2021 volumne 42卷 Issue 01
HUA Yicun1, LIU Qiqi2, HAO Kuangrong1, JIN Yaochu1,2
Abstract: In reality, the Pareto fronts of multi-objective optimization problems are often irregular. Evolutionary algorithms for such problems have gradually become a hot topic. This paper provides a survey of the existing evolutionary algorithms for the multi-objective optimization problems with irregular Pareto fronts, gives a general mathematical description of the multi-objective optimization problems, and introduces the relevant definitions in the research field such as dominated and non-dominated solutions. It suggests a taxonomy of the typical multi-objective optimization test problems with irregular Pareto fronts, as well as the actual multi-objective optimization test problems with irregular Pareto fronts such as car crash test problem. The existing evolutionary algorithms for multi-objective optimization problems with irregular Pareto fronts are divided into four categories: the methods of adjusting the reference vectors according to the population distribution, the fixed reference vectors merging other auxiliary methods, the methods of reference points, and the methods of clustering and partitioning. Their strengths and weaknesses are discussed. Although evolutionary algorithms for multi-objective optimization problems with irregular Pareto fronts have achieved certain success, existing algorithms generally perform well only on some irregular Pareto front problems. Algorithms that can work efficiently on all kinds of irregular Pareto front problems are yet to be developed. High dimensional, dynamic and the data-driven multi-objective problems with irregular Pareto fronts remain to be solved. More intelligent evolutionary algorithms that can identify and handle multiple types of multi-objective optimization problems with irregular Pareto fronts are the focus of future research. Hybrid approaches, transfer learning or multi-task learning and optimization combined with evolutionary computation are possible solutions.
WANG Shenwen1,2, ZHANG Jiaxing1,2, CHU Xiaokai1,2, LIU Hong3, WANG Hui4
Abstract: In multimodal multi-objective optimization problem, the same position of Pareto front often corresponded to multiple Pareto optimal solutions in decision space. However, the existing multi-objective optimization algorithms could only obtain one of the Pareto optimal solutions. Therefore, in this paper, a two-stage search multimodal multi-objective differential evolution algorithm was proposed, which divided the optimization process into two stages: elite search and partition search. In the elite search stage, elite mutation strategy was used to generate high-quality individuals to ensure the search accuracy and efficiency of the population. In the stage of partition search, the decision space was divided into several subspaces, and the detected population was used to explore each subspace in depth, so as to reduce the complexity of the problem and to improve the expansion and uniformity of the population in the decision space. The performance of the algorithm was compared with five classical algorithms NSGAII、MO_Ring_PSO_SCD、DN-NSGAII、Omni-Optimizer、MMODE on 18 multimodal and multi-objective optimization test functions, such as MMF1. Experimental results showed that there were 16 test functions in the performance index of Pareto approximation (PSP) of the proposed algorithm, which were better than the other five comparison algorithms.
WANG Kewen, YE Mengyu, LIU Yanhong
Abstract: In order to reduce the calculation time of small-disturbance stability analysis of large-scale power systems, the method of rapid formation of power system state space matrix was optimized. According to the plug-in modeling technology, the formation process of the state matrix was analyzed; the incomplete LU decomposition method (ILUTP) with double thresholds was used to adjust the position of the non-zero elements in the correlation matrix, and the matrix was converted to a diagonally dominant form; double conjugate gradient stabilization method (BICGSTAB) was used iteratively to solve the processed large sparse matrix; the storage method of the matrix was row-compressed sparse storage; the algorithm characteristics of ILUTP and BICGSTAB were used to realize the parallel computing based on Open MP technology; two calculation examples of 23 generators and 98 generators were included respectively. The generators adopted the sixth-order generator model. The excitation adjustment module and the prime mover speed control block were the actual parameters of the system, and the traditional methods and optimization methods were compared. The time taken to solve the state matrix method verified the feasibility and effectiveness of the proposed method. The results showed that this method could speed up the formation of the state space of large power systems, and its parallel speedup ratio was close to 3.
LIAO Xiaohui, CHEN Chuanchuan
Abstract: To solve the problems of low power quality disturbance detection accuracy and weak anti-interference, a VMD-HT method based on energy convergence factor and PSO optimization was proposed in this paper. Firstly, the improved VMD method was used to adaptively select the optimal values of k and α, and then the Hilbert transform was used to obtain the instantaneous amplitude-frequency simulation diagram of the IMF component after VMD decomposition, and to locate the start and end moments of the disturbance to realize the power quality disturbance detection. By adding white noise to simulate the actual power quality disturbance and randomly changing the start and end time of the power quality disturbance mathematical model, a large number of detection simulation experiments were conducted. The detection accuracy of the start and end moments of the transient oscillation disturbance signal in an environment containing 20 dB white noise, compares with the HHT, increased by 0.007 s and 0.006 s respectively, and the probability of its detection error within 0.003 s reaches 99.84%. The detection simulation results of composite power quality disturbances also showed that the improved VMD-HT method had better detection effect than the traditional HHT power quality disturbance detection method, and could avoid the problems of low detection accuracy, weak anti-interference and poor fault tolerance due to the phenomenon that EMD was prone to modal aliasing.
ZHANG Zhen, LI Haofang, LI Mengzhou, MA Junqiang
Abstract: In the current community video surveillance system, only a face camera was used to collect the entrance and exit face data, other valuable human information was negelected. In this paper, a human information detection method that combined improved YOLOv3 network and calling human information recognition module was proposed. The K-means++ algorithm was used to obtain the prior frame of the data set; the new bounding box regression loss function GIoU was used to improve the detection accuracy, and then multi-scale training was performed to obtain the human detection network model. Finally, the human detection model was used to detect human targets; and the human body information recognition module was used to analyze and save human body information. The experimental results showed that the method could detect human targets quickly, and accurately obtain various attribute information of human targets. Among them, the mAP of human detection model on the test set reached 91.8%, and the recognition speed was 45 f/s.
YAO Li, DU Junkang, LI Changshun
Abstract: This paper focused on solving the problem of artifact in image mosaic technology, and proposed a whole set of mosaic scheme to solve the problem. Image stitching method was divided into two stages: image registration stage and image fusion stage. In the stage of image registration, in this paper, an image registration method based on grid optimization was proposed. By using the feature extraction algorithm combining point feature and line feature, the quantity and quality of features could be improved. At the same time, a prior screening method of RANSAC feature point pairs was proposed, which could eliminate mismatching point pairs and improve the speed of model calculation. In order to further improve the alignment effect and optimize the artifacts, a number of image alignment constraints were proposed. Finally, the cylinder projection model was used to project the image pairs to the same plane. In the stage of image fusion, a seam-line detection algorithm based on the enhancement of the details of the overlapping areas of the image was used to optimize the stitching effect in the multi-focus scene. The gradual image fusion algorithm was used to fuse the projection image of the same plane, and a wide field of vision high-resolution image mosaic result was obtained. The existing splicing methods were compared with the method in this paper. The results showed that the methods in this paper maintained the PSNR value verified in multiple scenarios in the range of 35-40 dB with low distortion, which could deal with the problem of artifacts better, and could have better performance in image stitching quality.
SUN Ning1, WANG Longyu 1,2, LIU Jixin1, HAN Guang1
Abstract: In the scene recognition, RGB images present appearance information and depth image contains geometry information,which complement each other. In order to use the complementary information contained in the depth images and the RGB images in the test phase with only RGB images, this paper uses the depth image as the privilege information, and proposes an end-to-end trainable deep neural network model to combine the privilege information and attention mechanism. In the proposed method, the image encoding, feature decoding and then image encoding are used as the fr<x>amework to establish a mapping relationship from RGB images to depth images to high-level semantic features of depth images. By using of the attention mechanism, the high-level semantic features of RGB images are fused with the corresponding high-level semantic features of the depth image. And these two features are fed into the classification network to make the final prediction. In the test phase, only RGB images need to be used, the performance of scene recognition can be improved with the help of privilege information extracted from depth image. Extensive experiments are conducted on two RGB-D scene recognition benchmarks including SUN RGB-D and NYUD2, the validity of the proposed method in this paper is verified.
ZHAO Junjie1, WANG Jinwei1, 2
Abstract: Due to adversarial examples′ serious interference to the detection models based on deep learning, a recovery method of adversarial examples based on stochastic multihlter statistical generative adversarial network (SmsGAN) was proposed in this work. To achieve high-precision forensics of adversarial examples, this paper proposed the feature statistical layer in the stochastic multihlter statistical network (SmsNet). The feature map output from each convolution layer was directly transferred to the feature statistical layer to get global feature values. Stochastic multihlter statistical generative adversarial network (SmsGAN) used SmsNet as its discriminator, and its generator used a multi-scale convolution kernel parallel structure to avoid checkerboard artifacts. The generator′s loss function consisted of two parts, discriminative loss and guidance loss, to form a target guidance generator. The adversarial examples entered the down-sampling network to obtain local statistical features, and then these features were sent into SmsGAN for reconstruction to get denoised examples. Using SmsGAN to recover the adversarial examples, the recovery rate reached 91.3%, and the average PSNR reached more than 32. The visual quality was better than the traditional signal processing method, and the purpose of removing the anti-disturbance was achieved.
LIU Yuxiang, ZHANG Maojun, YAN Shen, LI Jingbei, PENG Yang
Abstract: The selection of the initial image pair was the key to the incremental structure from motion (SfM). However, traditional selection methods had some problems such as low computational efficiency and poor robustness in some special scenes. In this paper, an initial image pair selection network based on multi-task learning was proposed to improve the efficiency of selection, and a selection strategy combined with the scene connection graphs was proposed. The strategy firstly constructed the topological structure between the images, and then judged whether the initial image pair was in the center area of the scene based on the density of the connections between the images, so as to avoid the incomplete reconstruction in some special scenes due to the selected initial image pair being in the edge of the whole scene. Compared with traditional SfM (structure from motion) methods, the selecting speed of the proposed method in a variety of different scenes was increased by more than 5 times. At the same time, the proposed selection strategy combined with scene graphs could increase the number of reconstructed spatial points in special scenes by 10 times, and reduce the reprojection error by 0.05 px, which significantly improved the robustness of the initial image pair selection in special scenes. This proved the effectiveness of the proposed method. While improving the efficiency, it could ensure the completeness and stability of the reconstruction of special scenes.
YAN Fuyou1, CUI Hao2, LI Junchao1, GAO Xinjun1
Abstract: Based on the concept of modified mean effective stress, the effect of cementitious bonding and the loss of bonding during loading for cement-treated clay were described, in which most of the parameters could be got automatically by program calculation except for a few parameters estimated by other software. And then, the bounding surface model with two surfaces was developed in order to describe the behaviour of cementation and cementation degradation of cement-treated clay, and the basic model parameters were selected in the deviator strain and deviator stress plane from undrained shear tests. The reliability and correctness of the proposed constitutive model and their parametric estimation methods were shown by comparing the simulation and test results of the undrained shear tests from the literature under different confining pressure for different kinds and different cement contents cement-treated clay. Finally, to further illustrate the advantages of the proposed model, a numerical example was conducted on triaxial undrained shear test under cyclic loading condition for cement-treated clay, and the results showed that it could reveal the main stress-strain characteristics of cement-treated clay such as stress, excess pore water pressure and stress path and so on under cyclic loading condition.
LI Rong, YANG Meng, LIU Linxia, LIANG Bin
Abstract: In order to overcome the lack of prediction method for the vibration of submerged functionally graded (FG) cylindrical shell, a prediction method for the critical buckling pressure and natural frequency of submerged FG cylindrical shell was presented. The coupled vibration characteristic equation of system was established based on the Flügge theory and wave propagation method. The data of natural frequencies under different hydrostatic pressures could be obtained by solving this equation. The results showed that the squared fundamental natural frequency of submerged FG cylindrical shell was linearly related to hydrostatic pressure. Based on this conclusion, the predicted values of critical buckling pressure could be obtained accurately by using three data values of natural frequencies. In addition, the data of natural frequency for FG cylindrical shell under any hydrostatic pressure could be easily got by using this linear fitting method. The calculation results indicated that the predicted results were consistent with the reference results. The validity and usefulness of the proposed method in engineering application were proved.
ZHANG Wengang1, WANG Fang2, DING Longting3
Abstract: Snowfall in northern China often occurs, and melting snow salt, which was mainly composed of sodium chloride, was widely applied to remove snow. And at the same time, it had a serious corrosion effect on the asphalt pavement. In this paper, melting snow salt was used to prepare different concentrations of solution. And different types of asphalt mixtures, such as Stone Matrix Asphalt (SMA-13), Asphalt Concrete (AC-13) and Open-graded Friction Courses (OGFC-13) were tested in different corrosion conditions——different concentration of melting salt solution and different drying-watering cycle times. The dynamic stability (DS), residual marshall stability (MS′), freeze-thaw splitting strength ratio (TSR), and maximum flexural-tensile strain (εB) were tested after corrosion, and the change rules were analysed also. The residual pavement performance percentage of asphalt mixtures was predicted at last. The results indicated that as the concentration of melting salt solution and the drying-watering cycle times increased, the performances of asphalt mixtures were all became worse and worse, the corrosion effect of melting snow salt on asphalt mixture was serious, especially on OGFC.
WANG Decai1,2, HAO Peiwen3, SUN Yang2, LI Ruixia2
Abstract: Aiming at the stability of emulsified asphalt for cold regeneration, laboratory tests were carried out on the types and dosage of emulsifier, particle size, Zeta potential and interfacial tension of SBR modified emulsified asphalt, and the index requirements for characterizing the stability of emulsified asphalt were put forward. The results showed that the particle size and distribution of the emulsified asphalt prepared by different emulsifier types were different. With the increase of the amount of emulsifier, the average particle size, D90 and ununiformity coefficient regeneration and the stability increased. The type and amount of emulsifier had a significant effect on the Zeta potential and interfacial tension. As the amount of emulsifier increased, the Zeta potential increased significantly and the interfacial tension decreased. When the amount reached a certain level, the emulsion stability tended to be stable. The particle size and distribution of SBR modified emulsified asphalt were basically the same as those before unmodified, but the Zeta potential increased, and the storage stability slightly reduced. The grey entropy correlation analysis showed that the interfacial tension and particle size ununiformity coefficient had the best correlation with the stability of emulsified asphalt, followed by D90 and average particle size, and the Zeta potential had the smallest. It was recommended that when D90≤7 μm, average particle size ≤4 μm, ununiformity coefficient ≤4, Zeta potential>28 mV, interfacial tension <3.3 mN/m, it could meet the stability requirements of the current specification.
LIU Hailin, DU Siyi, BAO Penghui
Abstract: The development of BIM technology was the key to the transformation and upgrading of China′s construction industry. And the ground-based inspection BIM technology still remained explored.Based on the application flow of the foundation-based inspection information management system, this paper proposed a ground-based detection information model that would adapt to the ground-based detection content without changing the existing inspection workflow. That was, the foundation-based detection P-BIM model could ensure the information exchange between the ground-based detection and other participants. The informationization level of foundation-based testing was further improved, and the comprehensive supervision of ground-based testing was strengthened, which was conducive to the healthy development of the ground-based testing industry.
MEI Shengkai,LI Song, YUAN Wei, GUO Qianjian
Abstract: In order to improve the predictive ability of the error model of the five-axis machine tool, an indirect error measurement method of the five-axis machine tool based on the sample test method was proposed, and the optimized design of the stepped axis sample was completed. By inversely calculating the spatial error model of the machine tool, the function analysis relationship between the sample machining error and different error terms of the machine tool was established, and then the dynamic errors of the machine tool in different machining states were obtained. The reliability of the method was verified through experiments. The results showed that the method was in good agreement with the laser measurement results. The method could improve the predictive ability of the error model of the five-axis machine tool, thereby improving the machining accuracy of the machine tool.
WANG Junlei1, ZHANG Chengyun1, CHEN Weizhe1,2, WU Yipeng3, WANG Dingbiao1, JIN Zunlong1
Abstract: In this paper, through wind tunnel test, the energy harvesting characteristics of the square column oblique body (θ =30°,45°,60°) and the bare square column with galloping piezoelectric energy harvester were studied. The variation rule of their output power and voltage under different wind speed were analyzed. The results showed that under the optimal load R=0.6 MΩ, the output voltage and power of the bare square column and the square column oblique body (θ =30°, 45°, 60°) piezoelectric energy harvester increasd with the increase of wind speed, while the growth rate showed a trend of increasing first and then decreasing. The output voltage and power of the bare square column were greater than that of the square column oblique body. When the wind speed was U=3.194 m/s, the output power collected by the bare square column galloping piezoelectric energy harvester was 0.006 3 mW, which was 1.28 times, 1.39 times and 1.55 times of the square column oblique body with the oblique angle θ of 30°, 45° and 60°, respectively. Under the same wind speed, the output voltage and power of the square column oblique body with the oblique angle θ of 30° were the highest, followed by 45°, and the lowest was 60°. The study of this paper could provide reference for the design of relevant piezoelectric energy harvester and the design of vibration reduction in engineering.
LI Yumin, LIU Yang, WANG Xianshun
Abstract: It was significant to delimit reasonably the space radiation scope of airport aviation logistics and to study on the evolution pattern of airports in the development and decision-making of airport. The principal component analysis method was used to evaluate the aviation logistics comprehensive strength of 19 large airports in 2007,2012 and 2017 in China.On the basis of this, the weighted Voronoi diagram method based on the breaking point model was adopted to divide the aviation logistics radiation scope of each airport, which revealed their evolution pattern and influencing factors of evolution in recent years. The paper drawed the following conclusions: China′s eastern and southern airports had strong comprehensive strength in aviation logistics, that of western and northern regions was weak. In recent years, most of the airports that their aviation logistics radiation ranges expanded, were located in the central and western regions of China. Most of the airports that their radiation ranges shrank and had no significant change were located in the southeast coast, north and northeast of China. The overall economic and export-oriented industry development of the city where the airport was located had gradually became key factors affecting the development of airport aviation logistics.
Copyright © 2023 Editorial Board of Journal of Zhengzhou University (Engineering Science)