2023 volumne 44卷 Issue 05
YU Kunjie, YANG Zhenyu, QIAO Kangjia, LIANG Jing, YUE Caitong
Abstract: To address the difficulties of slow convergence and difficulty in finding feasible solutions when solving large-scale constrained multi-objective optimization problems, an adaptive two-stage large-scale constrained multiobjective evolutionary algorithm was proposed. In the first stage, the algorithm adaptively selected some variables for optimization according to the nature of the decision variables, without considering any constraint to make the population quickly cross the infeasible region and approach the unconstrained Pareto front. In the second stage, the algorithm considered all the constraints and optimizes the variables as a whole using the ε constraint-handling technique. At the same time, the feasible and non-dominated solutions obtained in the evolutionary process were saved and updated using archive to continuously improve the convergence and diversity of the population. Finally, the proposed algorithm was experimentally compared with the other six algorithms on 37 test functions, and the results showed that the proposed algorithm could achieved the best results on 25 functions and outperforms the comparison algorithm on at least 31 functions, respectively; meanwhile, the feasibility rate of the proposed algorithm in more than 90% of the functions could reach 100%, which could effectively solve large-scale constrained multi-objective optimization problems.
OUYANG Cong1, GUAN Jing2, YANG Ming1
Abstract: The precise grouping method might not constantly improve the algorithm performance and sometimes even lead to performance degradation when the cooperative co-evolutionary algorithm was used to solve large-scale global optimization problems with entirely separable or fully non-separable decision variables. A cooperative co-evolutionary algorithm (RG-CCFR3) based on resource allocation and dynamic grouping was proposed to address the above problems. The algorithm was based on CCFR3, where the array with the array index was first set for determining the group size at each round of optimization when the decision variables were fully divisible or fully indivisible. Secondly, the decision variables were randomly grouped according to the group size, which made the assignment of each group of decision variables more diverse in different rounds of optimization. Finally, the processing logic in CCFR3 at each round of optimization was modified to ensure a consistent number of rounds of optimization. The benchmark test functions in CEC2013 and CEC2010 were selected to examine the algorithm′s performance. And RG-CCFR3 was compared with CCFR3, MMO-CC, and CBCC-RDG3 and tested for significance. The experimental results showed that, compared with the CCFR3 algorithm, the RG-CCFR3 algorithm would perform better in most cases when dealing with problems with entirely separable or non-separable decision variables; it was also competitive with the MMO-CC and CBCC-RDG3 algorithms.
SHEN Xiaoning 1,2,3,4 , MAO Mingjian 1 , SHEN Ruyi 1 , SONG Liyan 5
Abstract: This study aimed to solve the scheduling problem of large-scale agile software project. It was decomposed into three strong-coupled subproblems: story selection, story allocation and task allocation. Dynamic events such as the addition and deletion of user stories, the change of employee′s working hours in each sprint, and other constraints such as team development speed, task duration and skills were introduced. To maximize the total value of user stories completed by the project, a large-scale agile software project scheduling mathematical model was established. According to the characteristics of the problem, the Markov decision process was designed. Ten state features were used to describe the agile scheduling environment at the beginning of each sprint; 12 composite scheduling rules were designed as candidate actions of the agent; and rewards were defined according to the objective function of the scheduling model. A priority experience replay double deep Q network algorithm based on composite scheduling rules was proposed to solve the built model. The double Q network strategy and priority experience replay strategy were introduced to avoid the over-estimation problem of deep Q network and improve the utilization efficiency of trajectory information in the experience replay pool. In order to verify the effectiveness of the proposed algorithm, experiments were carried out in six large-scale agile software project scheduling numerical examples, and the convergence of the proposed algorithm was analyzed. According to the performance measurement of the algorithm, it was compared with the existing representative algorithm DQN, double deep Q network and 12 single composite scheduling rules. The results showed that it had the highest average cumulative reward value in 6 different numerical examples.
LI Xi , LI Shuai, FENG Yanhong, LI Mingliang
Abstract: In traditional differential evolution algorithm solves optimization problems generally starting from zero-knowledge and searching independently without using the information of similar problems that have been solved. In this paper, the transfer learning technique was introduced into DE. Firstly, the extracted key information from the optimized population of the source problem and the current population of the target problem was mapped to the high-dimensional Hilbert space by the joint distribution adaptation method. Then a new population was constructed from the mapped matrix to replace the population of the target problem. Lastly, the subsequent evolution was completed. Two transfer modes were implemented: transferring the effective information of the source problem to guide the search direction of the algorithm during the initialization; utilizing the transferred effective information after a certain number of iterations to accelerate the population convergence. The proposed algorithm was tested with nine sets of multi-task test functions, and compared with the DE without transfer and transferred DE without adaptation technique. The results showed that, the proposed algorithm outperforms the traditional DE on seven sets functions in term of solution quality. Moreover, the proposed algorithm converged faster than DE. Therefore, the transfer learning-based differential evolution algorithm was effective in improving the solution accuracy and convergence speed of the objective optimization problem.
DENG Chuanyi, SUN Chaoli, LIU Xiaotong, ZHANG Xiaohong, LI Chunpeng
Abstract: Challenges in expensive large-scale optimization problems, such as high coupling between variables, easy falling into local optimal solution, and computationally expensive objective function, resulted in the difficulty to achieve the global optimal solution. An inertial grouping and overlapping feature selection technique for cooperative coevolutionary ( IG-OFSA) algorithms was proposed to solve expensive large-scale optimization problems. In the proposed algorithm, firstly, a large-scale optimization problem was decomposed into several low-dimensional overlapping sub-problems by using overlapping feature selection technology, and each sub-problem was optimized independently with the assistance of a surrogate model. Then, promising solutions found for each sub-problem would be merged into a context vector for expensive objective evaluation. In addition, an inertial grouping technology was used to control the frequency of regrouping during the optimization to extend the cycle of exploitation of the grouping scheme, and correspondingly improved the performance of optimization. The performance of IG-OFSA was tested on 15 CEC2013 benchmark problems and compared with three state-of-the-art algorithms. The experimental results showed that the performance of IG-OFSA was competitive to solve the expensive large-scale optimization problem, especially, good for solving problems with partially separable, overlapping or completely non-separable decision variables.
ZHANG Zhen1, CHEN Kexin2, CHEN Yunfei1
Abstract: Aiming at the problems that traditional controlled knife detection methods relied too much on specific equipment and environment, and the target detection was applicable to small scope of application, with weak anti-interference ability and low precision, a YOLOv5 model with optimized clustering and CBAM was proposed for controlled knife detection was proposed. Firstly, a decision diagram could be drew and cluster centers could be obtained by calculating the local density of each point and the minimum distance between it and others with higher density. Then the core points could be extracted by calculating all points in the local class the average distance with its class center, the global search assignment strategy was used to classify the test points. Finally, the statisticallearning strategy was employed to allocate the remaining points, these unprocessed points were used as noise points, and they were classified into the class of its nearest neighbor. The improved density peak clustering algorithm was used to analyze the bounding box of the controlled knife, optimize the size of the priori box, and improve the matching degree between the priori box and the size of target object, so as to solve the problem that the clustering effect of the K-means clustering algorithm in the YOLOv5 model such as its unstability and the low convergence rate of large-scale data. Moreover,The C3 module in the Backbone network was combined with the CBAM attention mechanism, named CBAMC3 module, which could improve the model′s ability to extract target features, solve the problem that the YOLOv5 algorithm was not effective for small target detection, and improve the model accuracy. The experimental results showed that the values of P, R, mAP @ 0. 5, mAP @ 0. 5 ∶ 0. 95 of the improved model YOLOv5-Plus on the customed data set were 98. 14%, 95. 80%, 97. 56%, and 76. 68%, respectived, which is 1. 64%, 1. 59%, 1. 51%, and 3. 26% higher than that of YOLOv5 before improvement, and also verified that the proposed model could effectively improve the detection performance of controlled knife in public areas.
HUANG Zijuan, TU Juan, DAI Zunxiang
Abstract: Bioelectric signals belong to weak low-frequency signals with strong noise, therefore it is necessary to filter out power frequency interference. In order to ensure the accuracy and effectiveness of the filtering during power frequency offset, local outlier factor based on frequency density, and combines empirical mode decomposition was proposed to carry out adaptive denoising of signals. Firstly, the local outlier factor was used in the frequency domain by the short-time Fourier transform, and the frequency offset and the offset time and frequency were found by FLOF. Secondly, the signal was segmented according to the offset time, and the average instantaneous power frequency within the segment was used as the actual power frequency within the segment. Finally, each signal segment was decomposed by EMD to generate multiple local feature components of different time scales. More useful information could be reserved only for the component filtering containing power frequency signals. The frequency estimation accuracy of this method was high, and the SNR, RMSE, and SIM were improved after filtering in different dB. Taking -30 dB as an example, compared with the least mean square error filtering and recursive least squares filtering, the SNR increases by 16. 266 and 7. 671 dB, the RMSE decreased by 16. 017 and 4. 388 dB, and the SIM increased by 0. 200 and 0. 013. It proved that the filtering effect in this study was better than the conventional adaptive filter.
CUI Jianming1, LIN Fanrong1, ZHANG Di1 , ZHANG Luning1, LIU Ming2
Abstract: As an important part of autonomous driving, trajectory prediction aimed to forcast the vehicle′s driving path, so that the vehicle could make path planning according to the driving estimation, so as to make safe and accurate decisions. Firstly, in order to improve the accuracy of vehicle trajectory prediction, the directed graph method was used to construct a high-definition driving scene map, and the directed graph method vectorized the map information to effectively extract the map topology. Secondly, GAIL was used to learn the driving strategy of the dataset through the confrontation game between the generator and the discriminator, so as to adopt the corresponding driving behavior according to the current state. Finally, the multimodal prediction trajectory scheme was obtained by sampling traversal. Simulation was carried out on the nuScenes motion prediction dataset. The quantitative results showed that compared with other methods, when K = 5, the minimum final displacement error MinFDE5 was increased by 10. 8%; when K = 10, the minimum fianl displacement error MinFDE10 increased by 17. 53%, the minimum average displacement error MinADE10 increased by 9. 52%, and the error rate MissRate10 decreased by 28. 26%. The evaluation showed that the generated trajectories were multimodal, could conform to the basic structure of the scene, with improved accuracy.
WANG Jie1,2, GE Lina1,2, ZHANG Guifen1
Abstract: To solve the problem of unreasonable distribution of PoS block rewards, a proof of stake based on incentive (Incentive-PoS)consensus algorithm was proposed. Firstly, the research problem was described. A PoS determined that nodes with more coins have a greater chance of obtaining accounting rights, and the block reward was exclusively owned by the block producer. Secondly, in order to solve the problem of reward distribution, a PoS consensus algorithm based on incentive mechanism was proposed, Shapley′s principle in game theory was uesd to redistribute block rewards. Nodes with high credibility and active participation in consensus would receive dividends, and made small nodes more likely to obtain benefits. Finally, the simulation experiment and result analysis of the improved algorithm were carried out. Compared with the original algorithm, the improved scheme had a more reasonable performance in the distribution of income, and increased the number of nodes receiving dividends, reduced the gap between the rich and the poor, and improved the enthusiasm of consensus. And the throughput, latency, and security were significantly improved. It was beneficial to improve the stratification phenomenon caused by the excessive wealth gap in the blockchain, and could further promote the healthy operation and development of the blockchain network.
LI Jingli , REN Junyue, YUAN Hao, WANG Zijian, LEI Hong, ZHAO Zijing
Abstract: Chinese urban low-voltage distribution network capacitance current rose sharply, single-phase ground fault arc was difficult to self-extinguish and easy to produce arc light ground overvoltage and lead to accidents. The resonant grounding system (NES) arc ground fault reliable line selection method to ensure the stable and safe operation of the distribution network basis was explored. The equivalent circuit of arc ground fault in NES was first established, and the difference between the zero-sequence current between the fault and the sound line was analyzed. Secondly, the Schwarz arc model was improved by introducing dynamic arc length parameters, and the Schwarzarc simulation model was improved by using MATLAB / Simulink, and on this basis, the arc-optic grounding fault model of the pure cable line system in the scenario of exceeding the capacitance current of a medium-voltage distribution network in a city was constructed. Then, the typical fault conditions were set to simulate the arc grounding fault of any line, and the zero sequence current of each line was extracted for coif wavelet decomposition, and it was proposed that when the arc grounding fault occurred in the outlet line, the polarity of the fault line and the nonpower frequency zero sequence current mode maximum of the sound line was opposite. When the arc ground fault occurred on the busbar, the polarity of the non-power frequency zero sequence current mode maximum of all out lines was the same as the arc ground fault selection criterion. Finally, the typical fault conditions were set to verify the effectiveness of the method for selecting wires for NES arc ground faults. The results showed that in different typical fault conditions such as transition resistance, fault point distance, and phase angle at fault time, this method could correctly select the arc light ground fault line.
CHEN Jinghua1, ZHANG Linjuan1, LU Dan1, GUO Pu1, REN Junyue2 , LI Jingli2, LI Zhongwen2
Abstract: With the development of new power system construction, the proportion of grid-connected distributed power gradually increased. In order to solve the difficult problem of the regulation of distributed power in new distribution network, the improved particle swarm optimization algorithm was proposed to cluster the large-scale distributed power. Firstly, on the basis of the modularity division standard, the active power matching degree and reactive power matching degree of the group internal load were introduced, and the comprehensive performance indexes weighted by the three were proposed to construct the cluster division optimization model based on the comprehensive index system. Then the inertia weight dynamic decline strategy was used to improve the binary particle swarm optimization algorithm to make the inertia weight change dynamically, optimizing the updating process of particle position and velocity, and improving the optimization efficiency of particle swarm optimization algorithm. Finally, the improved binary particle swarm optimization algorithm was used to optimize the clustering optimization model based on the comprehensive index system. Based on this, cluster division of IEEE33 nodes and a 10 kV actual distribution network feeder system was carried out. The results showed that the cluster partitioning method based on the comprehensive index system could improve the active power matching degree and reactive power matching degree by about 30%, respectively, on the basis of keeping the modularity of the partitioning result basically unchanged. The improved particle swarm optimization algorithm had obvious effect on improving each index value of partitioning results.
GAO Chao1 , LIU Zehui1, CAO Dong2, YAO Lina2
Abstract: In order to improve the accuracy of fault diagnosis of power cable failure to ensure a cable fault detection method based on convolution neural network (CNN) and bi-directional long short term memory (BiLSTM) was proposed in this paper. The simulation model was built through Simulink to extract the voltage signals of single-phase ground short circuit, two-phase ground short circuit, two-phase phase short circuit and three-phase short circuit faults, and to generate the fault sample. Then the fault voltage signals were input into the network model, to obtain the local features through CNN, to obtain the fault signal timing information through BiLSTM, and to realize the diagnosis of cable fault based on the automatically extracted features. The simulation results showed that this method could accurately classify the four short-circuit faults of power cables, the accuracy rate of single-phase grounding short-circuit fault and three-phase short-circuit fault was 97%, the accuracy rate of two-phase ground short circuit and two-phase short circuit was 92%, and the overall accuracy rate was 98. 37%. In addition, through the analysis of loss function curve and accuracy curve, it was proved that this method could achieve better cable fault diagnosis effectiveness. Finally, the actual data was used to verify the feasibility of the method.
HUANG Yuan1,2 , WANG Li2
Abstract: To explore the effect of torsion on the progressive collapse resistance of RC frame, a finite element model was established by LS-DYNA. Based on the reasonable verification of the torsion tests of beams and the antiprogressive collapse tests of frames, the influence mechanism of torsion on the progressive collapse resistance of the 2-D frame was studied, and the anti-collapse performance of the spatial frame was analyzed. The results showed that compared with non-torque members, when the initial torque reached 20%, 40%, 60%, and 80% of the torsional resistance of the beam, the compressive arch action capacity decreased by 0. 5%, 3. 4%, 10. 9%, and 15. 6%, respectively. With the increasing torque, the fracture of the beam bottom reinforcements showed a trend of advancing and then delaying. While the fracture of the beam top reinforcements near the middle column kept advancing. The coordinated torsion caused by transverse beam bending in the spatial frame would make the bottom reinforcements fracture earlier than the planar frame. The collapse resistance of the spatial frame was equal to the superposition of the capacities of its corresponding planar frame and transverse beam before the reinforcements fracture.
YANG Yaxun1,2, WANG Chengzhi1 , CHAI Wenhao1, ZHANG Yuhang1, ZHANG Fuhua1
Abstract: In order to study the mechanical response of long-span curved cable-stayed bridge caused by cable breaking, Xigu Chaijiaxia cable-stayed bridge in Gansu Province was taken as an engineering example, the finite element model was established by using the beam lattice method, and 15 typical cable breaking conditions were selected to analyze the changes of residual cable force, main beam deflection, main beam stress and tower top offset before and after cable breaking at different positions and different numbers. The results showed that single cable breaking would only produce local effects, the peak value of the change occurred near the broken cable anchorage zone and decreased to the surrounding. The combined fracture condition had a greater impact than the single cable fracture condition. Long cable fracture had the greatest impact, followed by medium cable and short cable. The impact of cable fracture in the middle span was greater than that in the side span. The influence of cable fracture on the broken span was greater than that of non broken span, and the influence on the broken cable surface was greater than that of non broken cable surface. The effect of cable breaking on the peak stress of the main beam and the stress distribution of the whole bridge was very small, but only on the main beam stress near the anchorage zone.
ZHANG Hua, PENG Zhaohui, ZHANG Qiang, MA Mengmeng, WANG Yawei, LI Zongkun
Abstract: Traditional Bayesian network (BN)method was difficult to ensure the dynamic assessment of construction progress risk, on the basis of analyzing the risk sources of construction progress lag, the time factor was introduced to horizontally expand the Bayesian network into a dynamic Bayesian network (DBN) with multiple time periods, and the construction progress risk was dynamically assessed in combination with the construction progress monitoring data. El Sillar highway project was taken as an example to verify the rationality of the method. The results showed that the construction progress risk of the El Sillar Highway Project was controllable as a whole, and the probability of completion in each period was about 60%. The risk of delay showed a trend of increasing first and then decreasing. The main factor causing the delay of the construction period was the environmental risk, the DBN model could consider the dynamics of risk factors, so it was superior to the static model. The dynamic risk analysis of highway construction progress based on DBN could provide reference for similar projects.
GUO Yinchuan, LIU Yiwei, SHEN Aiqin, LI Zhennan, WU Jinhua, ZHANG Jialong
Abstract: In order to reduce the generation and expansion of reflective cracks in asphalt pavements, to improve the overall crack resistance of asphalt pavements and prolong their service life, glass fibres, which were economically efficient, convenient to obtain and have good toughening effect, were used to improve the shrinkage and crack resistance of cement stabilised macadam, and the effect of glass fibres on the softening and crack resistance of cement stabilised macadam was analysed through drying shrinkage tests, temperature shrinkage tests, bending toughness tests and fracture energy tests. The results of the study showed that three different types of glass fibres were used to improve the shrinkage and cracking resistance of cement stabilised macadam. The results showed that the drying shrinkage coefficients of cement stabilized macadam with three different glass fibre doping levels decreased by 6%, 13% and 16% respectively at 30 d compared with those of ordinary cement stabilized macadam; the average temperature shrinkage coefficients decreased by 6%, 16% and 19%, respectively; the average temperature shrinkage coefficient decreased the most when the glass fibre doping level was increased from 0. 05% to 0. 10%, reaching 10%, and the best improvement effect was achieved at this time. The bending and tensile toughness of the three different glass fibre admixtures increased by 25. 9%, 48. 1% and 150. 0%, respectively; the ultimate breaking load increased by more than 20% and the deflection at breakage increased by 53% after the addition of 0. 10% glass fibre modification compared to ordinary cement stabilised macadam; the fracture energy gain ratio of cement stabilised aggregates reached 1. 225. The addition of glass fibre could effectively inhibit the drying shrinkage and temperature shrinkage deformation of cement stabilised macadam, while the modified cement stabilised macadam were much tougher and consumed much more energy at breakage than ordinary cement stabilised macadam, which was important for enhancing the shrinkage and crack resistance of cement stabilised macadam.
XU Ouming, XU Rentao, LI Yang, HAN Sen
Abstract: In view of shear failure of thin asphalt overlays on old cement concrete pavement, three types of tack coat oil ( ordinary emulsified asphalt, SBS modified emulsified asphalt and self-made high-performance emulsified asphalt) , three types of overlay mixtures (AC-10, SMA-10 and DTFC-10) and three kinds of underlying surface roughness ( initial texturing, chiseling, and grooving) were selected to prepared compound pavement. The effects of various factors, such as the type and dosage of the tack coat oil, the type of the overlay mixtures and the surface roughness of the existing cement pavement, on the interlayer shear strength were evaluated by the direct shear test. In addition, the influence degrees of various factors were analyzed though gray theory. The results showed that the type of tack coat oil had a significant effect on the interlayer shear strength. Furthermore, the interlayer shear strength first increased and then decreased with the increase of the tack coat oil consumption. And the shear strength reached the maximum value when the amount of the tack coat oil was 0. 8 L / m 2 . The effects of the overlay asphalt mixtures type on the interlayer shear strength were significant, and the values of ultra-thin wear course DTFC-10 were greater than other two mixtures. The surface textural behavior of concrete pavement also had positive contribution to the interlayer shear strength. The shear strength values of underlying surface treated by chiseling method were slightly higher than that treated by grooving method. While, the values of those two textured surface were obviously greater than the initial surface. Gray correlation analysis revealed that the interlayer shear strength were notable dependent on the thin HMA overlays and the tack coat oil. Moreover, the surface texture of cement concrete pavement and the consumption of tack coat oil also had a certain influence.
Copyright © 2023 Editorial Board of Journal of Zhengzhou University (Engineering Science)