2017 Vol. 43, No. 1

Display Method:
Progresses in Pigeon-inspired Optimization Algorithms
DUAN Haibin, YE Fei
2017, 43(1): 1-7. doi: 10.11936/bjutxb2016090003
Abstract:

In recent years, the bio-inspired intelligent optimization has always been a very popular field of study in intelligence computing and has been widely applied in life science, system science, control science, computer science, management science, sociology and other subjects. The pigeon-inspired optimization (PIO) algorithm is a swarm intelligent optimization algorithm that proposed in recent years, which is inspired by the autonomous homing behavior of pigeons in nature. In this paper, the nature of the pigeons mechanism and flock optimization basic principles are described, and the latest developments of flock optimization model were introduced. The typical applications of unmanned aerial vehicle (UAV) formation, control parameter optimization and image processing were reviewed. Finally, the future development direction was forecasted.

Risk Pre-warning Model of Doctor-Patient Relationship Based on Particle Swarm Optimization BP Neural Network
WANG Zongjie, XING Mingfeng, WANG Hongbo
2017, 43(1): 8-12. doi: 10.11936/bjutxb2016040071
Abstract:

In this paper, an optimized modeling method based on particle swarm optimization (PSO) toward back propagation (BP) neural network was proposed to raise the prediction accuracy for the doctor-patients risk pre-warning case. The PSO method was applied to optimize the initial weights and biases of the conventional BP neural network to raise the prediction accuracy. By contrast and analysis of the results, the optimized method achieved a more effective prediction with much lower error. Therefore, the proposed PSO-BP neural network provides a more promising prediction method with faster convergence and higher accuracy.

Detecting Functional Module Method Based on Cultural Algorithm in Protein-protein Interaction Networks
JI Junzhong, GAO Guangxuan
2017, 43(1): 13-21. doi: 10.11936/bjutxb2016100034
Abstract:

To achieve function module detection in protein-protein interaction (PPI) networks, a PPI network functional module detection method based on cultural algorithm (CA-FMD) was proposed. First, an ordered adjacency list encoding scheme was used to model an individual in the population space. Then, the evolutionary mechanism of cultural algorithm was designed and employed to obtain the optimal solution, where the upper mechanism simulated the evolution of the group experience in the belief space, and the lower mechanism described the evolution of individuals in the population space. Finally, the optimation of solutions was completed by the interaction and influence of the two spaces. Experimental results on three datasets show that the CA-FMD method has obvious advantages in some evaluation metrics compared with other algorithms.

PCA Shuffling Initialization of Convolutional Neural Networks
LI Yujian, SHEN Chengkai, YANG Hongli, HU Haihe
2017, 43(1): 22-27. doi: 10.11936/bjutxb2016060070
Abstract:

To initialize convolutional neural networks better, an effective method named principal component analysis (PCA) Shuffling initialization was proposed. The method consisted of three steps. First, for the first convolutional layer, all receptive field of each feature map on training set was sampled. Then, principal component analysis of image patches separately for each feature map was conducted, and projection matrix was used to initialize filter of first convolutional layer. Finally, the first two steps on the other convolutional layers layer-wisely were performed. Experimental results on MNIST and CIFAR-10 dataset show that the proposed initialization has advantages of accuracy and speed of convergence compared to the common method such as random initialization and Xavier initialization.

Dynamic Convolutional Neural Network Extreme Learning Machine for Text Sentiment Classification
JIA Xibin, LI Ning, JIN Ya
2017, 43(1): 28-35. doi: 10.11936/bjutxb2016040093
Abstract:
Aim

ing at improving the generalization performance of the dynamic convolutional neural network on text sentiment classification, a dynamic convolutional extreme learning machine algorithm was proposed. This algorithm modified the output layer of dynamic convolutional neural network by replacing the fully connection layer with the shallow random neural network. By utilizing the perturbation ability of the random generation of parameters, it is prone to mitigate the dependence on training samples and avoid over-fitting to improve the classification performance. Experiments on several public data sets show that this approach outperforms the dynamic convolutional neural network and extreme learning machine under the evaluation metrics including accuracy rate, F1-measure, etc.

Machine Translation of Mongolianand Chinese Natural Language Based on Statistical Analysis
SU Yila, WU Nier, LIU Wanwan
2017, 43(1): 36-42. doi: 10.11936/bjutxb2016070044
Abstract:

In order to change the current situation of the development of Mongolian and Chinese Machine Translation in Inner Mongolia. A machine translation method was presented based on statistics by using phrase as the basic unit of translation. A maximum entropy model was selected, methods of participle and word alignment were provided. The translation was output through the results of adjustment. Results show that the BLEU value obtained by the method is improved. The method proposed in this paper can provide reference for the application study of Mongolian and Chinese.

Image Super-resolution Reconstruction Based on Residual Dictionary Learning
DU Weinan, HU Yongli, SUN Yanfeng
2017, 43(1): 43-48. doi: 10.11936/bjutxb2016060049
Abstract:

In this paper, an image super-resolution method was proposed based on 2D separable dictionary and residual dictionary to improve the quality of reconstructed image and to preserve space information as well as high-frequency information. Unlike the conventional super-resolution method based on 1D dictionary, the 2D separable dictionary was constructed from a set of training images in 2D matrix form without vectorization, so it had the capacity of learning the inherent texture structure of images. Additionally, it was a pair of compact dictionaries that were of smaller size and less storage space compared with the conventional dictionary. To restore more high-frequency information, the residual dictionary was introduced based on the reconstructed images with the 2D separable dictionary, which captured the high-frequency information of edges, angles and corners in images. Combining the two dictionary learning procedures into one framework, the proposed method was expected to synthesize high resolution images with high quality. The proposed algorithm was tested on public natural image set. The experiment results show that the proposed image super-resolution method based on 2D separable dictionary and residual dictionary is effective and superior.

2-UPS/RR Parallel Mechanism Used in Human Hip Joint Power Assist and Kinematic Performance
LI Jianfeng, LI Shicai, TAO Chunjing, JI Run, XU Chenghui, ZHANG Zhaojing
2017, 43(1): 49-57. doi: 10.11936/bjutxb2016020013
Abstract:

To help patients with impaired hip joint walking exercise, a 2-UPS/RR parallel mechanism was presented, which is suitable for hip joint assisting. It meets the requirements of hip movement for flexion/extension adduction/abduction, pronation/supination, and its mechanism centres of rotations accurately match different patient’s hip. The mobility of the mechanism was calculated, inverse kinematics was solved by using the analytical method, and the velocity Jacobian matrix was established. By demarcating the driving parameters, the work space of the mechanism was coped with. Finally, in view of the Jacobian matrix, the kinematics performance of the mechanism was analyzed when driving mobile deputy and two axes of Hooke deputy, respectively. Results show that the mechanism is of favourable operability, flexibility and stiffness characteristics within the specified workspace. Compared wtih some existing hip joint power assist robots, the mechanism in simple structure can effectively avoid unnecessary interference between branches and significantly reduce the occupancy volume of the entire unit, absolutely with more advantages. Therefore, the designed mechanism is suitable for hip assist training.

Identification of CNC Machine Tools’ Geometric Errors Based on Circular Tests
WANG Min, LI Jiafu, ZAN Tao, MA Gangjian, ZHOU Shuang
2017, 43(1): 58-64. doi: 10.11936/bjutxb2016010070
Abstract:

To identify the geometric errors of three-axis vertical CNC machine tools, the circular motion trajectories of three-axis vertical machine tools were modeled by using the theory of multi-body system. Based on the numerical simulations of circular motion trajectories, the influence mechanism of the typical geometric errors on the circular motion trajectories was analyzed. Considering the influence of multiple geometric errors on the circular motion trajectory, this paper put forward a geometric errors identification method, which can identify all 21 geometric errors of three-axis vertical machine tools based on the least squares method and polynomial based error modeling. Numerical simulation results demonstrate the accuracy and feasibility of this method.

Optical Flow Based Mobile Robot Obstacle Avoidance Method in Unstructured Environment
YU Naigong, ZHENG Yuling, XU Li, CAI Jianxian
2017, 43(1): 65-69. doi: 10.11936/bjutxb2016050002
Abstract:

In order to solve the obstacle avoidance problem in complex unstructured environment, inspired by the optical flow navigation strategies of insects, an algorithm of visual obstacle avoidance for autonomous mobile robot was developed in this paper, which is suitable for complex unstructured environment. In order to reduce the influence of illumination changes and noise, improve the calculation accuracy and robustness of the optical flow algorithm, the gradient constancy assumption was introduced and the local was weighted. And the flow divergence was used to calculate the relative depth time-to-contact (TTC). Then an obstacle map was constructed based on the TTC. When robots were at risk of collision with the obstacles, the heading direction decision algorithm was designed for robot to choose the optimal direction. Simulation and physical experiments were done in this paper. Experimental results show that the presented method can guarantee a mobile robot wander without collision in unknown complex unstructured environments.

Projection Filtering Performance of Alternate Polarization Array
DOU Huijing, XIAO Dengliang, ZHANG Shaofei
2017, 43(1): 70-75. doi: 10.11936/bjutxb2016040087
Abstract:

In view of the traditional projection filtering algorithm based on polarization sensitive array, there are some problems like the large structure of the array, the high demands of polarization parameters and the uncertainty of filter loss.In this paper, alternate polarization sensitive array was chosen as signal reception model. With half equipment, the model can filter the interference signal effectively,which solved the problem of low cost structure of polarization array. Meanwhile, traditional projection filtering algorithm was improved. While oblique projection filter with unknown expected signal polarization parameters, the oblique projection filtering operator was deduced. Then the acquired operator was used to extract the interference signal, and the principle of elimination was used to obtain pure expected signal,which effectively solved the high polarization parameter demands in traditional filter. With regard to the new orthogonal projection filter, which cannot be quantitatively analyzed, an orthogonal projection filter operator was constructed to obtain the changes of filter loss. Further analysis shows that the improved oblique projection filtering can filter out the interference signal, which greatly expands the limits of the projection filtering. The improved orthogonal projection filtering make quantitative analysis of the loss of the desired signal, which has great theoretical significance.

Research Development and Prospect of Solar Cells Surface Defects Detection Based on Machine Vision
QIAN Xiaoliang, ZHANG Heqing, CHEN Yongxin, ZENG Li, DIAO Zhihua, LIU Yucui, YANG Cunxiang
2017, 43(1): 76-85. doi: 10.11936/bjutxb2016040063
Abstract:

Considering the advantages of simple operation and high detecting accuracy, all aspects involved in solar cell surface defect detection methods based on machine vision were reviewed in this paper. First of all, the various imaging techniques and common defect types of solar cells surface were summarized. Secondly, the existing detection methods were introduced and compared with each other according to the different idea of mathematical modeling. Finally, a brief summary of this article and perspective of future research are presented. It can be concluded that the solar cell surface defect detection methods based on machine vision have made great progress. However, there is still room for improvement in algorithm design of feature extraction, such as feature extraction algorithm based on deep neural networks.

Discrete-time Observer Repetitive Control System Based on a Two-dimensional Model
XIE Wei, DUAN Jianmin, FANG Zeping, ZHENG Banggui
2017, 43(1): 86-93. doi: 10.11936/bjutxb2016030058
Abstract:

To implement less conservative than those that ignore the difference between the control and learning actions, an observer using discrete-time repetitive control (RC) design schema was presented, based on the two-dimensional (2D) system theory for linear systems with uncertainty. First, to describe both the learning process occurring between two cycles and the control behavior during a repetitive cycle, a two dimensional discrete-time model was established accordingly. Furthermore, by rebuilding the state variables, the repetitive controller designing was transformed to the feedback gain design problem. Thus, through linear matrix inequality (LMI), the stability conditions that also gained quotient of repetitive system were obtained. By comparison, the applicability of the proposal is stronger, significantly improving the robustness and tracking speed. Test numerical simulation results show that the method can achieve high control accuracy within only a few learning cycles, and the two-dimensional system theory has broad application prospects in repetitive control design and analysis.

Microblog Retrieval Results Re-ranking Using Graph Model Based Decision
YANG Zhen, ZHANG Guangyuan, FAN Kefeng
2017, 43(1): 94-99. doi: 10.11936/bjutxb2015090041
Abstract:

As a typical short text, microblogging retrieval suffers from the problem of the insufficient samples both in users’ query and documents that makes the probabilistic-like models unreliable. To remedy this problem, a graph model was designed and implemented based on topic clustering algorithm to re-rank microblog retrieval results. The graph model was built by the content similarity between micro-blogs. By comparing the cosine similarity, the dice coefficient, and the one-way dice coefficient with the experimental results. Results show that the performance of the search depends on the ratio of related topics, therefore decision tree algorithm was used to remedy the influence of the ranking position relevant topics.

Scheme of Trusted Bootstrap Based on General Smart Card
YAN Lin, ZHANG Jianbiao, ZHANG Ai
2017, 43(1): 100-107. doi: 10.11936/bjutxb2015100063
Abstract:

The risk of the key authentication information being bypassed and the potential safety hazard of booting data being tampered with both exist in the booting mechanism of the traditional operating system. Based on the theory of trusted computing, combined with the technology of smart card with CD-ROM file system, a scheme of trusted boot based on general smart card was proposed. Without changing the structure of hardware and firmware of the smart card and terminal device, through the transformation of storage data in the smart card and disk booting data, the security objective of binding the user’s identity information, the smart card and the terminal device were achieved. The trusted computing mechanism was extended from power on to the application layer to ensure that the initial state of operating system was trustworthy. Through the analysis of security and performance, the security of terminal device bootstrap was proven, which has been verified in practical applications.

Research of Strength and Freezing-thawing Durability of Saline Soil Solidified by Modified Sodium Silicate
LÜ Qingfeng, MENG Huifang, WANG Shengxin, ZHANG Qing, CHEN Hui
2017, 43(1): 108-112. doi: 10.11936/bjutxb2016070045
Abstract:

In order to solve relevant engineering problems about sulphate salty soil and the collapsibility of saline, research of saline soil solidified by coal ash which was stimulated by temperature modification sodium silicate, was conducted in this paper. Strength and freezing-thawing durability of saline soil solidified by sodium silicate modified by temperature were studied by means of freeze-thaw cycle test, unconfined compressive strength test, X-ray diffraction spectra and microstructure test. The result shows that the strength of saline soil solidified by modified sodium silicate decreases against the rise of temperature. The shear strength of saline soil solidified by sodium silicate which is modified under different temperature decreases while the times of freeze-thaw cycle tests increase and its decreasing range is lessen after 10 times. There is no appearance of new crystal materials in saline soil solidified by sodium silicate modified by temperature. With the increasing of modification temperature, the diffracted intensity of mirabilite and dolomite gets enhanced, the apparent porosity becomes bigger, and strength decreases. The microstructure is damaged after freeze-thaw cycle tests.

Oretical Analysis on the Dynamics of Line Plume in Thermal-stratified Environment
LI Junmei, ZHANG Ren, LI Yanfeng, ZHAO Yuhang, TIAN Yang
2017, 43(1): 113-117. doi: 10.11936/bjutxb2016050010
Abstract:

Thermal-stratified environment is often seen in atria or other large space buildings with glass roofs due to the solar radiation. In order to study the effect of the thermal-stratified environment on the smoke spread in fire cases, theoretical analysis of the thermal dynamics of the line plume in the environment with constant vertical temperature gradient was carried out based on the former research in this paper. Variations of the radius, central velocity, density difference of the line plume with height were derived, and the maximum height of the plume was obtained. The upward buoyancy due to the line source changed its direction at the height of 76% of the maximum height which the plume could rise.

Effect of Shaking Table Errors on Specimen Response and Its Correction Measures
LIN Shuchao, TANG Zhenyun, HUANG Li, GUO Jun, LI Zhenbao
2017, 43(1): 118-126. doi: 10.11936/bjutxb2016040073
Abstract:

Shaking table test is the main testing method for engineering structures research, which plays an important role in the research for seismic performance of structures. In order to improve control accuracy problems of shaking tables, and to get the authentic structural response, firstly, the influence of the shaking table dynamic on structural response was analyzed. And then a correction measure was developed based on the input and output of the shaking table. Correction method of better compensation of the influence of vibration table error was provided. And specimen corrected response can be better representative specimen’s response at the desired ground motions. Finally, the performance of the developed method was verified experimentally.

Influence of Mineral Mixture Performance on Dynamic Modulus of Hot Mix Asphalt
MIAO Yinghao, ZHENG Xiaoheng, WANG Wentao
2017, 43(1): 127-134. doi: 10.11936/bjutxb2016040043
Abstract:

In order to solve the problem of high temperature performance of asphalt mixture, an investigation into the influence of mineral mixture performance on the dynamic modulus of hot mix asphalt (HMA) was conducted. Nine various AC-20C gradations were determined in accordance with a Bailey method parameters based on orthogonal experimental design. The performances of the mineral mixtures with each gradation were evaluated by using the California bearing ratio (CBR) and the voids of vibrated aggregate (VVA) in laboratory. Three gradations with significantly different mineral mixture performances were picked out to make HMA. Then, the dynamic moduli of HMAs corresponding to each of the 3 gradations were tested under multiple temperatures and loading frequencies. The influence of mineral mixture performance on the dynamic modulus of HMA was analyzed in accordance with the experimental results. The results show that the mineral mixture performance has significant influence on the dynamic modulus of HMA at high temperature, while has almost no influence at low temperature. The high temperature performance of HMA can be improved through investigating the CBR and VVA of aggregate in design. Higher CBR and lower VVA imply better high temperature performance of HMA.

Road Performance and Antifreeze Property of Antifreeze Micro-surfacing Asphalt Mixture
WU Ping, WANG Xuancang
2017, 43(1): 143-149. doi: 10.11936/bjutxb2016080014
Abstract:

To delay the snow or ice forming on road pavement, as well as to prevent the pavement-ice bonded layer, antifreeze micro-surfacing asphalt mixtures were designed based on the volume displacement method. Two types of asphalt mixture replaced with aggregate-salt and filler-salt were investigated by Cantabro test, loading wheel test, abrasion test, texture depth test, and conductivity test respectively. Results indicate that the abrasion test value increase by the addition of salt, meanwhile, the rutting ratio increases and the texture depth of the surface is reduced. It is suggested that the engineering performance of micro-surfacing containing filler-salt is better than that of containing aggregate-salt. In actual application, the mixture’s type, salt content, environmental temperature and service time have great impacts on the antifreeze performance of antifreeze micro-surfacing asphalt mixtures.

Analysis of the Significant Level of Sludge Rheological Models With Different Moisture Contents
CAO Xiuqin, YIN Weiqi, ZHAO Zhendong
2017, 43(1): 150-157. doi: 10.11936/bjutxb2016020015
Abstract:

There are seven typical non-Newtonian rheological models. Significant level of these models was analyzed by combining with sludge in order to establish the relative optimal model. The rheological experiments of sludge with moisture content in the range of 93.99%-98.72% were carried out at 20℃ by using a rotating viscosity meter, and sludge rheological characteristics were analyzed. The seven rheological models were fitted with the experimental results and white noise test was carried out to verify the effectiveness. Results display that the moisture content has bigger effect on the viscosity of sludge at the low shear rate of 0-150s-1. The apparent viscosity is below 0.5Pa·s when the sludge is in the range of moisture contents of 96.31%-98.72%, which is suitable for long-distance transportation. While the sludge in the range of moisture contents of 96.31%-98.72% is not suitable for long-distance transportation as it has higher apparent viscosity. The significance level of Power-law model is better than the other two parameter models when the moisture contents are among 93.99%-95.52%. When the moisture contents are among 93.99%-98.72%, the Sisko one is better than the H-B among the three parameter models, the Carreau model is better than the Cross among the four parameter models.

Numerical Simulation of Temperature in the Adsorbent Bed in Insolation
YUAN Zhongxian, GAO Dongdong
2017, 43(1): 158-162. doi: 10.11936/bjutxb2016010049
Abstract:

To deeply study the internal temperature changes of the adsorbent bed in insolation, based on the fully understanding of solar radiation variation regularity and the configuration of evacuated tube bed, a two-dimensional model of heat transfer was established. To identify the effect of the internal cooling mode, the temperature field in the adsorbent bed under two cooling conditions was investigated with FLUENT techniques. Results show that the maximum temperature difference for the uniform wall temperature condition and natural convection of the air condition can reach 100℃. Meanwhile, the temperature distribution of the bed presents an obvious non-uniform phenomenon. It can be concluded that the uniform wall temperature condition has an obviously better cooling effect than that of the natural convection of the air condition. In addition, the non-uniform distribution of the temperature in the bed hinders the adsorption cooling performance.