28 December 2017, Volume 41 Issue 6
    

  • Select all
    |
  • ZHAO Baojun, LI Zhenzhen, ZHAO Boya, FENG Fan, EDNG Chenwei
    Beijing Jiaotong University. 2017, 41(6): 1. https://doi.org/10.11860/j.issn.1673-0291.2017.06.001
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    Spaceborne SAR(Synthetic Aperture Radar) ship detection has been widely used in the sea rescue, territorial security and so on. As the traditional detection methods still have some shortages like high false alarm rate, this paper introduces the convolutional neural network (CNN) that has a powerful characterization for the spaceborne SAR ship detection. Aiming at the accurate and rapid demand of SAR ship detection,it proposes a spaceborne SAR ship detection algorithm based on low complexity CNN. The algorithm first combines the characteristics of spaceborne SAR images, uses the ROI extraction method to achieve the target rough extraction, getting the suspicious target slices and their corresponding location information, then accurately classifies all the suspicious target slices by the constructed CNN with low complexity to determine the target of the ship so as to realize the target detection of the ship. The experimental results show that the algorithm can achieve accurate spaceborne SAR ship detection.Compared with the traditional two-parameter CFAR and the methods based on the existing network frameworks (LeNet, GoogLeNet), the proposed algorithm has better performance and shorter detection time, which can effectively reduce the missed rate and the false alarm rate.
  • WU Xinxiao, WU Kun
    Beijing Jiaotong University. 2017, 41(6): 8. https://doi.org/10.11860/j.issn.1673-0291.2017.06.002
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    Most video representation methods are supervised in the field of computer vision, requiring large amounts of labeled training video sets which is expensive to scale up to rapidly growing data. To solve this problem, this paper proposes an unsupervised video representation method using deep convolutional neural network. The improved dense trajectory (iDT) is utilized to extract the video blocks which alternately train the convolutional neural network and clusters. The deep convolutional neural network model is trained by iteratively algorithm to get the unsupervised video representations. The proposed model is applied to extract features in HMDB 51 and CCV datasets for tasks of motion recognition and event detection respectively. In the experiments, a 62.6% mean accuracy and a 43.6% mean average prevision (mAP) are obtained respectively which proves the effectiveness of the proposed method.
  • MA Wenting,ZHAO Yao,WEI Shikui,ZHANG Shiyin,LIAO Lixin
    Beijing Jiaotong University. 2017, 41(6): 13. https://doi.org/10.11860/j.issn.1673-0291.2017.06.003
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    At present, morphological methods for detecting diabetic retinopathy have high complexity.The traditional deep learning methods avoid the exclusion of physiological structure, manual design features.However, they have to do large calculation and the speed is relatively slow. In order to solve these problems, this paper presents a cascade detection framework based on deep learning. Firstly, the fundus images are divided into blocks to detect whether there are lesions, and then pixels in these lesions are classified into four categories:microaneurysms, haemorrhages, hard exudates and soft exudates. The experimental results show that on the public DIARETDB1 fundus image database, the detection of four kinds lesions with sensitivity are 88.62%, 94.91%, 98.91% and 92.91%, respectively.Compared to morphological methods, the accuracy has improved 17.39% in microaneurysms and 15.18% in haemorrhages. And the detection time is only a quarter of the traditional deep learning methods.
  • ZHANG Chen, DU Gang, DU Xuetao
    Beijing Jiaotong University. 2017, 41(6): 21. https://doi.org/10.11860/j.issn.1673-0291.2017.06.004
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The rapid development of mobile communication technology has greatly promoted the communication experience of users. How to identify and filter out the illegal content in a large amount of data is crucial for improving the ability and level of illegal information management in China Mobile. Towards this end,this paper proposes an ensemble deep model (EDM) to classify illegal images. In this approach, several deep models with diverse network structures and complementary information are integrated by using the proposed scheme, and the illegal images with diverse distributions will be distinguished. To evaluate the effectiveness of the proposed approach, we first collect and set up an illegal image dataset, and compare the proposed approach with the traditional support vector machine(SVM) based image classification method and Alexnet-based, VGG-based and Googlenet-based methods. Experiments show that the proposed approach clearly outperforms the existing methods and obtains excellent classification performance in accurate (94%), precision (84%) and recall (98%).
  • GE Shiming, ZHAO Shengwei, LIU Wenyu, LI Chenyu
    Beijing Jiaotong University. 2017, 41(6): 27. https://doi.org/10.11860/j.issn.1673-0291.2017.06.005
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    Deep learning has been widely used in face recognition system due to its powerful ability in feature representation. However, the high inferring complexity and feature representation reduce the efficiencies in feature extraction and retrieval respectively, which hinders the practical deployments of face recognition system. To address these issues, this paper proposes deep feature distillation in order to uniformly compress the deep network parameters and feature dimensions by distilling the knowledge from large teacher network and domain related data via multi-task deep learning. Combined feature regression and face classification, the method uses a pre-trained large depth network as a teacher network to guide the training of small network, which the knowledge transferred to the lightweight student network to achieve efficient feature extraction. The experimental results on LFW benchmark show that in the condition of the student model recognition accuracy is reduced by 3.7% compared with the teacher model, the network has been compressed to about 2×107 in model size and 128 dimensional feature, which achieves the reductions of 7.1 times in model parameters, 32 times in feature dimension and 95.1% in inferring complexity. The results demonstrate the validity and efficiency of the proposed method.
  • WANG Lei, ZHAO Yinghai, YANG Guoshun, WANG Ruoqi
    Beijing Jiaotong University. 2017, 41(6): 34. https://doi.org/10.11860/j.issn.1673-0291.2017.06.006
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    Combined the big data acquisition, the key technologies of deep neural network have widely applied in the field of image classification, object detection, speech recognition, natural language processing, et al. With the developing of the deep neural network model performance, the model size and the required calculation need to be improved, so that it is reliance on high power computing platform. This paper is focus on the deep neural network model compression technology for embedded applications in order to solve the problems of storage resource, memory access speed constraints and computing resources limit in embedded system. It aims to reduce the model size and the complex computation. Meanwhile, it could optimize the process of calculation. This paper has summarized the state-of-the-art model compression technologies including model pruning, fine model designing, tensor decomposition, model quantization, etc. Through the summary on the model development, it could provide the references for the studies of the deep neural network model compression technologies.
  • YUAN Jian, WANG Peng, WANG Yue, YANG Xin
    Beijing Jiaotong University. 2017, 41(6): 42. https://doi.org/10.11860/j.issn.1673-0291.2017.06.007
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    With the expanding of urban rail transit network revenue length and the increasing of its passenger volume, the passenger volume for some subway stations is susceptible to rapid changes, which can easily incur uneven distribution of the entire network traffic. It could therefore increase the difficulty of rail transit operations and probability of operational incidents. On the basis of passenger volume data collected from practical operation, this paper analyzes the temporal and spatial characteristics of the passenger volume. It also proposes a passenger volume prediction method using the Bayesian network to predict the passenger volume of certain subway stations. Based on practical data, these numerical experiments demonstrate that the proposed method can achieve an mean absolute percentage error below 0.1 when predicting, proving the model is highly accurate.
  • ZHAO Peng, LI Yunfeng ,LI Bo
    Beijing Jiaotong University. 2017, 41(6): 49. https://doi.org/10.11860/j.issn.1673-0291.2017.06.008
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The high-frequency train operations which like the bus operation have become a normal organization mode for meeting the rapid growth travel demand in some certain high-speed railway lines. The historical ticket data show that travelers have obvious differences in choosing the train under the condition of parallel trains between the same origin and destination station (OD). This paper proposes a train choice model for travelers based on the Logit model considering the departure time preferences. Firstly, two aspects of the train selection rate and ticket buying are used to describe the difference between the choices of parallel trains, and indicate that the departure time is a key factor affecting the choice of passengers. A combined survey with RP and SP is applied to get the expected departure time of the passenger and the potential willingness to pay. The departure time preference is integrated into the passenger’s choice of the generalized cost function to construct the passenger’s choice model for the parallel trains. The empirical analysis of the Beijing-Shanghai high-speed railway shows that the average relative error between the model predictive results and the actual situation is 6.62%, which verifies the validity and accuracy of the method. The research can provide theoretical basis and decision-making method for the analysis and management of the travel demand.
  • XU Hong, CHEN Nan
    Beijing Jiaotong University. 2017, 41(6): 55. https://doi.org/10.11860/j.issn.1673-0291.2017.06.009
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    This paper aims at reducing the train delay in urban rail transit, which is caused by disturbance from the random passenger arrivals. The dynamic changeable arrival rate of passengers is set as a random variable which follows a Poisson distribution. Appling stochastic programming theory, this paper proposes an expectation model for train operation which tries to minimize the deviations of the departure time of trains and the deviations of the time intervals between successive arriving trains. The control strategy for adjusting the deviations is obtained by solving the model with the Kuhn-Tucker condition from optimal method.In the base of the real survey data, the simulation chapter obtains the delay of trains by simulating trains operate under the influence of random passenger arrivals and external factors. The paper also compares the deviations of departure time and the levels of optimization before and after using the adjustment strategy. In this way, the paper verifies that the designed strategy can effectively reduce the train delay by 25%, enabling the urban rail transit trains to operate more steadily according to the timetable.
  • HU Zhigang, LI Haodong, HE Shiwei
    Beijing Jiaotong University. 2017, 41(6): 61. https://doi.org/10.11860/j.issn.1673-0291.2017.06.010
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    Daily operation of marshalling yard mainly focuses on optimizing the hump sequence of arrival trains and assemble sequence of departure trains. Based on the wagon-flow allocation plan, this paper presents two sequencing models of hump and assemble, such as the sequencing model with one hump engine and one-yard engine, as well as a general model with multi-engines, which can be improved to take the optimization of technical inspection sequence of trains into consideration. At last, numerical examples are proposed to verify those models. The calculation results of different numbers of shunting engines show that the efficiency of parallel hump process does not always better than single hump process and the hump sequence and assemble sequence are interacted, and the capacities of two parts should be coordinated with each other. The efforts on the efficiency of the whole dispatching system appears insignificant, if only the capacity of the hump side or assemble side is increased.
  • ZHANG Xiaodong, DONG Baotian, WANG Yue
    Beijing Jiaotong University. 2017, 41(6): 69. https://doi.org/10.11860/j.issn.1673-0291.2017.06.011
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The contemporary railway marshalling station wagon forecast method cannot meet the requirements compared with the actual railway transport organization. Combined with the characteristics of the stability of the wagon operation time data, the goods-wagon type mapping function is designed. Considering the impact of the difference between the wagon type and the goods type, the new time series is dynamic updated. A new forecasting model of marshalling yard wagon transit operation is put forward with time series analysis method. The parameters in the model are estimated using the maximum likelihood estimate method. The experimental results indicate that the proposed model is conducive to forecast and analyze the transportation organization activities, could accurately forecast the wagon transit time, conform to the actual development trend and improve the accuracy of railway traffic flow forecast.
  • WANG Yongxing, BI Jun, SUN Huanhuan
    Beijing Jiaotong University. 2017, 41(6): 76. https://doi.org/10.11860/j.issn.1673-0291.2017.06.012
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    In car rental peak hours, difficulty to reasonably assign the vehicles to users is a main problem for the operators of time-share car rental. Focused on the reservation assignment problem under time-share car rental mode, a reservation assignment optimization model is proposed based on the analysis of the main affecting factors. The model aims at maximizing the profits of the time-share car rental operator. A genetic algorithm (GA) is designed to solve the model. The model and GA are applied to the simulation example based on the assumption of simulation scene. The results show that the GA is suitable for solving the proposed model. The maximum profits of the operator and the corresponding optimal scheme of reservation assignment can be obtained through solving the model. Through comparison analysis, it can be found that the profits of the operator under the model are superior to the conventional reservation assignment method. The feasibility and practicality of the model and GA are verified.
  • WANG Wenjing, BAI Jinyi, LIU Wei
    Beijing Jiaotong University. 2017, 41(6): 82. https://doi.org/10.11860/j.issn.1673-0291.2017.06.013
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    In view of the practical difficulies facing the traditional nominal stress approach applied in fatigue analysis of weld plate structure, the hot spot stress method based on the surface extrapolation is introduced comprehensively. It has been developed in marine structural engineering well. The specimens are loaded under different stress levels and the fatigue life with 97.7% reliability is obtained. The hot spot stress of the specimen weld toe is calculated by the spot stress method and the fatigue life is given by FAT90 modified S-N curve considering the plate thickness effect. The results show good accuracy and better reliability in the assessment of weld structure fatigue life. Taking the railway antenna beam as example, the hot spot stress method is used to get the maximum stress point, which is consistent with the fatigue crack position in service. The fatigue life prediction of the antenna is 7.1×105 and it coincides with the actual operation life basically. The analysis method can be applied on the anti-fatigue design of weld joints of plate structure.
  • SONG Zhikun, YUE Renfa, HU Xiaoyi, LI Xiang
    Beijing Jiaotong University. 2017, 41(6): 88. https://doi.org/10.11860/j.issn.1673-0291.2017.06.014
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    For the polygonal wheel phenomenon becoming increasingly prominent, a rigid-flexible coupling vehicle system dynamics model is established in ANSYS and SIMPACK considering the flexible wheelset. The influence of polygonal wheels on the vehicle vibration and the wheel rail contact force is studied. The results show that the fluctuation of the wheel/rail vertical force increases with the increase of the polygon wave depth, but does not increase with the increase of the polygon. The bending vibration frequency of the wheel pair can be coupled with the nod frequency of the wheel set and the rotating arm. If the vibration frequency generated by the polygon of the wheel is in the range of the frequency, the resonance will occur. According to the wheel/rail force upper limit value 170 kN, the 1~20 polygon wave depth limit value at the speed of 300 km/h is proposed, especially the wave depth of the 11 wheel polygon is smaller than 0.07 mm.
  • GOU Qingwei, ZHANG Xinmin, ZHAO Chang, ZHANG Xin
    Beijing Jiaotong University. 2017, 41(6): 94. https://doi.org/10.11860/j.issn.1673-0291.2017.06.015
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    In order to analyze the influence of structure parameters on high pressure common rail injector response characteristics, this paper takes a particular heavy-duty diesel engine high pressure common rail injector as the research object, builds the simulation model by using the AMESim software and verifies the accuracy of the model by experiments. The results show that the increase of the inlet hole diameter will enhance the open response speed of the fuel injector but lower its close response speed. With the acceleration of the fuel injector open response speed, the increase of the control plunger diameter will enhance the close response speed of the fuel injector but lower its open response speed; the increase of needle valve sealing diameter will improve the opening and close response speed at the same time, the increase of the needle valve spring pre-tightening force will prolong the needle valve turn-on time and lower its turn-off time. Therefore, in the process of the injector design, the inlet hole diameter should be chosen carefully, and the outlet hole and needle valve sealing diameter should be increased appropriately. It can be applied as theoretical guidance for the optimization of structural parameters of high pressure common rail injector.
  • ZHANG Jibao, ZHANG Xin, QIN Zhihan, HOU Xiaosen
    Beijing Jiaotong University. 2017, 41(6): 101. https://doi.org/10.11860/j.issn.1673-0291.2017.06.016
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The integration and miniaturization of diesel engine ECU as well as poor working conditions are likely to result in the failure of ECU’s components. In order to study the thermal load of ECU, thermal test and analysis of diesel engine ECU under different operation modes and environment temperatures are performed through the ECU thermal test system. It’s found out that the rotational speed and the torque of diesel engine will affect the power loss of ECU’s components. About 75% of components’ temperature reaches maximum when diesel engine operates at rated power, but the temperature of components in injector solenoid valve drive module reaches maximum under maximum torque condition and are significantly affected by the operation conditions of diesel engine.The environment temperature affects heat transfer so that the temperature difference between ECU’s components and environment decreases first and then increases with the increase of environment temperature. As environment temperature up to 105 ℃, temperature of ECU components is closed to 125 ℃ which is permissible temperature and the improvement of cooling condition for components is required.
  • LIU Beisheng
    Beijing Jiaotong University. 2017, 41(6): 107. https://doi.org/10.11860/j.issn.1673-0291.2017.06.017
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The use of BIM for fine engineering project management requires online display and interaction of 3D BIM models. Although traditional 3D graphics visualization technology can display models, there are still some problems such as poor compatibility of the model file format, high hardware requirements of devices and lack of necessary interactive functions. In order to solve these problems a cloud rendering based 3D BIM model visualization solution is proposed. The solution includes BIM model lightweight conversion technology, HOOPS based graphics engine development technology, cloud rendering based large model display and interactive technology. Through the research and development of the system, BIM models designed using Revit, Catia, Microstation, Tekla etc. are converted into lightweighted formats, and a cloud rendering based graphics engine system is developed and verified in Beijing to Zhangjiakou railway project.
  • WANG Kunpeng, XIA He, ZHANG Nan
    Beijing Jiaotong University. 2017, 41(6): 114. https://doi.org/10.11860/j.issn.1673-0291.2017.06.018
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The creep of simply-supported concrete beam on high-speed railway will induce additional deformation of the track structure and affect train operation quality. A train-bridge system which takes car-body flexibility into consideration is therefore established. Taking the 32 m and 24 m simply-supported prestressed concrete beams, which are the common span beams, on high-speed railway as a study case, the additional deformation of CRTS III ballastless slab track induced by concrete creep and its effect on the safety and comfort of vehicles running through the deformed bridge are studied. The calculation results show that prolonging the track construction time can effectively prevent additional deformation of track structure induced by creep from exceeding code limit. The creep deflection makes the convex angles appear around beam end of the track structure, when a wheel goes through the convex angle, the wheel-rail force will decrease sharply, which could even induce wheel-rail separation. The influence of creep on the comfort index is far greater than that on the safety index, and the influence of creep on the vehicle operation comfort could be underestimated when ignoring car-body flexibility.
  • LIU Meilin, HOU Yanjuan, ZHANG Dingli, FANG Qian
    Beijing Jiaotong University. 2017, 41(6): 122. https://doi.org/10.11860/j.issn.1673-0291.2017.06.019
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    Based on the research data collected at Xiawafang station, which is a cross-transfer subway station of Tianjin subway line 5 and line 1, the horizontal displacement of foundation pit support structure, the surface settlement behind the retaining wall, and the deformation of existing station caused by deep foundation pit crossing the existing subway station are analyzed. The results show that the maximum horizontal displacement of support wall is 0.064%H (in which H is the excavation depth), where is about 0.63H under the ground surface. The maximum ground surface settlement behind the retaining wall is 0.025%H, where is about 0.71H away from the wall. The influence zone extends to about 2H from the wall. The ratio between the maximum ground surface settlement and the maximum wall deflection is among 0.38 to 1.04, and its average value is about 0.77. The vertical deformation of the existing station which crosses the foundation pit is mostly uplifted. The horizontal displacement protrudes outside the railway station. The deformation of the existing station is mainly horizontal. By reinforcing the existing station with waterproofing sealing and reinforcing its soft soil layer behind the retaining wall, the existing structural deformation can be significantly reduced.
  • LIU Qibin
    Beijing Jiaotong University. 2017, 41(6): 130. https://doi.org/10.11860/j.issn.1673-0291.2017.06.020
    Abstract ( ) Download PDF ( ) HTML   Knowledge map   Save
    The static and dynamic analysis models of No.12 alloy steel single turnout with combined frog are established based on finite element analysis software. The fatigue sensitive areas of turnout rail are located by the dynamic simulation results of the dynamic model when the train is passing through the turnout in a straight or branch line, the load spectrums of each sensitive area are extracted for fatigue analysis. The internal stress of rails in each sensitive area under unit load are researched by the static model. The fatigue crack initiation lives of rails in sensitive areas are predicted based on Miner linear cumulative rule. The effects of axle-weight and vehicle speed on the fatigue crack initiation lives of rails are studied. A prediction formula of fatigue crack initiation lives of rails in sensitive areas under different axle loads and vehicle speeds is given. The results show that the fatigue crack initiation lives of rails in different sensitive areas of turnout varies greatly, and axle-weight and vehicle speed when the train is passing through the turnout in the branch line have great effects on the fatigue crack initiation lives of rails. The fatigue crack initiation life decreases significantly with the increase of axle-weight, and it also decreases with the vehicle speed when the train is passing through the turnout in the branch line.