Table of Content

    16 December 2011
    Volume 41 Issue 6
    A method of feature selection for continuous attributes
    LI Guo-he1,2, YUE Xiang1,2, LI Xue3, WU Wei-jiang1,2, LI Hong-qi1
    Abstract ( 348 )   PDF (1089KB) ( 851 )   Save
    Related Articles | Metrics

    Feature selection is one of the methods for reduction of data sets, which improves efficiency and effectivity of machine learning. In terms of the distribution of objects and their classification labels, the continuous feature space was partitioned into a variety of subspaces, each one with a clear edge and unique classification label. After the projection of all the subspaces for  each feature, the quality of each feature was  estimated for a subspace opposite all  the other subspaces with different classification labels by means of statistical significance. Through construction of a matrix by all the estimate qualities of all features of  the subspaces, all  features were ranked from the highest classifying power to the lowest on the matrix for the feature space. After the information gain function was defined by the subset of features, the feature subset was optimally determined on the basis of ranked features by gradually adding features. Experiments on the data sets from UCI(University of California Irvine) repository by the feature selection obtained feature subsets,  by which the performance and classification accuracy of machine learning were improved, illustrating that the feature selection was feasible.

    An improved extreme learning machine based on Akaike criterion
    YIN Jian-chuan1,2, ZOU Zao-jian1,3, XU Feng1
    Abstract ( 379 )   PDF (1008KB) ( 934 )   Save
    Related Articles | Metrics

    To reduce the dimension of a neural network and improve the generalization capability of the extreme learning machine (ELM) network, Akaike information criterion (AIC) was implemented to choose a suitable number of hidden units, and the modified Gram-Schmidt (MGS) method was also implemented to automatically adjust the network parameters. In comparison with the conventional ELM learning method on several commonly used regressor benchmark problems, the improved ELM algorithm could achieve  a  compact network with much faster training speed and satisfactory accuracy.

    Feature engineering for Chinese part-of-speech tagging
    YU Jiang-de1, ZHOU Hong-yu1, YU Zheng-tao2
    Abstract ( 395 )   PDF (977KB) ( 1809 )   Save
    Related Articles | Metrics

    Context features have a major impact on  the performance of Chinese part-of-speech tagging. In order to improve  the performance, the feature engineering for Chinese part-of-speech tagging was explored by the using maximum entropy model. Two key issues of feature engineering, the size of the feature window and the feature templates, were  studied. Closed evaluations were performed on PKU, NCC and CTB corpus from the Bakeoff-2007. Then,   comparative experiments about the training process and tagging accuracy for Chinese part-of-speech tagging were performed on different feature windows,  the “5 words” and “3 words” feature windows, and different feature templates: single-word, doubleword and mixing feature templates. Experimental results showed  that the feature window including 3 words was better  than that of 5 words, and the performance increased 10% using single-word feature templates than double-word feature templates. All the results  showed  that the feature window including 3 words and single-word feature templates were  appropriate for Chinese part-of-speech tagging.

    Study of bilingual words of part-of-speech(POS) disambiguation in the English-Chinese parallel corpus
    FENG Min-xuan1, QU Wei-guang2,3*
    Abstract ( 362 )   PDF (954KB) ( 1486 )   Save
    Related Articles | Metrics

     A part-of-speech disambiguation approach was given based on idiosyncratic rules in a parallel corpus unaligned at the lexical level. This approach focused on those words that occurred in the corpus at  very high frequency, while the part-of-speeches were difficult to determine. A number of idiosyncratic disambiguation rules were  constructed and an algorithm built on these rules was  applied on five typical words, among which were three Chinese words, “guoqu”, “jihua” and  “yu” and two English words, “back” and “so”. Experiments on a large scale parallel corpus obtained an F-score of 98.45% for the disambiguation of these words, and the results showed that the constructed rules would not be constrained by the length of context and the number of templates.

    An activity mining model for surveillance video
    LIANG Hao-zhe, XU Shu-kui, LI Guo-hui, ZHANG Jun
    Abstract ( 339 )   PDF (2979KB) ( 1158 )   Save
    Related Articles | Metrics

    The activity pattern mining technique is  the  key component of semantic analysis for surveillance video. Because of the lack of prior and highdimensional feature constraints, the complexity of the model structure of the parametric mining model is  difficult to be precisely defined. Nonparametric clustering of motion feature by infinite Gaussian mixture was used to get the elementary activity patterns, based on which duration distribution was estimated. The  partial-dimension test for feature validated the motion similarity hypothesis existing  in the mining model. The results showed that the obtained activity patterns precisely reflected motion semantics of the scene, and that the multi-modality temporal distribution existing  in activity can  be further used to discover the hidden knowledge of motion.

    An outlier detection algorithm based on attribute reduction and relative entropy
    HU Yun1,2, LI Hui1, SHI Jun1, CAI Hong1
    Abstract ( 297 )   PDF (1310KB) ( 1047 )   Save
    Related Articles | Metrics

    A new outlier detection algorithm combining a  rough set and information entropy technology was proposed. This approach could obtain similar outlier sets by means of searching in an attributes subspace, which  could lead the analysis of outlier detection to focus better on narrow and specific object fields. This algorithm divided the original attribute space into several segments, which filtered out those subjects with largest relative entropy negative relative cardinality as the outliers. To prove this algorithm’s effectiveness,  experiments on a  real world dataset were conducted. Theoretical analysis and experimental results showed that this method of outlier detection was efficient and effective.

    Community mining on non-binary graph sequences
    TANG Jun, CHEN Song-can*
    Abstract ( 371 )   PDF (1303KB) ( 1186 )   Save
    Related Articles | Metrics

    Against the defects of existing graph sequences community mining methods, a community mining method on nonbinary graph sequences based on the minimum description length principle was proposed. According to its nature of complete NP-hard problem, it was processed by  preprocessing on the problem and  a relatively good  initial input was obtained. Based on the concept of graph sequences coding length, an optimization problem was solved by regrouping rows and columns to  integrate  gray information. And then a  community mining problem was effectively  solved. It could avoid being trapped in the local minimum by using the random and optimization mind of genetic algorithm in the processing. In addition, the change of community structure could be detected with passage of time which is  critical for reality problems. Finally,  an experiment validated the effectiveness of this method and its high performance.

    A new clustering algorithm for user access patterns based on network virtual environments
    CHEN Ming-zhi1, 2, CHEN Jian3, XU Chun-yao3, YU Lun3, LIN Bo-gang1, 2
    Abstract ( 363 )   PDF (1402KB) ( 1083 )   Save
    Related Articles | Metrics

    In order to efficiently implement  personalized information services in network virtual environments, a new clustering algorithm for user access patterns was proposed, which was  the MPF, i.e. the fuzzy C-means (FCM) clustering algorithm based on multi-objects particle swarm optimization (MOPSO). The MPF could combine the respective advantages of PSO and FCM. Through the global spatial search of PSO, it could avoid that  FCM was susceptible to initial value, noisy data and easily falling into the local optimum. In order to improve the clustering effect,  a particle fitness function was designed based on dualobjectives (intra-class distance and inter-class distance) in PSO. Finally, the standard data set and simulation data set were applied to verify the effectiveness of this MPF. Experimental results showed that this algorithm had  good performance in clustering precision.

    Quadratic filtering of stochastic systems with multiplicative
    noise and delayed measurements
    XING Guo-jing, ZANG Cheng-hui*, ZHAO Hui-hong
    Abstract ( 293 )   PDF (1254KB) ( 1154 )   Save
    Related Articles | Metrics

    The problem of quadratic filtering for a bilinear stochastic system with state-dependent multiplicative noise  and single delayed measurements was studied. Due to the presence of multiplicative noises, the system parameter matrices  were random. So, the classical Kalman approach could not be  directly  used in the presence of multiplicative noises and would delay the measurements. Based on Kronecker algebra, the original system was changed into a linear augmented system, whose states and observations included the original states, observations and their second order Kronecker product. Then, the augmented system was transformed into a delay free system via the reorganized innovation approach, and the linear optimal filter for the augmented system was designed through projection theorem. Finally, the quadratic filter of the original system was derived by extracting the first n elements of the augmented state estimation. Compared with the widely used linear optimal filter, estimation accuracy of the quadratic filter increased 27%, and the overall performance was improved.

    Optimal fusion filtering for systems with stochastic parametric
    uncertainties and packet dropouts
    SUN Jia-bing1,2, ZHANG Cheng-jin1*
    Abstract ( 250 )   PDF (1269KB) ( 880 )   Save
    Related Articles | Metrics

    The distributed optimal fusion problem for the state estimation of multi-sensor discrete-time systems with stochastic parametric uncertainties and packet dropouts was studied. By introducing fictitious noises, the original system was transformed into an equivalent system without uncertain parameters. For each subsystem of the equivalent system with packet dropouts, the local filtering estimate and the local filtering error covariance were obtained by using the innovation analysis method. After the filtering error cross-covariance matrices between local estimates were obtained, the distributed optimal (i.e., linear minimum variance) fusion filters were developed by the fusion rule weighted by matrices. The simulation example showed that the fusion filter was better than each local filter.

    Numerical analysis of deforming force for the backward extrusion
    process based on slipline field theory
    QIN Xiao-qiong1,2, LIU De-xue1*
    Abstract ( 392 )   PDF (1194KB) ( 954 )   Save
    Related Articles | Metrics

    The calculation of deforming force depends on experience formula in extrusion process design. It may leads to large calculation error and can not be used in optimization design. Based on the classical plasticity theory, a more accurate and convenient calculation model of extrusion force was studied  by mathematical methods. Based on slip line field theory, the backward extrusion process with a cup-shaped mold was studied. Series slip-line field features of different compression ratio were analyzed and superposed. The stress field and deforming force were calculated. The analytical relationship between slip-line field parameters and mold geometrical parameters were analyzed. In  this way, the numerical analysis model of the deformation physical parameters of  geometric parameters was established. The analytical solution obtained  by the numerical analysis model has more obvious physical significance than that  by the slip-line field method. If deformation geometric parameters are   known, the analysis model could  be conveniently used to solve extrusion force. Finally, the extrusion force was also measured with experiments and  combined with experience formula for verification. The results showed  that the numerical analysis model calculation results were  more accurate, which could  be used as an objective function to realize processing optimization design.

    Design and implementation of the AMT data acquisition system
    based on LabVIEW
    ZHAO Mei1, 2, HU Tian-liang1, 2*, ZHANG Qian3, ZHANG Cheng-rui1, 2
    Abstract ( 321 )   PDF (2639KB) ( 1018 )   Save
    Related Articles | Metrics

    In order to meet the requirement of date acquisition of an automatic transmission system, an AMT data acquisition system based on LabVIEW was designed. The basic principle of the AMT data acquisition system was  introduced, the basic system structure was built, the hardware and the software design  written based on LabVIEW were studied, and the function of the data acquisition that based on LabVIEW was analyzed. The results showed that the function of the data acquisition, display and storage were  realized. With this design, the user interface could be made more intuitive and easier to operate, so that the engineers could promptly get the system run-time information.

    Study of an  automatic test system for local fans
    ZHANG Yong-chao, ZHANG Yong-jian, CHEN Qing-guang, ZHANG Zhen-dong, LI Jing-lei
    Abstract ( 373 )   PDF (1616KB) ( 956 )   Save
    Related Articles | Metrics

    To improve the safety of coal production and reduce the production energy consumption, it is necessary to test and evaluate the performance of lots of local fans  running in a coal mine production site. Software processing technology,  sensor technology and  automation technology were used to construct the general structure of the performance testing system for local fans. In addition, the key problems related to this system such as  pressure measurement, flow adjustment, performance curve fitting and software filter were studied. When  practically applied to the developed system, the  accuracy, rapidity and automatic operation of local fan performance test was realized. On this basis, the developed performance automatic test system of AFT-2 local fan was widely used at the  production site, proving the feasibility and applicability of this research result.

    Experimental  research of energy consumption of an electric forklift hydraulic lifting system with pressure accumulators
    HAO Qian-hua1, HE Qing-hua1,2*, ZHU Jun-lin1, LI Sai-bai1, CHEN Zheng1, SHU Min-fei1
    Abstract ( 412 )   PDF (1554KB) ( 1480 )   Save
    Related Articles | Metrics

    In order to reduce the energy consumption of an electric forklift, a new hydraulic lifting system was proposed, which was based on a combination of the technologies of a VVVF control and hydraulic pressure accumulator. The energy consumption of the electric forklift hydraulic lifting system with pressure accumulators and the one of VVVF controlled electric forklift hydraulic lifting system without pressure accumulators were tested under different load working conditions. The results showed that the latter system has higher energy efficiency and remarkably saves energy  compared with the former. The maximum energy saving rate of the latter system can reach 20.35%, and it has  promising value in  engineering applications.

    Simulation analysis of a seepage field in an anisotropic complex geological body
    LIANG Bing, GAI Di, SUN Wei-ji, ZHAO Fang-fang
    Abstract ( 314 )   PDF (3085KB) ( 1189 )   Save
    Related Articles | Metrics

    Based on the saturated and unsaturated seepage theory, a numerical simulation on the seepage field in an anisotropic complex geological body was done. A mine which had a  typical fold shape geological structure in Fuxin was taken as research background, a three-dimensional geological model was built using the Kriging interpolation method, and then the pressure gradient changes in the geological body  were simulated and analyzed, when the principal direction of hydraulic conductivity and global coordinate system were of different angles. The results showed that the pressure gradient  reached  the maximum at 90°. The type  of geological structure and formation were  the key factors of the direction and density of the pressure gradient.

    The effect of pre-embedded phase change materials on temperature rise caused by the hydration heat evolution of concrete
    GAO Gui-bo1, QIAN Chun-xiang2, YUE Qin-yan3, WANG Yong-wei1, LU Tong-wei1
    Abstract ( 291 )   PDF (1689KB) ( 1085 )   Save
    Related Articles | Metrics

    In order to evaluate the infuluence of pre-embedded phase change materials(PCM)for the inner temperature rise of a concrete, a semiadiabatic device was adopted to measure the  inner temperature change of a common concrete specimen and the concrete specimen with  pre-embedded PCM at different hydration ages, and the cooling influence factors of pre-embedded PCM were calculated and analyzed. The results showed that  the inner temperature of concrete that had pre-embedded PCM was related to the pre-embedded content and the specific heat of PCM, the phase transition heat and phase transition temperature of PCM, the dosage of cement materials per unit volume, the hydration heat of the cement, the specific heat, density and initial temperature of the concrete. The peak of inner temperature was reduced and the appearance time of the temperature peak was prolonged after the PCM was pre-embedded into the concrete specimen, while the inner temperature of concrete over long a period  could not be reduced.

    HPLC fingerprint spectra analysis of Buzhong Yiqi Pills and Shiquan Dabu Pills by the relative peak area difference ΔSr
    ZOU Hua-bin, DONG Feng-juan, ZHANG Xin-ling, DU Ai-qin
    Abstract ( 455 )   PDF (1208KB) ( 1092 )   Save
    Related Articles | Metrics

    An efficient and simple pattern recognition methodrelative peak area difference (ΔSr) grade sequence classifying method was established. It is suitable for analyzing HPLC fingerprint spectra of herbal medicines Buzhong Yiqi Pills and Shiquan Dabu pills based on both information of peak positions and contents of compounds, and it is qualified to recognize the two traditional Chinese compound formulae which are so-called “black box” systems. The HPLC fingerprint spectra of these two medicines’ components extracted with absolute ethanol were measured, and the ΔSr sequences of these samples were constructed. Then their most similar sample groups characteristic sequences named ΔSr grade sequences were determined depending on the similarity scale ΔSr≤ΔSr+xSv,-3≤x≤+3. In addition, clustering and classification of these samples were performed relying on the characteristic sequences. These two kinds of  medicine were accurately and subtly recognized by this new method, with the correct ratio of 100%. The results were greatly superior to those  obtained by means of coefficient r and vectorial angle cosine(cosα) methods. The results also indicated that the ΔSr grade sequence clustering/classifying method was an accurate and simple approach suitable for quantitative identification and quality control of traditional Chinese compound formula, a kind of complex “black box” system.

    Theoretical study of the inclusion interaction of β-cyclodextrin with trans-Resveratrol
    ZHOU Mei-juan1, TIAN Chun-hua1, WANG Su-na1, CHEN Xiao-hua2, LIU Ji-feng1, ZHANG Chong1*
    Abstract ( 402 )   PDF (1933KB) ( 1000 )   Save
    Related Articles | Metrics

    The inclusion process of β-cyclodextrin (β-CD) with trans-Resveratrol was studied by the ONIOM (B3LYP/6-31G*:PM3) method. The passing process and cycling process were simulated using β-cyclodextrin (β-CD) with trans-Resveratrol as monomers respectively. The two processes probed the stability of  trans-Resveratrol entering into the cavity of β-cyclodextrin from the narrow side (the primary hydroxyl group side) compared with its wide side (the secondary hydroxyl group side), the possible driving forces for the inclusion compounds, the factors influencing stability of the inclusion compounds and the driven type of forming inclusion compounds in the gas phase. The calculated data indicated that the transResveratrol could easily tend to enter into the cavity of β-cyclodextrin from the narrow side, the possible driving forces for the inclusion compounds were the charge transfer and dipoledipole interactions, hydrogen bonds played an important role in the stability of the inclusion compounds, and the formation of natural β-CD inclusion compounds was an enthalpy-driven process with temperature of 298-15 K and pressure of 1 atm in the gas phase.

    Wireless monitoring platform development for electric vehicles based on GPRS and CAN bus
    LIU Yang1, CHENG Yong1*, JI Shao-bo1, HUANG Wan-you1, LI Chuang2, ZHANG Xiao-wen2
    Abstract ( 332 )   PDF (2426KB) ( 1102 )   Save
    Related Articles | Metrics

    A wireless monitoring platform based on GPRS and CAN bus was developed to evaluate electric vehicles real-time operating performance and the matching of vehicle’s key components under  real road conditions. The vehicle motor’s and battery set’s working parameters as well as their fault messages were collected by the platform’s monitoring terminal with CAN bus, RS485, analog or digital interface and sent to the monitoring center by GPRS, and the monitoring center could monitor real-time data by transmission protocol analysis. To meet the requirements of the wireless monitoring platform, the terminal system was designed using  MC9S12XET256, and the system’s transmission protocol was constituted with data frames. Laboratory experiment and road test results showed that the transfer accuracy could meet the demands of vehicle parameters’ real-time monitoring at 5kb/s communication speed.

    Exergy analysis of the R404A/CO2cascade refrigeration system
    LAI Yan-hua1, WANG Qing-wei1, L Ming-xin1, SHAO Chang-bo2, KONG De-min1
    Abstract ( 356 )   PDF (2573KB) ( 938 )   Save
    Related Articles | Metrics

    In order to decrease the energy consumption of the R404A/CO2 cascade refrigeration system, exergy analysis was applied to this system. An exergy analysis method was adopted to obtain  the influence of the design and operating parameters on the coefficient of performance ε, the exergy destruction X and the exergetic efficiency ηe of this system. The design and operating parameters include the evaporating temperature Te, the condensing temperature Tk and the temperature difference ΔT in the cascade-condenser. The results indicate that there were a maximum ε, ηe and a minimum Xtot at the same optimal condensing temperature of the cascade-condenser T4opt when Te, Tk and ΔT were constant in the system. It is helpful to reduce X and improve ηe  and ε through increasing Te, decreasing Tk and ΔT. The total exergy destructions of the throttling device and the compressor of the R404A circuit, the cascade-condenser and the compressor of the CO2 circuit were  80% of the total exergy destruction of the system. A Multiple linear regression analysis was employed in term of  Te, Tk and ΔT to develop mathematics expressions for t4opt, εmaxand ηe,max.

    Influence of the fluctuation of module parameters on the output properties of photovoltaic arrays
    QIN Jing-yu1, SUN Cheng-shuai1,2, GU Ting-kun3
    Abstract ( 331 )   PDF (1967KB) ( 1475 )   Save
    Related Articles | Metrics

    By experiments, it was found that each solar cell in the photovoltaic (PV) module generally had different temperatures  whose value could be well approximated by normal distribution. The abnormal behavior in the I-U curve was also observed. This study showed that it was reasonable to employ the relative standard error of the temperature as that of the electrical parameters of PV modules. Based on this assumption,  cell parameters were randomly generated, and then the abnormal behavior in the I-U curve was well simulated by the series connection of these cells. The same method was applied to  a virtual 5 kW PV array, and the power loss was estimated around 200W due only  to the electrical parameter fluctuation. The results revealed that the consistency of the electrical parameters of the PV modules should be stressed in the construction of a large PV array.

    A bridge-type protection topology for 3-phase 4-wire current  output inverters and its parameter selection method
    GONG Yu-lei1,2, TAN Xing-guo2, LI Qing-min2, WANG Hui2, SUN Yong3
    Abstract ( 417 )   PDF (2761KB) ( 1323 )   Save
    Related Articles | Metrics

    The inverter’s actual model was built with the consideration of passive devices’ un-ideal models and the impact of distributed impedance. The causes and potential damages of pulsed current during the switching transition process of the current source inverter bridge in normal operation state was  analyzed in detail. A LCRD bridge protection circuit was presented. Detailed analysis as well as experiments based on this bridge protection topology fully validated that the topology could effectively decrease the pulsed current, current transition rate and voltage transition rate in the switching process, and could also protect  power devices from inappropriate triggers and possible damage. In addition, the parameter selection method was also given for establishing the bridge protection circuit.

    Location and algorithm of multi-level electric vehicle charging stations
    ZHANG Guo-liang, LI Bo*, WANG Yun-fa
    Abstract ( 515 )   PDF (1538KB) ( 2184 )   Save
    Related Articles | Metrics

    The charging station is  part  of the crucial infrastructure of  the electric vehicle, and the location and scale of the charging stations are  significant for the development of the electric vehicle industry. A location and solving algorithm of multi-level electric vehicle charging stations was proposed. Concerning  the user’ distributions,based on the objective program thought, the multi-level electric site model was developed to minimize the initial construction cost and the total users’ charging service cost. An improved tabu search algorithm was presented to solve the model. The new combination of tabu coding and the generating of initial solutions were  designed to characterize the two decision variables of the multi-level sites and the allocation to the users, using the strategy of 2-opt search neighbor,  locating sites,   station levels and the allocation of the demand points to the stations were simultaneously determined. Finally, a numerical example demonstrated that the algorithm had better global searching performance and convergence property and the proposed model was also effective.