power ratio (PAPR), total degradation (TD), bit error rate (BER), and complexity are presented. The simulation results demonstrate that the proposed block scalable waveforms can obtain better performance and lower complexity than existing NR waveforms in different NTN uplink scenarios, and can therefore be promising candidate waveforms for NR NTN.

Multimodal Aspect-Based Sentiment Analysis (MABSA), as an emerging task in the field of sentiment analysis, has recently received widespread attention. Its aim is to combine relevant multimodal data to determine the sentiment polarity of a given aspect in text. Researchers have surveyed both aspect-based sentiment analysis and multimodal sentiment analysis, but, to the best of our knowledge, there is no survey on MABSA. Therefore, in order to assist related researchers to know MABSA better, we surveyed the research work on MABSA in recent years. Firstly, the relevant concepts of MABSA were introduced. Secondly, the existing research methods for the two subtasks of MABSA research (that is, multimodal aspect sentiment classification and aspect sentiment pairs extraction) were summarized and analyzed, and the advantages and disadvantages of each type of method were analyzed. Once again, the commonly used evaluation corpus and indicators for MABSA were summarized, and the evaluation results of existing research methods on the corpus were also compared. Finally, the possible research trends for MABSA were envisioned.

To solve the problem of prostate puncture biopsy in confined spaces, a wire-driven continuum-type biopsy robot is designed based on the bionic principle in this paper. The robot consists of a continuum-type robot arm and a puncture end, where the continuum-type robot arm consists of multiple rigid nodes and four sets of driving wires. The rigid nodes of the continuum structure are used as the skeleton and the four sets of wire wires are used as the driving wires so that each of the two segments of the continuum has two bending degrees of freedom. The kinematics of the robot were solved using a combination of the D-H method and geometric analysis,and the kinematic model of the system was solved using MATLAB software.Finally, the principle prototype was fabricated for experiments. The positioning error of the robotic arm is generally less than 2.5 mm, which has good kinematic performance and can accomplish the demand of prostate biopsy puncture under narrow space.

This paper is devoted to the study of delay-dependent stability of time-varying delay systems. A delay-derivative-partitioning approach is proposed. By constructing an augmented Lyapunov functional that contains two delay-product-dependent terms and using the delay-derivative-partitioning approach, an improved stability condition is established. The derived condition is described as a set of linear matrix inequalities, which can be readily implemented. Finally, four examples are carried out so as to attest the effect and merit of the presented approach.

The target tracking algorithm of mobile wireless sensor networks involves target motion trend prediction and subsequent node guidance. This study aims to solve the problems of global consistency of node information and significant errors in forecasting fast-moving targets’ trajectories through traditional distributed tracking methods in sensor networks. Initially, the average consistency algorithm is used to average the local measurements of each node to achieve global consistency. Then, semantic moving computing of the Internet of Things calculates and analyzes the node movement to support the subsequent movement guidance of nodes and target movement prediction. Finally, the simulation experiment is carried out to evaluate the commonly used target trajectory prediction model. The simulation results show that the node movement algorithm by average consistency can effectively improve the positioning accuracy of the network for moving targets. Besides, the positioning error decreases with the increase of the sensing radius R, the number of moving nodes nm, and the total number of nodes ns deployed in a particular range in a two-dimensional (2D) space. The positioning error after node movement in 2D space is about 20%–30%R lower than that in a static state. After node movement in a three-dimensional (3D) space, the positioning error is about 40%–50%R lower than in a dormant state. When the target moves at a speed greater than 7m/s, the consistency-based moving computing algorithm’s target loss rate and tracking errors are about 0~10% and 1.5%~2% lower than the target tracking algorithm via Kalman Filter. Therefore, the algorithm reported here can precisely track the high-speed moving target. The existing research on point target tracking has problems of insufficient accuracy and robustness. The algorithm proposed here has stronger robustness, reduced data error in multi-node, and more flexible node movements, providing a reference for the subsequent research on distributed point target tracking.