Table of Content

13 February 2020, Volume 0 Issue 01

Previous Issue    Next Issue

LSTM-based Working State Prediction of Industrial Internet Equipment

LI Zhao-tong, ZHANG Wei-shan, GUO Wu-wu

2020, 0(01):  1.  doi:10.3969/j.issn.1006-2475.2020.01.001

Asbtract ( 189 )   PDF (746KB) ( 174 )  

References | Related Articles | Metrics

With the development of industrial Internet technology, working state prediction of industrial Internet equipment is of great significance for improving the reliability of equipment. In practical industrial scenarios, simple single-signal prediction and threshold methods are ineffective because the data is highly discrete and coincides over multiple time periods. This paper presents a working state prediction model of industrial Internet equipment based on LSTM (Long Short-Term Memory) neural network. Firstly, this paper uses the SMOTE algorithm for data skew processing and the PCA algorithm for data dimensionality reduction, then builds the working state prediction model of industrial Internet equipment based on LSTM neural network. Finally the model is evaluated by F1-Score. This paper is based on real air conditioning compressor data for experimental verification. The experimental results show the effectiveness of the proposed method.

Hydraulic Turbine Governing System Based on Neural Network Inverse Control

CHEN Yan-lin, LI Zhi-hua， XIE Xue-han

2020, 0(01):  6.  doi: 10.3969/j.issn.1006-2475.2020.01.002

Asbtract ( 107 )   PDF (1432KB) ( 142 )  

References | Related Articles | Metrics

According to the identification ability of neural network to the model of non-linear system, the inverse model of hydraulic turbine generator unit is modeled by combining the neural network with the adaptive inverse control, and a new turbine regulating system is born. Based on the theory of inverse system and system identification, a neural network inverse controller is established for the frequency and load disturbance of the hydraulic turbine generator unit, and the simulation results are compared with the traditional PID control. From the simulation results, it can be seen that the proposed control scheme can effectively control the hydraulic turbine generator unit and make the system have better dynamic performance and robustness.

Reviews on Event Knowledge Graph Construction Techniques and Application

XIANG Wei

2020, 0(01):  10.  doi:10.3969/j.issn.1006-2475.2020.01.003

Asbtract ( 1782 )   PDF (912KB) ( 1480 )  

References | Related Articles | Metrics

Given its rich and flexible semantics by the graph structure, knowledge graph which describes the things in the objective world and their relationships has received extensive attentions in many fields. In objective world, event knowledge graph focuses on various dynamic events, entities and their relationships in terms of structured graph for more efficient management of massive data. In particular, the mining of dynamic event information and event logic in the application field are of great significance for understanding the laws of world development and assisting various intelligent applications. The construction techniques and typical applications of event knowledge graph are reviewed in this article, including event knowledge representation, event knowledge extraction, and event relation extraction. The challenges and research perspectives are also discussed.

Research and Implementation of Real-time Analysis Technology of Moving Target Data

HOU Bo, NIE Ying

2020, 0(01):  17.  doi:10.3969/j.issn.1006-2475.2020.01.004

Asbtract ( 268 )   PDF (1102KB) ( 545 )  

References | Related Articles | Metrics

Aiming at the real-time trajectory data of moving object， this paper analyzes the problems consisting in the existing research, and proposes two solutions for real-time analysis. The first one is trajectory prediction method based on five-point method, which can predict the next point’s position of moving object rapidly and has strong real-time performance. The second one is Storm-based historical frequency statistical analysis method, which analyzes historical track frequency with high accuracy. These two methods solve two important problems in real-time analysis: real-time, accurate, and have high practicability.


Image Segmentation Method Based on Optimization of PSO Algorithm #br# and K-means Clustering Algorithm

CAO Shuai-shuai， CHEN Xue-xin， MIAO Pu， BU Qing-kai

2020, 0(01):  22.  doi:10.3969/j.issn.1006-2475.2020.01.005

Asbtract ( 247 )   PDF (1980KB) ( 194 )  

References | Related Articles | Metrics

In order to improve the quality and efficiency of image segmentation, and considering the weakness that particle swarm optimization (PSO) algorithm is easy to fall into local optimum and that K-means algorithm is sensitive to initial clustering center, combining PSO with K-means algorithm, an optimization algorithm is proposed through the adjustment of inertia weight and learning factor. First, the image is denoised and pre-processed, and the processed color image is converted to the HSV space to improve the color quality. Then, the formula and parameters of the inertia weight and learning factor in the particle swarm optimization algorithm are improved to avoid falling into local optimum. Finally, according to the fitness of the particles, the K-means algorithm is switched to perform a local search, so that the cluster center is continuously updated to achieve fast convergence. In the process of image segmentation, the experimental results show that this improved algorithm has strong ability in global search and it performs well in faster convergence speed and higher segmentation accuracy.

A Method Based on Markov Model for Accelerating Path #br# Convergence in Information-Centric Networking

MA Pu-fang1,2, WANG Jin-lin1,2, YOU Jia-li1,2

2020, 0(01):  28.  doi:10.3969/j.issn.1006-2475.2020.01.006

Asbtract ( 169 )   PDF (881KB) ( 119 )  

References | Related Articles | Metrics

Multi-homed host can own multiple addresses because of the separation of the identifier and address of the host in Information-Centric Networking (ICN). The packet with multiple destination addresses can obtain multiple output ports at each hop after matching the routing table. Thus, the router can dynamically select the path of the multi-address packet at each hop to improve the throughput of the transmission. However, this forwarding method breaks the shortest path forwarding rule according to the routing table, and the multi-address packet may hop back and forth in the network and cannot quickly converge to the destination. This paper proposes an address trimming method based on Markov model, which trims the addresses according to historical information of address trimming. The experimental results show that the method can improve the path convergence with reducing the average hop by about 16% compared with the benchmark method, while the transmission rate keeps almost the same.

Design and Optimization of Content Diffusion Strategies in Heterogeneous Edge Network

XUE Han-xing1，2, YOU Jia-li1，2, WANG Jin-lin1，2

2020, 0(01):  34.  doi:10.3969/j.issn.1006-2475.2020.01.007

Asbtract ( 160 )   PDF (1055KB) ( 98 )  

References | Related Articles | Metrics

Deploying the replications of popular content in the edge network through the content diffusion system can effectively alleviate the load on the backbone network. Because the capacities of terminals and the interests of their users may differ, the content diffusion strategies need to fully consider the heterogeneous. This paper proposes a layered topology of network and a decentralized content diffusion strategy to deploy the replications. Based on the capabilities, load and interests of terminals, this paper further optimizes the diffusion of content. The optimization strategy includes the adjustment of target of diffusion based on the interest features, the heuristic caching node selection strategy, the coverage rate-aware selection of transmission content and transmission load-aware selection of receiving node. Experimental results show that the optimized diffusion system can provide a more stable success rate of services, improve the probability of local services by about 25.5%, and reduce the time to complete the diffusion of content by about 7.6%.

A Topology Construction Method for Incompletely Measurable Networks

LIAO Yi1,2, SHENG Yi-qiang1,2, WANG Jin-lin1,2

2020, 0(01):  41.  doi:10.3969/j.issn.1006-2475.2020.01.008

Asbtract ( 183 )   PDF (1526KB) ( 84 )  

References | Related Articles | Metrics

Overlay network technology is popular in the research fields of next-generation Internet, the data center network of cloud, and software-defined network. The measurement-based overlay network construction technology can utilize the real-time network state data by active measurement techniques, which is benefit for the overlay network to adapt to the dynamics of network. However, the measurement-based overlay network construction methods also face the incompletely measurable problem, i.e., the global network status information required to be measured when the node joining cannot be completely measurable or it is difficult to get enough node information in a limited time, resulting in the failure of the node joining process. To solve this problem, this paper proposes a Topology Construction method for Incompletely Measurable (TCIM) networks. TCIM includes a high-precision node joining method and a low-complexity node joining method to construct a tree topology based on latency. In TCIM, the high-precision node joining algorithm uses the edge relationships of the latency triangle to select a suitable parent node for the node to join under small-scale or static/low-dynamic conditions; the low-complexity node joining method select constant number of nodes which have joined in the overlay network to measure and select the node with the smallest latency as the parent node, which can be used for large-scale, high-dynamic and network incomplete measurable conditions. The simulation results show that the tree-based overlay structure generated by TCIM has lower latency stretch comparing with state-of-art methods underdifferent network topology models, and TCIM has smaller topology maintenance cost in the Waxman and the BA graph model.

Design of Power Operation and Maintenance Audit System Based on Hadoop

SU Lin-ping, AN Ran, LI Wei, CUI Wen-chao, ZHANG Xiao-liang

2020, 0(01):  49.  doi:10.3969/j.issn.1006-2475.2020.01.009

Asbtract ( 210 )   PDF (752KB) ( 120 )  

References | Related Articles | Metrics

State Grid is becoming more and more informative. The amount of data generated by single-machine operation and maintenance audit system is increasing. The performance of high-efficiency storage analysis for massive data is seriously degraded, and system stability is reduced. In order to meet the requirement of State Grid for data storage analysis and system stability of the operation and maintenance audit system, based on the Hadoop open source architecture, the paper proposes a massive data distributed storage method based on Hadoop cluster and heartbeat detection technology based on Heartbeat, and designs a power operation and maintenance audit system based on Hadoop. The experimental results show that the availability of power operation and maintenance audit system is increased by 8.42% compared to stand-machine system. The performance of massive data for storage analysis is greatly improved. It has the advantages of stable system operation and uninterrupted service.

A Cloud Application Decomposition Method Combining GWO with FM Algorithm

JIANG Kai-hua1,2, SUN Peng1,2, HAN Rui1

2020, 0(01):  53.  doi:10.3969/j.issn.1006-2475.2020.01.010

Asbtract ( 154 )   PDF (746KB) ( 84 )  

References | Related Articles | Metrics

The rapid development of the Internet brings challenges to the data process mode in cloud service. In this regard, the Chinese Academy of Sciences proposes the SEA service and SEA-Cloud collaboration system, in which the decomposition strategy of cloud application is an important link to affect the performance of the system. However, current mainstream methods aim to deal with the simple graph in the cloud environments, which mismatch the directed weighted graph in the SEA-Cloud collaboration scenarios. So, this paper proposes a cloud application decomposition method combining Grey Wolf Optimizer(GWO) and〖JP2〗 Fiduccia-Mattheyses(FM) algorithm to solve the problem. Benefitted from the rapid convergence of GWO, the proposed algorithm takes the outcomes of GWO as the initial inputs of FM algorithm to avoid the sensitivity of FM to the initial partitions. The simulation results illustrate that the combination method outperforms the current method. The obtained partition matches the distribution of SEA resources and the rate of edge cut decreases dramatically, which means that the communication overhead is reduced.

Intelligent Attendance System Based on Dual Positioning Technology

LIANG Xiao-qi1, DAI Yong-hui2, ZANG Hong-yan1

2020, 0(01):  58.  doi:10.3969/j.issn.1006-2475.2020.01.011

Asbtract ( 178 )   PDF (1195KB) ( 111 )  

References | Related Articles | Metrics

In view of the current situation that the class time of adult education students is not centralized, the classroom is not fixed, and WiFi attendance effect is poor, an intelligent attendance system for adult education students is proposed, which uses outdoor GPS and indoor WiFi dual positioning technology to solve the above problems. This paper describes the design and implementation of the system, including the system architecture design, function design, attendance check-in process design, as well as the implementation of GPS and WiFi positioning attendance, course reminder, attendance statistics and other functional modules. The test and actual operation show that the system runs stably, realizes the students’ intelligent attendance and statistical functions, which not only improves the accuracy of attendance data, but also effectively prevents the phenomenon of “Delegate attendance, false attendance”, which provides a new method for improving the efficiency of teaching management.

An Automatic Text Summarization Model Construction Method Based on BERT Embedding

YUE Yi-feng, HUANG Wei, REN Xiang-hui

2020, 0(01):  63.  doi:10.3969/j.issn.1006-2475.2020.01.012

Asbtract ( 339 )   PDF (863KB) ( 267 )  

References | Related Articles | Metrics

Aiming at the problem that the traditional word vector can not effectively represent polysemous words in text summarization, which reduces the accuracy and readability of summarization, this paper proposes an automatic text summarization model construction method based on BERT (Bidirectional Encoder Representations from Transformers)Embedding. This method introduces the BERT pre-training language model to enhance the semantic representation of word vector. The generated word vectors are input into the Seq2Seq model for training to form an automatic text summarization model, which realizes the rapid generation of text summarization. The experimental results show that the model can effectively improve the accuracy and readability of the generated summarization on Gigaword dataset, and can be used for automatic text summarization generation tasks.

A Collaborative Filtering Recommendation Algorithm Based on Adjusted Cosine Similarity

LI Yi-ye 1,2, DENG Hao-jiang1,2

2020, 0(01):  69.  doi:10.3969/j.issn.1006-2475.2020.01.013

Asbtract ( 284 )   PDF (800KB) ( 213 )  

References | Related Articles | Metrics

In the traditional collaborative filtering recommendation algorithm, there exists a problem of low accuracy of classification in the case of sparse data. To solve this problem, a collaborative filtering recommendation algorithm based on adjusted cosine similarity is proposed. Converting the data to the feature matrix by the embedded layer, the mean square error between the adjusted cosine similarity matrix obtained by the calculation and the unit matrix is used as the loss function for improving the accuracy of classification of the proposed algorithm in the case of sparse data. Experimental results indicate that, compared to the FM, FFM and DeepFM models, AUC and the Logloss of the proposed collaborative filtering recommendation algorithm based on adjusted cosine similarity are better within the acceptable range of training time.

Collaborative Filtering Recommendation Based on Dynamic Changes of User Preferences

JIANG Shu-hao1,2, ZHANG Li-yi1,2, ZHOU Na1

2020, 0(01):  75.  doi:10.3969/j.issn.1006-2475.2020.01.014

Asbtract ( 173 )   PDF (855KB) ( 140 )  

References | Related Articles | Metrics

Traditional collaborative filtering methods focus only on rating data to generate recommendation， without considering the evaluation time， project category and other informations, which affects the quality of the recommended system. This paper proposes a personalized recommendation model based on the dynamic changes of user preferences. The method is based on the project category, and different weighting functions are set up according to the user scoring time (recent, long and periodic). The experimental results from Movielens data set show that this method weakens the influence of short-term preference on recommendationquality, reflects the dynamic changes of user preferences accurately, and improves the accuracy of recommendation effectively.

FFMPEG Online Video Conversion System Based on Autonomous Controllable Platform

Abudukelimu Yusufu1,2, WANG Liang-liang1

2020, 0(01):  81.  doi:10.3969/j.issn.1006-2475.2020.01.015

Asbtract ( 329 )   PDF (787KB) ( 120 )  

References | Related Articles | Metrics

As a part of the research on technology of autonomous controllable platform, an on-line FFMPEG video conversion system based on autonomous controllable platform is researched and implemented, which provides online video format conversion service and video encoding processing environment for embedded platform. This paper mainly designs the system framework, builds the system operating environment, and researches the implementation method of the related core functional modules of the system. The core modules include video upload module, Socket service module, video conversion, compression, key frame injection module, and video play module. Finally, an experiment is done to test the function and performance of online video conversion system. Experiments show that the system can support batch upload, conversion, and compression of large-capacity video files, and can support the conversion of different video file formats, and the client player supports arbitrary dragging of the video progress bar and playing.

GL Studio Chinese Technology Based on GB2312 Standard

WANG Zhi-le1, DONG Jun-yu2

2020, 0(01):  85.  doi:10.3969/j.issn.1006-2475.2020.01.016

Asbtract ( 162 )   PDF (930KB) ( 117 )  

References | Related Articles | Metrics

Aiming at the problem that the instrument display software developed by GL Studio has not supported Chinese character display, the texture technology is used to establish the first-level Chinese character texture font of GB2312 standard, and the font library corresponding to the texture font library and the basic Chinese character display unit are established in GL Studio environment. The storage of abstract Chinese characters shows the data model. Taking the airborne MFD display software as an example, according to the Chinese characters of the input MFD, the corresponding national standard code and the coordinates of the Chinese characters in the font table are searched, and the Chinese character dynamic control model and the Chinese character display model are established based on the basic Chinese character display unit. The development of airborne MFD simulation software is realized in the GL Studio environment. The experiment shows that the method is fast, effective, scalable and reconfigurable, and can be used not only in the development of real equipment, but also in the development of simulation simulator.

Automation Software Deployment Algorithm for Military Information System

DAI Wen-bo, XU Luo, WEI Jin-yi

2020, 0(01):  90.  doi:10.3969/j.issn.1006-2475.2020.01.017

Asbtract ( 165 )   PDF (796KB) ( 100 )  

References | Related Articles | Metrics

In the large-scale information system for the military field, because the existing automatic deployment software such as Jenkins, apt, Docker, and other software can not meet the large varieties of large information system software, complex software dependencies, cross-platform deployment requirements, this paper first proposes a set of standardized software deployment models which is described by Document Architecture Description (hereinafter referred to as XSD). Secondly, based on the dependency conflict detection algorithm, the software deployment sequence generation algorithm is proposed by improving the depth-first traversal algorithm (DSP). And the algorithm is verified through experiment.


A Symbolic Execution Optimization Method for Safety-critical Software System

DAI Yan-jun1, WU Zhi-qiang2, LIU Jie1, LIU Zhao-hui1, CHEN Zhi2, XIAO An-hong2

2020, 0(01):  96.  doi:10.3969/j.issn.1006-2475.2020.01.018

Asbtract ( 203 )   PDF (622KB) ( 132 )  

References | Related Articles | Metrics

In the aviation, nuclear power and defense military fields, the software of the Safety-Critical System (SCS) is very important, and its reliability must be guaranteed by testing or formal methods. Symbolic execution is widely used as an efficient test case generation method. However, the coupling between the modules of SCS software system is high, which makes symbolic execution constraint solving difficult. This paper proposes a decoupling method with a minimum set of weights to provide a new idea for the automated testing of safety-critical software systems.

A Conditional Proxy Re-encryption Supporting Multi-keyword Search

QIN Lu-lu, ZHOU Li-jing, WANG Min

2020, 0(01):  100.  doi:10.3969/j.issn.1006-2475.2020.01.019

Asbtract ( 180 )   PDF (550KB) ( 112 )  

References | Related Articles | Metrics

In order to realize fine granularity access control and multiple keyword data search, this paper introduces access tree and strong one-time signature, and proposes a conditional proxy re-encryption scheme which can be used for multiple keyword search. Compared with the existing schemes, this scheme has obvious improvement in computational efficiency. The scheme can resist the selective cipher-text attack and satisfy the security of the chosen cipher-text attack under the random prophet model. The scheme proposed in this paper plays an important role in solving the security problem of sensitive data in industries such as finance.

An Improved Grasshopper Optimization Algorithm Based on Levy Flight

ZHAO Ran1,2, GUO Zhi-chuan1,2, ZHU Xiao-yong1,2

2020, 0(01):  104.  doi:10.3969/j.issn.1006-2475.2020.01.020

Asbtract ( 274 )   PDF (737KB) ( 289 )  

References | Related Articles | Metrics

Grasshopper optimization algorithm is a meta-heuristic optimization algorithm that can be used to solve task scheduling problems. The existing improved grasshopper optimization algorithm lacks randomness, and its ability to jump out of local optimum is weak. The improvement effect is not obvious enough. To solve this problem, this paper proposes an improved Grasshopper Optimization Algorithm Based on Levy flight（LBGOA）. The algorithm introduces a local search mechanism based on Levy flight to enhance the randomness of the algorithm and adopts a random jumping strategy based on linear decreasing parameters to improve the ability of the algorithm to jump out of local optimum. The experimental results of the CEC test show that the proposed algorithm has strong search ability, and 17 optimal solutions and 6 suboptimal solutions are obtained by LBGOA in the results of 30 test functions. The proposed improved algorithm is applied to the task scheduling problem in edge computing. The results of task scheduling simulation experiments show that the proposed algorithm can effectively improve the search results. Compared with GOA, OBLGOA, WOA, ALO, DA and PSO algorithms, the search results by LBGOA are promoted by 7.4%, 7.5%, 4.8%, 27.7%, 29.9%, and 20.7%respectively.

A Method of Sample Updating and Target Repositioning Based on KCF

WU Shi-yu， LI Zhi-hua， WANG Wei

2020, 0(01):  111.  doi:10.3969/j.issn.1006-2475.2020.01.021

Asbtract ( 171 )   PDF (2601KB) ( 106 )  

References | Related Articles | Metrics

In order to solve the problem that the Kernelized Correlation Filter (KCF) algorithm leads to target tracking failure due to the accumulation of measurement error, a sample quality evaluation mechanism is proposed to screen the sample to update the classifier. In order to solve the problem of repositioning after target occlusion, the Kalman filtering algorithm is used to estimate the target position, and then the estimation results are evaluated.In order to solve the problem that the target location is difficult to predict, the ORB feature point matching algorithm is used to complete the relocation of the target. A partial sequence in the TB dataset is selected for testing. Experimental results show that when the target appears in short-term and long-term occlusion, the improved algorithm improves the accuracy and success rate to a certain extent.

A Multi-scale Feature Fusion Meter Box Rust Spot Detection Algorithm Using Cascaded RPN

WANG Wen, ZHOU Chen-yi, XU Yi-bai, LU Shan, ZHOU Meng-lan

2020, 0(01):  117.  doi:10.3969/j.issn.1006-2475.2020.01.022

Asbtract ( 273 )   PDF (1391KB) ( 113 )  

References | Related Articles | Metrics

The consequences of corrosion of the power distribution cabinet may result in poor contact, which may even lead to fire and explosion of some electrical control equipment. To this end, this paper proposes a neural network based multi-scale meter rust spot detection method. Firstly, based on a large number of rust spot data, a convolutional neural network (CNN) model for identifying rust spots is trained. Secondly, the trained CNN model is used to detect the position of the meter box, and real-time recognition of the rust spot on the surface of the meter is realized. The algorithm combines the multi-scale feature mapping through the cascaded RPN network, and makes full use of the location information of the low-level features and the strong semantic information of the high-level features to enhance the detection effect. For the collected rust spot dataset of the meter, the meter detection reaches 94.9% precision, which is better than 91.1% precision achieved by YOLOv2, and the rust spot classification precision reaches 94.5%. The recognition rate, robustness, real-time and stability of rust spot recognition can better meet the needs of practical applications.

Speaker Recognition Based on DNN and Pitch Period

ZHANG Xue-xiang, LEI Ju-yang

2020, 0(01):  122.  doi:10.3969/j.issn.1006-2475.2020.01.023

Asbtract ( 194 )   PDF (851KB) ( 114 )  

References | Related Articles | Metrics

 Traditional speaker recognition frameworks are mostly based on the Gauss mixture model (GMM), but this shallow learning model can not effectively represent the high-order correlation between data features, thus the recognition effect is poor. In this paper, a speaker recognition method based on Deep Neural Network (DNN) and Pitch Period (PP) is proposed. The logarithmic Meier filter bank feature parameters are used as the input of DNN for mainline identification, and the voiceprint characteristics of the speaker are extracted through training DNN model. To eliminate the subjective influence of threshold setting in DNN model, dynamic time warping technology is used to match pitch period of the speaker for assistant recognition. The experimental results show that equal error rate (EER) of this dual recognition method reaches 1.6%, which decreases respectively by 1.2% and 2.4% compared with DNN system and EM-GMM system, and this method still has good robustness in noise environment.