Abstract:To analyze the saturation and coupling characteristics of SynRM and realize inductance identification with small disturbance and low error, an online inductance decoupling identification algorithm for SynRM is proposed. The influence of magnetic saturation and coupling on the voltage and flux linkage equations is first described to interpret the saturation and coupling characteristics of the inductance, and a decoupling motor model is developed by introducing a coupling angle. This model enables the analysis of saturation and coupling effects from a decoupling perspective. Then, an online identification strategy based on a virtual-axis equivalent impedance model is designed to identify both the coupling angle and inductance in real time. The proposed method is validated on a 3 kW SynRM experimental platform under various operating conditions. Experimental results demonstrate that the proposed algorithm effectively realizes online inductance decoupling identification, with identification errors for both the coupling angle and inductance within acceptable limits. Moreover, the inductance decreases with the increase of current, and the coupling angle increases with the increase of current. The changing trends of coupling angle and inductance identification results also verify the accuracy of motor saturation and coupling characteristic analysis. Compared to other inductance identification algorithms, the proposed algorithm does not require high chip computing power. While simplifying inductance calculations, it can also follow motor control in real time and output accurate values.

Abstract:To understand and investigate the impact of summer thermal reflection from glass curtain walls on the microclimate of surrounding environments, and to raise awareness of this rarely acknowledged issue, which is often overlooked due to its invisibility, this study employed optical simulation, scale modeling, and field measurements to identify reflection zones and quantify thermal reflection effects. Firstly, the principles, variation patterns, and potential hazards of thermal reflection from curtain walls are explored using optical mirror reflection principles. Secondly, taking Chunxi Road Plaza and its surrounding curtain wall buildings in Chengdu as an example, optical simulation and experimental modeling methods were used to determine the hourly positions of thermal reflection areas on the square. Through on-site thermal environment testing and comparative analysis, the impact of curtain wall thermal reflection was quantified. Finally, a series of strategies were proposed to control curtain wall thermal reflection and mitigate its hazards. The results indicate that compared to normal areas, the average radiant temperature in thermal reflection areas was higher by 4 to 13 ℃, air temperature was elevated by 1 to 2 ℃, humidity was reduced by 5% to 10%, and the universal thermal climate index was increased by 2 to 5 ℃. These findings confirm the significant influence of summer curtain wall thermal reflection on microclimate environments and pedestrian thermal comfort. Therefore, incorporation of thermal reflection management strategies into urban planning and curtain wall design is imperative to reduce its negative impacts.

Abstract:To investigate the robustness and trustworthiness of neural networks under security threats, this study focuses on their vulnerability to poisoning attacks. Based on a systematic analysis of the characteristics of type I and type II adversarial attacks, and in light of the structural deficiencies in neural network feature learning, the concept of type I poisoning attack is proposed. Theoretical modeling and analysis demonstrate fundamental feature-level distinctions between type I poisoning attacks and existing methods, such as “clean-label” or feature collision poisoning. A type I poisoned sample generation framework is built based on supervised variational autoencoders, and experiments on widely-used deep neural network architectures including ResNet50, VGG16, and MobileNetV2 are conducted. Results demonstrate that the proposed type I poisoning method effectively disrupts model classification decisions while preserving label consistency, successfully inducing misclassification across typical neural network architectures. Moreover, the defense experiments reveal that type I poisoning attacks can bypass existing mainstream defense mechanisms, rendering current primary countermeasures ineffective. With its strong stealth and disruptive capabilities, type I poisoning represents a novel security threat worthy of in-depth investigation. The development of this attack methodology holds significant implications for building more secure and robust neural network systems in the future.

Abstract:To address the challenges faced by existing temporal graph representation learning methods in fully exploiting structural features and dependencies of the network, a time-constrained walk fused memory enhanced temporal graph neural network (TWMTGN) is proposed. Firstly, a specific type of temporal constraint walk sequence is constructed for the interaction node, and the memory module is used to capture the long-term dependence of the network. The walk sequence features are fused into the memory of interaction node, which is dynamically updated upon the occurance of events. Secondly, a feature attenuation layer is designed according to the node type and time interval to model short-term dependencies, which improves the model′s ability to identify key historical interaction nodes. Finally, the aggregated historical interaction features of the target node are fed into a causal convolutional network to further uncover the potential association between them. Experimental results on real datasets show that the proposed network can improve the performance of temporal link prediction tasks while maintaining relatively low complexity. Parameters such as the length and frequency of time-constrained walks affect the mode′s performance. The proposed time-constrained walk sequences can effectively capture the structural features of the network, while the node memory and feature attenuation layers help to capture the network dependencies.

Abstract:In order to solve the issue of multi-bit upset in nano-scale SRAM (processes below 100 nm) in aerospace applications, this study optimizes and improves traditional serial encoding and decoding methods based on multi-bit upset （MBU） characteristics of nano-scale SRAM. A parallel encoding and decoding approach is employed to implement a reinforcement method based on RS(2,8,4) code, enabling encoding and decoding outputs within one single clock cycle. The effectiveness of this reinforcement method in terms of delay and error correction capability is validated based on an FPGA platform. The test results show that, compared to the built-in Hamming code of Xilinx Block RAM, the proposed method has an equivalent output delay but with 5 to 8 times greater error detecting and correcting capability than those of Hamming code. Furthermore, when compared to encoding and decoding methods such as FUEC-QUAEC, CLC, the correction rate for consecutive 5-bit upset errors is elevated to 100%. Using parallel coding and decoding method, the implemented RS(2,8,4) code is effective for reinforcing multi-bit upset in nano-scale SRAM. At a minimum latency cost, it allows for the correction of any two symbols (up to 8 bits) within a single codeword (48 bits), fully correcting errors involving consecutive 5-bit upset within a single word in space radiation environment. The proposed MBU reinforcement method can be extended to external memory control interface or internal cache of CPUs addressing the issue of multi-bit upset errors in caches of existing aerospace processors that rely on single-error correction codes.

Abstract:To address the issues of large repair degree, high repair bandwidth of parity nodes, and the inability to achieve fast repair of multiple nodes in existing piggybacking codes, a construction scheme of substripe-added piggybacking (SAP) codes is proposed in this paper. Based on maximum distance separable (MDS) codes, the proposed SAP codes extend the substripe, embed the data blocks of the information nodes by region regularly, and place the data blocks of the parity nodes using cyclic shifts. Through theoretical derivations, the average repair bandwidth rates and average repair degree rates of information nodes and parity nodes in SAP are determined. Finally, SAP is compared with RSR-I, RSR-II, and OOP in terms of three aspects: storage overhead, repair bandwidth overhead, and repair degree. The results show that, compared with RSR-I, RSR-II, and OOP, the SAP coding scheme not only achieves optimal repair degree but also significantly reduces parity node repair bandwidth while maintaining low information node repair bandwidth. Additionally, it enables rapid repair of multiple parity node failures, effectively addressing the issue of excessively high repair bandwidth in multiple parity node failures. The SAP coding proposed in this paper significantly improves the data recovery efficiency of piggybacking codes. In particular, a fast repair algorithm is provided to address multiple parity node failures, offering an effective approach for optimizing the piggybacking codes.

Abstract:In order to solve the limitations of existing infrared and visible images fusion algorithms in preserving pixel-level information and extracting semantic features, an infrared and visible image interactive fusion method based on semantic driven was proposed. First, the image fusion network and the image segmentation network were jointly operated to form a semantic-driven effect, enhancing the retention of information features of the image in both pixel domain and semantic domain. Then, a cross-domain interactive integration module was constructed to capture features of infrared and visible images, allowing for the interactive transfer of features across different spatial locations and independent channels, thereby mapping features from local to global, and enhancing the complementary characteristics of the two types of images. Finally, a semantic loss function was introduced to constrain the network training, preserving the intrinsic semantic features of the source images. Pixel-level fusion experiments and semantic-level segmentation experiments were conducted on multi-band data sets and multi-spectral road scene data sets. These experiment results were then compared with six other advanced fusion algorithms. The results of fusion experiments show that the proposed algorithm achieves improvements of 47.92%, 6.15%, 0.87%, 44.31%, 35.99% and 36.88% across six objective evaluation metrics, including gradient-based similarity measures, information entropy, peak signal-to-noise ratio, spatial frequency, standard deviation and visual fidelity. The results of segmentation experiments indicate that the proposed algorithm outperforms all other evaluation metrics. Therefore, the proposed method exhibits superior performance in both qualitative analysis of subjective visual effects and quantitative indicators of quality evaluation compared to existing algorithms. The fusion images effectively balance both visual quality and high-level semantic tasks, thereby enhancing utility for human visual observation and machine vision perception.

Abstract:In order to analyze the thermal effects of soil under lightning strikes and describe the thermal diffusion process caused by lightning strikes in soil, a lightning strike soil model is established based on the electro-thermal coupling theory in this work. The model, by considering the nonlinear electric-thermal characteristics of the soil, can reflect the thermal effect. The transient computation was performed to investigate the characteristics of lightning current and the influence of soil characteristic parameters on thermal effects. In addition, the difference between the nonlinear electrothermal characteristics model and the traditional model in simulating the response of the lightning strike soil was compared. Finally, the rationality of the model was validated by the field observation data from previous researchers. The results show that: the energy released instantaneously causes a sharp increase in temperature near the contact point when lightning strikes the ground. The thermal effect of the soil reaches a peak at around 30 μs after a lightning strike, and the temperature influence radius is less than 40 cm. Besides, the lightning current and soil characteristic parameters have a significant impact on the thermal effect of lightning strikes on soil. The peak value of a lightning current can affect the radius of the area near the lightning strike point that reaches the melting temperature. Different lightning waveforms can determine the speed of current change, thereby affecting the speed of dissemination and range of heat in the soil. Higher initial resistivity of the soil will increase the thermal effects caused by lightning strikes, and a larger specific heat capacity will increase the thermal stability of the soil. The results of the current work are expected to have important theoretical and practical significance for understanding and predicting the thermal response of lightning-struck soil, which can further be applied to guide the design of lightning protection and grounding systems to improve the safety of infrastructures.

Abstract:To solve the problem of low accuracy of existing fault diagnosis models under imbalanced data distribution caused by insufficiency of fault samples in practical chemical process, a fault diagnosis model based on cost sensitive multireceptive fields spatio-temporal graph attention network (CSMRFSTGAT) is proposed. This model converts the corresponding variable data collected from chemical process into topological graph data through maximum information coefficient (C_MI) weighted calculation. Using the fault diagnosis model of the graph convolutional network (GCN), multireceptive fields graph convolutional module (MRFGCM) and space-time graph attention module (STGAM) are designed. Then, a hybrid margin-aware focal loss function is proposed to impose more penalties on samples which are difficult to recognize. The proposed model is applied to evaluate its diagnostic performance in multiple imbalanced scenarios of the Tennessee Eastman process (TEP) and the three-phase flow (TPF) dataset. The results show that the proposed model achieves the classification precision and F1 score of more than 91% and 92% in the TPF dataset, and meanwhile the classification recall rate and F1 score in the TEP dataset both break through 99%, respectively; It can recognize faults more efficiently compared with the machine learning model, deep learning model and graph deep learning model. The proposed model has excellent generalization performance in dealing with the data imbalance problem, and can effectively realize chemical process fault diagnosis under sample imbalance.

Abstract:To address the issues of insufficient parameter flexibility and low code rate in locally repairable codes (LRCs) within current distributed storage systems, this paper introduces two new types of locally repairable codes. First, this paper constructs a class of iterative matrices based on the combination of all-zero matrices and all-one vectors, and then proposes a construction algorithm for all symbol-locally repairable codes (AS-LRCs) with (r,t)-locality using the constructed iterative matrices as parity-check matrices. Subsequently, by improving the structure of the parity-check matrix of AS-LRCs based on iterative matrices, a construction algorithm for information symbol-locally repairable codes (IS-LRCs) with (r,t)-locality is further proposed. Experimental and theoretical analyses show that AS-LRCs meet strict availability requirements, and when the availability parameter t=2, their code length reaches the theoretical minimum bound, making them the optimal LRCs in terms of code length. The minimum distance of IS-LRCs reaches the Singleton-like optimal bound, making them the optimal LRCs in terms of minimum distance. Both AS-LRCs and IS-LRCs construction algorithms support flexible configuration of arbitrary locality and availability. The code rates of the two construction algorithms are significantly higher than existing methods, reaching the theoretical optimal bound of code rate when t=2. The construction algorithms of the two types of LRCs not only ensure efficient data repair but also support more flexible parameter configurations and achieve higher code rates. This provides more efficient coding strategies for distributed storage systems and thereby enhancing the overall performance of distributed storage systems.

Abstract:To address the high communication overhead caused by global gradient parameter updates at aggregation nodes in distributed data parallel training of deep neural network (DNN), a parallel optimization method of multi-step delay parameter updates for deep neural network is proposed. Firstly, an adaptive multi-step update interval selection strategy was designed. After completing multiple local iterative parameter updates, node gradients are aggregated to update the global model parameters, reducing the excessive communication overhead caused by frequent gradient aggregation. At the same time, to prevent the local model from deviating from the global model after several local updates, a parameter correction strategy is proposed to ensure the accuracy of model training. Secondly, during gradient aggregation, the gradient tensor is split into several sub-tensors. By combining sub-tensor priority scheduling, communication and computation during gradient aggregation are maximally overlapped, further accelerating the model training process. Finally, on the CIFAR-100 and ImageNet-mini datasets, the proposed method is compared with SSGD, Local SGD training methods. Results show that the proposed method can significantly reduce communication overhead due to parameter updating on the basis of ensuring model training accuracy. It can maximize the overlap of communication and computing, and make full use of computing resources to improve the speed of parallel training. The results of this study can provide a new resolution to reduce communication costs in the distributed training process of deep neural network.

Abstract:To improve the accuracy of compensating current tracking control for shunt active power filters (SAPF) under the condition of grid frequency fluctuation and suppress the harmonic pollution of power grid effectively, a frequency-adaptive repetitive control strategy is proposed in this paper. A multi-rate repetitive control system is built based on a selective repetitive controller with an adjustable sampling rate. The multi-rate repetitive control system is transformed into a frequency-adaptive uniform-rate selective repetitive control system (FUSRCS) by a finite impulse response (FIR) filter based on Lagrange linear interpolation. To enable the FUSRCS to adapt to grid frequency fluctuations, the coefficients of the approximate expression of the delay link in the repetitive controller are adjusted according to the grid frequency. The compensator is designed and the stability, convergence and steady-state error of FUSRCS are analyzed. The mathematical model of three-phase SAPF is established, and a compensation current compound repetitive control system based on FUSRCS is designed according to the actual data. Simulation and experimental results indicate the SAPF based on FUSRCS can maintain high compensation current tracking accuracy and good compensation effect under steady-state grid frequency deviations, dynamic grid frequency variations, and load switching. Compared with conventional repetitive control, FUSRCS not only offers frequency adaptability, but also reduces the computational burden of the control system, enhances the dynamic response speed, and solves the problems caused by multi-rate repetitive control systems.

Abstract:In order to address the issues of detail loss and poor interpretability in current deep learning-based dehazing algorithms, this paper proposes an interpretable dual-branch deep learning dehazing algorithm. The algorithm employs a dual-branch collaborative architecture to decouple the dehazing task: the upper branch focuses on haze extraction through a designed haze removal block (HRB) that captures haze features in the frequency domain, while incorporating a channel attention mechanism to enhance feature extraction in dense haze regions. The lower branch adopts an aggregated residual framework for detail restoration to correct texture details lost during feature extraction. By computing the negative residual between the hazy image and the haze feature image, a preliminary dehazed image is obtained, which is then refined by the lower branch to produce the final dehazed result. Experiments on the SOTS, NH-HAZE, and real-world datasets demonstrate that compared to existing mainstream dehazing algorithms, the proposed method achieves more thorough haze removal, more complete detail preservation, and significant improvements in objective evaluation metrics. This work not only establishes new research directions for deep learning in image dehazing field but also provides a practical solution for real-world image clarity enhancement.

Abstract:Diaphragm walls, as water interception, seepage control, load bearing, and water retaining structures, have many applications in deep foundations. To improve the waterproofing and mechanical properties of diaphragm walls after joint cracking, four-point bending, and waterproofing tests were carried out using high-density polyethylene corrosion protection liners on four pieces of wall with reinforced joints and two pieces of wall to compare with unreinforced walls with joints. To study the damage pattern, deformation capacity, load-bearing capacity, and waterproofing performance of walls before and after reinforcement with high-density polyethylene corrosion protection liners and reinforcement planting. The experimental results show that the diaphragm wall joints reinforced with corrosion protection liners can significantly improve the bearing capacity and waterproofing performance, postpone the concrete cracking, and increase the ultimate deformation capacity of the diaphragm wall joints. According to the “load-deflection” curve at the joints of the wall, the bearing capacity of the wall is divided into the initial cracking stage, the crack development stage, and the damage stage. When the wall reaches the cracking stage, the joint leakage problem must be considered quickly. As the water pressure increases, the waterproofing performance of the joints before and after corrosion protection liners decreases under the same loading conditions.

Abstract:To enhance the capability of DC-DC converter in power management integrated circuit to withstand single event transient (SET), this study thoroughly investigates the characteristics between SET and load transient in DC-DC converter. Based on this analysis, a radiation hardened by design (RHBD) circuit is proposed. This design outputs control signals to manage the RHBD circuit, by distinguishing SET from load transients, thereby enhancing transient response under dynamic conditions. thereby enhancing transient response under dynamic conditions. The design and validation of a Boost converter are completed based on a 180 nm BCD process. Experimental results demonstrate that with an input voltage range of 2.9 to 4.5 V, an output voltage range of 5.8 to 7.9 V, and a load current ranging from 0 to 55 mA, the detection circuit swiftly disables the hardening module during load transient events. This effectively prevents oscillations. Under the influence of SET, output fluctuations remain within maximum allowable ripple, achieving a SET suppression capability of over 86%. The system operates normally under ionizing radiation with linear energy transfer (LET) up to 100 MeV·cm²/mg. The hardened circuit proposed in this paper can maintain normal operation under varying load conditions while also mitigating SET.

Abstract:To enhance the quality and information integrity of fused images and tackle issues like inadequate feature extraction, insufficient texture details, and loss of global contextual information in infrared and visible images fusion, a fusion and decomposition network architecture for infrared and visible images is proposed. Firstly, a parallel structure of Restormer and Res2Net is utilized. Multiple deep convolutional heads with transposed attention mechanisms and multi-scale residual connections are employed to collaboratively capture the global contextual information and local detail features. Secondly, an invertible neural network with affine coupling structure is adopted to divide the shallow-level features of infrared and visible images into two parts, using alternating coupling transformations to achieve lossless feature preservation. Then, the reconstruction module generates high-quality fused images through concatenation and convolution operations. Finally, the decomposition network reverses the fusion image into source images by minimizing the decomposition loss function. Experimental results show that on the RoadScene dataset, the objective and subjective results of this method surpass most comparative methods. Specifically, compared to other methods, the standard deviation improves by an average of 8.5%, the difference correlation coefficient by 23.1%, the average gradient by 49.0%, and the spatial frequency by 56.1%. On the MSRS dataset, the proposed method outperforms SDCFusion method by 1.4% in standard deviation, 0.4% in visual information fidelity, 0.6% in average gradient, 4.3% in difference correlation coefficient, and 3.4% in spatial frequency. The proposed method shows significant advantages in improving the quality of fused images, preserving texture details, and retaining global information.

Abstract:To suppress motor common-mode voltage while maintaining good steady-state performance,a model predictive flux control method was proposed. Firstly, the H8 inverter′s candidate voltage vectors are expanded by incorporating zero vectors into the candidate set to reduce the current harmonic content. Secondly, according to the flux error vector, the voltage vectors are rapidly selected, which ensures low torque ripple and low current harmonic content. Moreover, Considering the influence of dead-time (DT) on common-mode voltage, the action sequence of the selected voltage vectors is defined to avoid potential equivalent zero vectors. Then, based on the flux beat control principle, a new calculation method of voltage vector duty cycle is designed to reduce flux pulsations. Finally, the proposed control method is compared with the three-vector method, five-vector method and eight-vector method which can suppress the common-mode voltage. The results show that the proposed control method can avoid the dead-time common-mode voltage peak. Comparing with the three-vector method and the eight-vector method, the common-mode voltage amplitude is reduced by 66.67% and 33.33%, respectively. At the same time, the proposed control method can effectively reduce the torque ripple, flux ripple, and current harmonic content of the motor. The proposed control method can improve the steady-state performance of the motor while suppressing the common-mode voltage, and reducing the computational complexity.