2024 volumne 45 Issue  02
WANG Dingbiao 1,2 , DUAN Hongxin 1,2 , WANG Guanghui 1,2 , SHEN Aoqi 1,2 , LIU Heyu 1,2 , QIN Xiang 1,2
Abstract: As an important part of the transcritical CO2 cycle system, the control strategy played the key role to ensure the high efficiency and energy saving operation of the system. Studies of control strategies were examined such as the feedback control based on the empirical calculation of the optimal diacharge pressure of the system and the Buckingham π theorem, the real-time online control based on gradient tracking and extreme seeking, and the predictive control based on the neural network, etc. The development history and future development trend of the system control strategy were analysed and summarized in detail. Off-line control was easy to establish with low cost, but it was easily affected by environmental factors and changes in system components, resulting in reducing control performance; The real-time online control strategy could track the discharge pressure corresponding to the maximum energy efficiency of the system in real time, but due to the long optimization process, the convergence time of the control system was too long. Model predictive control system could realize real-time optimization and rapid convergence, and had a good development prospect. Combined with the practical scenarios of new energy vehicles, building heating, rail transit, commercial refrigeration, military industry and other practical scenarios, the application characteristics and future development trend of the control strategy of the transcritical CO2 cycle system were explored, and it was further explained that improving the applicability of the control strategy was an important direction for future research. The feasibility of applying adaptive methods such as generalized predictive control and reinforcement learning to the control strategy of transcritical CO2 cycle system was proposed, and the significance of developing control strategy for large-scale cycle system and energy storage system in China with the background of “ double-carbon” was discussed.
ZHANG Zhen 1 , WANG Xiaojie 1 , JIN Zhihua 1 , MA Jijun
Abstract: In order to improve the detection speed of road traffic signs, an improved model based on lightweight YOLOv5 was proposed. Firstly, Ghost convolution and depthwise convolution were used to build a new Bottleneck, which could reduce the amount of computation and parameters. Then the BiFPN structure was introduced, which could enhance the feature fusion ability. CIoU loss function was replaced by SIoU loss function, which focused on the angle information of ground true box and prediction one, so that it would improve the detection accuracy. Secondly, the TT100K dataset was optimized, and 24 categories of traffic sign pictures and labels with more than 200 were screened out. Finally, the experiment achieved 84% accuracy, 81. 2% recall and 85. 4% mAP@ 0. 5. Compared with the original YOLOv5 model, the number of parameters was reduced by 29. 0%, the amount of computation was reduced by 29. 4%, but the mAP@ 0. 5 was only reduced by 0. 1 percentages, and the detection frame rate was improved by 34 frames/ s. Using the improved model for detection, the detection speed could be significantly improved, could basically achieve the goal of compression model on the basis of maintaining the detection accuracy.
FAN Wenbing, ZHANG Lulu
Abstract: Aiming at the problem that KCF tracking algorithm might decrease the tracking performance or even tracks unsuccessfully in the occlusion scene, an anti-occlusion model adaptation image tracking algorithm was proposed by combining KCF and KF prediction. Firstly, considering the lack of occlusion evaluation in the traditional KCF target tracking algorithm, the peak sidelobe rate of the response map was introduced to judge the occlusion of the image target, and the occlusion types were divided into partial occlusion and severe occlusion. Then different model update strategies were adopted according to the severity of occlusion. When the target was not occluded or occluded partially, instead of using a fixed learning rate to update the model in the traditional KCF tracking algorithm, the target appearance model was updated by adjusting the model learning rate adaptively to avoid tracking drift. When the target was severely occluded, stopped updating the KCF model. Finally, the state space and position output models of Kalman filter were constructed by applying the motion information before severe occlusion. The Kalman filter prediction algorithm was designed to predict the moving target trajectory and estimate the target position in the occlusion scene,so as to solve the problem of target tracking failure in occlusion scenes. The OTB-2013 standard dataset was utilized to conduct extensive experiments, the results demonstrated that the distance accuracy of the proposed hybrid tracking algorithm KCF-KF was 0.796, and the overlap success rate was 0.692. Compared with the other traditional tracking algorithms, the tracking accuracy and success rate of the hybrid algorithm were better, and the hybrid algorithm could achieve better tracking performance when encountering the target occlusion challenges and solve the occlusion interference in the tracking process effectively.
GAO Yufei, MA Zixing, XU Jing, ZHAO Guohua, SHI Lei
Abstract: For medical image segmentation tasks such as glioma image segmentation with dense prediction, both local and global dependencies were indispensable. To address the problems that convolutional neural networks lacked the ability to establish global dependencies and the self-attention mechanism had insufficient ability to capture local details, a convolutional and deformable attention-based method for glioma image segmentation was proposed. A serial combination module of convolution and deformable attention Transformer was designed, in which convolution was used to extract local features and the immediately following deformable attention. Transformer was used to capture global dependencies to the establishment of local and global dependencies at different resolutions. As a hybrid CNN-Transformer architecture, the method could achieve accurate brain glioma image segmentation without any pretraining. Experiments showed that the average dice score and the average 95% Hausdorff distance on the BraTS2020 glioma image segmentation dataset were 83. 56% and 11. 30 mm, respectively, achieving comparable segmentation accuracy compared with other methods, while reducing the computational overhead by at least 50% and effectively improving the efficiency of glioma image segmentation.
LI Wenju1, JI Qianqian1, SHA Liye2, CHU Wanghui1, CUI Liu1
Abstract: Aiming at the shortage of distance feature and local geometric structure information in feature extraction, a point cloud classification and segmentation network based on graph walk and graph attention was proposed. Firstly, a guided graph walk algorithm was used to supplement additional geometric information and remote feature information to the whole feature of point cloud. Secondly, the graph attention mechanism was embedded to make the model on the key areas of the point cloud and improve the feature extraction ability of the network. Finally, distance features were extracted from the initial point cloud and embedded into the network as initial residuals to avoid oversmoothing. Point cloud classification experiments were carried out on ModelNet40 dataset and ScanObjectNN dataset, and point cloud component segmentation and point cloud semantic segmentation experiments were carried out on ShapeNetPart dataset and Toronto-3D dataset, respectively. The experiment results showed that, compared with the benchmark network DGCNN, classification accuracy increased by 1. 3 percentages and 5. 6 percentages, respectively; The segmentation accuracy was improved by 1. 2 percentages and 33. 1 percentages respectively. Through the robust analysis on ModelNet40-C dataset, it was proved that the proposed network had strong robustness.
CHEN Yan1,2, LAI Yubin1, XIAO Ao1, LIAO Yuxiang1, CHEN Ningjiang1
Abstract: In response to the issues of limited annotated data, insufficient fusion between modalities, and information redundancy in multimodal sentiment analysis, a multimodal sentiment analysis model called CLIP-CA-MSA based on contrastive language-image pretraining(CLIP) and cross-attention mechanism was proposed in this study. This model employed models such as BERT which was pre-trained by CLIP, and PIFT to extract feature vectors from videos and textual content. Subsequently, a cross-attention mechanism was applied to facilitate interaction between image feature vectors and text feature vectors, enhancing information exchange across different modalities. Finally, the uncertainty loss was utilized to compute the fused features, and the ultimate sentiment classification results were generated from the outputs. The experimental results showed that the model could increase accuracyrate by 5 percentage points to 14 percentage points and the F1 value by 3 percentage point to 12 percentage point over other multimodal models, which verifieed the superiority of the model in this study. And uses of ablation experiments to verified the validity of each module of the model. This model could effectively utilize the complementarity and correlation of multimodal data, and utilize uncertainty loss to improve the robustness and generalization ability of the model.
JI Ke1,2, ZHANG Xiu1,2, MA Kun1,2, SUN Runyuan1,2, CHEN Zhenxiang1,2, WU Jun3
Abstract: With the rapid popularization of the Internet, the amount of Internet news has increased dramatically. In this case, how to effectively find relevant reports that are more in line with a specific topic has become an urgent problem to be solved. To address this issue, a topic matching algorithm based on the fusion of key entities and text abstracts was proposed in this study. Firstly, the W2NER model was used for named entity recognition to extract key entities using features such as word frequency, TF-IDF, lexical cohesion word-word similarity, and word-sentence similarity. Secondly, the Pegasus model was used for text summarization, and the deep semantic features of news texts were obtained by combining the key entity features with the text summary features using BiLSTM. Next, the cross-attention mechanism was employed to enhance the interaction between the matching news articles by performing feature interaction. Finally, the deep semantic features of the news texts and the text interaction features were fused together to participate in the determination of text topic matching. Comparative experiments were conducted on real data from Sohu, and the results showed that the proposed algorithm achieved similar accuracy and precision compared to other algorithms, while recall and F1 score were improved.
WANG Hairong, XU Xi, WANG Tong, JING Boxiang
Abstract: In order to solve the problems in studies of multimodal named entity recognition, such as the lack of text feature semantics, the lack of visual feature semantics, and the difficulty of graphic feature fusion, a series of multimodal named entity recognition methods were proposed. Firstly, the overall framework of multi modal named entity recognition methods and common technologies in each part were examined, and classified into BilSTM-based MNER method and Transformer based MNER method. Furthermore, according to the model structure, it was further divided into four model structures, including pre-fusion model, post-fusion model, Transformer single-task model and Transformer multi-task model. Then, experiments were carried out on two data sets of Twitter-2015 and Twitter2017 for these two types of methods respectively. The experimental results showed that multi-feature cooperative representation could enhance the semantics of each modal feature. In addition, multi-task learning could promote modal feature fusion or result fusion, so as to improve the accuracy of MNER. Finally, in the future research of MNER, it was suggested to focus on enhancing modal semantics through multi-feature cooperative representation, and promoting model feature fusion or result fusion by multi-task learning.
HU Hongchao 1 , ZHANG Shuaipu 2 , CHENG Guozhen 3 , HE Weizhen 3
Abstract: In addressing the inefficiencies and limitations in proactive defense against Regular Expression Denial of Service (ReDoS) attacks in cloud-native environments, we have developed a defense method based on Moving Target Defense (MTD) technology. Initially, we analyzed the behaviors of both attackers and defenders within microservice applications characteristic of cloud-native environments. Subsequently, leveraging Kubernetes, we designed an MTD-based defense system. This system incorporates dynamic and static multi-dimensional microservice weight indices based on topology information and request arrival rates, as well as service efficiency judgment indices based on queue theory. It also includes a method for selecting the timing of key microservice rotations to guide the selection and rotation timings of critical microservices. Finally, we introduced a multi-dimensional MTD heterogeneous rotation algorithm, grounded in heterogeneity and service efficiency, and conducted simulations using Python. Experimental results indicate that our proposed algorithm reduces defense latency by approximately 50% compared to dynamic scaling and that defense costs stabilize after the initial defense against an attack, preventing continuous growth.
YU Kunjie, WANG Siyu, YANG Duo, FU Hanwen, LIAO Yuefeng
Abstract: In order to reduce the equivalent hydrogen consumption of the hybrid system and delay the aging of the fuel cell, an Energy management strategy (EMS) was proposed based on multi-objective optimization and road condition classification. Firstly, the electrical model of the fuel cell and lithium battery hybrid system power was constructed, and the equivalent hydrogen consumption model and fuel cell aging model were introduced. Then, a rulebased multi-mode EMS was designed; on this basis, in order to further reduce the equivalent hydrogen consumption of the system and prolong its service life, the multi-objective beluga whale optimization algorithm (MOBWO) was proposed to optimize the control parameters. Furthermore, in order to make the designed EMS suitable for different road conditions, a real-time classification method of driving road conditions based on long short-term memory ( LSTM) network was proposed, aiming to switch the control parameters of EMS according to the classification results to achieve the optimal effect. Finally, the proposed algorithm was analyzed on the simulation platform. The results showed that the hydrogen consumption of the hybrid system with the proposed method was reduced by 2. 3% and the aging degree of the fuel cell was reduced by 1. 02% compared with the rule-based method,The proposed EMS could effectively reduce the equivalent hydrogen consumption of the hybrid system and delay the aging of the fuel cell.
ZHANG Wenyu, MA Keke, GUO Zhenhai, ZHAO Jing, QIU Wenzhi
Abstract: In order to improve the multi-step prediction of wind speed, a hybrid prediction model based on data signal decomposition and grey wolf optimization algorithm was proposed to optimize extreme learning machine. Firstly, the original wind speed time series was decomposed into several intrinsic mode functions and a residual sequence using the complete ensemble empirical mode decomposition with adaptive noise, and the partial autocorrelation function model input. Then, the model was built and the prediction was made on the decomposition subsequence. An extreme learning machine neural network with multi-input-multi-output strategy was constructed, and grey wolf algorithm was used to solve the weight and bias of the optimal hidden layer. Finally, the subsequence was reconstructed and the final prediction result was obtained. Simulation experiments were conducted using multiple sets of measured data with a time resolution of 15 minutes. The root mean square errors of the proposed model in the three wind farms were 0.859, 0.925, and 0.927, respectively, which were lower than other comparative models, verifying the effectiveness of the model in predicting wind speed in the next four hours,i.e. 16 steps prediction.
LUO Peng, CHEN Guanghao, YANG Donghong, GUO Lei
Abstract: To solve the problem that the output voltage of the new energy power generation device varied greatly and it was difficult to realize energy storage, a novel single-switch coupled buck-boost converter based on PI controller and feed-forward control was presented. The voltage gain could be adjusted by the turns ratio of the coupled inductor, and the voltage stress on the power switch was suppressed by the passive clamped circuit with recycled leakage inductor energy. Compared with traditional buck-boost converter, the proposed converter had the advantages of wider voltage conversion ratio, continuous input current, and low voltage stress on power switch. Combing PI controller with feedforward control strategy, superior input transient response of the converter during the whole input voltage range is obtained. The operating principles and steady-state characteristics of proposed converter were analyzed and derived in detail, respectively, and the performances were compared with other single-tube buck-boost converters. The small-signal model was derived, and the correctness of PI parameter design was verified by bode diagram. The design process of PI controller combined with feedforward control strategy was analyzed. Finally, an experimental prototype with a rated power of 100 W, 20 V to 60 V input, and 48 V output was built to verify the performance of the proposed converter in boost mode and buck mode, and the feasibility of PI controller combined with feed-forward control strategy. The measured maximum efficiencies with the boost and buck modes were 97. 08% and 97. 10%, respectively.
ZHENG Yuanxun, KONG Meng, WANG Boli, WANG Changzhu, CHEN Jing
Abstract: In order to study the effect of bridge pre-camber on the comfort of continuous rigid bridge, a number of continuous rigid bridges with main span diameters between 110 m and 200 m were used as research objects. Firstly, the time domain model of the pavement was established using the filtered white noise method, and the levelness of the new bridge deck was simulated by superimposing it with the formed bridge alignment set according to the cosine curve. Secondly, MATLAB / Simlink was used to build the vehicle-road system model. The root means square (RMS) acceleration value was used to evaluate the traffic comfort of the bridge side span and middle span deck. The maximum transient vibration value (MTVV) was used as the evaluation index of the driving comfort in the short time at the peak of the side span. Finally, the influence of the span diameter and design speed on the RMS value of the side and middle spans of the bridge with the empirical method of setting the pre-arch according to the cosine curve was analyzed. The influence of the pre-arch value on the MTVV value at 3L / 8 of the side span were analyzed. The results showed that the RMS values were less than 0. 315 m / s 2 at the side span and mid-span, indicating that the empirical method of setting the pre-arch of the bridge according to the cosine curve did not affect the traffic comfort at the mid-span. The MTVV value at the side span 3L / 8 was greater than 0. 345 m / s 2 in a short period of time, which mean that the peak of the side span was slightly uncomfortable in a short period of time. In order to improve the traffic comfort within the side span, measures were proposed to optimize the bridge deck alignment at 3L / 8 and the top of the side pier based on the leveling layer construction.
LIU Mingjian, ZHU Yunhe, ZHANG Sijia, SUN Hua
Abstract: The existing autonomous intersection control strategies lack foresight and are prone to deadlock, resulting in low execution efficiency of the control system. To address this issue, a vehicle road collaborative autonomous intersection control strategy based on maximum clique theory is designed. Firstly, the spatiotemporal trajectory of vehicles was modeled, and the conflict matrix describing the driving conflict relationship between vehicles was constructed. Secondly, the conflict matrix was transformed into a conflict relation graph. Through three established solution stages, the complement of the maximum clique in the conflict relation graph was solved as the set of accepted vehicle reservation requests, which could ensure more successful vehicle reservation requests passing through the intersection within each batch processing cycle and could improve the efficiency of intersection passage while ensuring the safety of vehicle driving at the intersection. Simulation results showed that compared with the first come first served control strategy, traffic signal control strategy, and Tabu-based control strategy, the average waiting time was reduced by 40%, 17%, and 8%, and the number of vehicles passing through the intersection per unit time was increased by 30%, 18%, and 9%, respectively. This proved the effectiveness of the strategy, which not only could improved the throughput of the intersection but also could effectively reduce the average waiting time of vehicles.
JIN Libing, WANG Zhenhao, WU Tian, XIE Zhiheng, ZHOU Pin
Abstract: In view of the damage to the durability of concrete structures caused by the coupled effects of loading and sulfate attack in coastal and saline soil environments, a numerical model for the sulfate ion diffusion in compressed concrete was proposed. Firstly, based on Fick′s second law, a theoretical diffusion model of sulfate ions in concrete under loading was established by considering the relationship between stress and concrete porosity. Secondly, a three-phase mesoscopic convex polygonal stochastic aggregate model of concrete containing cement mortar, interfacial transition zone, and natural aggregate was established by a self-programmed program, which enabled the mesoscopic simulation of sulfate ions diffusion in compressed concrete. Finally, the validity of the theoretical and mesoscopic models was verified by comparative analysis with the experiment results of full immersion of compressed concrete in a sulfate solution. Then the ion diffusion and damage process of compressed concrete specimens with different water-cement ratios in sulfate solutions of different concentrations were numerically analyzed. The findings indicate that the sulfate ion concentration at the same depth gradually decreased as the compressive stress level increased. The effects of sulfate concentration and the water-cement ratio on sulfate ion diffusion were more evident than with compressive stress. The extent to which compressive stress inhibited ion diffusion was influenced more by the water-cement ratio than by the sulfate concentration. Reducing the water-to-cement ratio appropriately made the compressed concrete more resistant to the sulfate attack.
Copyright © 2023 Editorial Board of Journal of Zhengzhou University (Engineering Science)