Published in last 1 year | In last 2 years | In last 3 years | All
 Select A survey of image captioning methods based on deep learning Zhifu CHANG,Fengyu ZHOU,Yugang WANG,Dongdong SHEN,Yang ZHAO Journal of Shandong University(Engineering Science)    2019, 49 (6): 25-35.   DOI: 10.6040/j.issn.1672-3961.0.2019.244 Abstract （2039）   HTML （57）    PDF（pc） （7881KB）（737）       Save Image captioning is the cross-research direction of computer vision and natural language processing. This paper aimsed to summarize the deep learning methods in the field of image captioning. Imgage captioning methods based on deep learning was summarized into five categories: multimodal space based method, multi-region based method, enconder-deconder based method, reinforcement learning based method, and generative adversarial networks based method.The datasets and evaluation metrics were demonstrated, and experimental result of different methods were compared. The three key problems and future research direction for image captioning were presented and summarized.
 Select Fast 4-points congruent sets for coarse registration of 3D point cloud Shiguang LIU,Hairong WANG,Jin LIU Journal of Shandong University(Engineering Science)    2019, 49 (2): 1-7.   DOI: 10.6040/j.issn.1672-3961.0.2018.244 Abstract （1992）   HTML （145）    PDF（pc） （6934KB）（644）       Save In order to solve the problem that the 4-points congruent sets (4 PCS) method suffered from low computational efficiency and high registration errors when the overlap rate of two pieces of input point clouds was low, fast 4-points congruent sets (F-4PCS) was put forward. A new method for selecting four-point basis was presented. The source point cloud and target point cloud were given, their boundaries were separately extracted and extended as the boundary feature bands, and then a consistent four-point basis set was chosen from the boundary feature bands. This method could avoid some unnecessary iterations. By limiting the characteristics of the four-point basis, the invalid four-point basis was removed, it could reduce the verification time of the algorithm and improve the computational efficiency. Experiments results carried out on the relevant data sets showed that the F-4PCS method was more efficient than conventional 4PCS method in the case of low overlap rate of input point clouds and the registration success rate was higher than state-of-the-arts.
 Select Object detection of 3D point clouds based on F-PointNet Peng WAN Journal of Shandong University(Engineering Science)    2019, 49 (5): 98-104.   DOI: 10.6040/j.issn.1672-3961.0.2018.348 Abstract （1964）   HTML （40）    PDF（pc） （3128KB）（733）       Save Aiming at the problem of poor detection accuracy of the current 3D point cloud object detection model, the F-PointNet model, which directly processed point cloud data, was used to detect cars, pedestrians and cyclists, and the model was fine-tuned to further improve the object detection accuracy. The model was tested by different parameter initialization methods, $\ell$2 regularization and modifying convolution kernels. The experimental results showed that the Xavier parameter initialization method converged faster 0.09s than the truncated normal distribution method, and the vehicle detection accuracy and the cyclists detection accuracy was about 3% and 2% higher respectively. By adding $\ell$2 regularization, the detection accuracy of pedestrians and cyclists was increased by about 2% and 1% respectively. By reducing the number of convolution kernels in the first layer of T-Net (Transformer Networks) to 128, the detection accuracy of cars and cyclists was increased by about 1% and 2% respectively, which confirmed that the model could effectively improve object detection accuracy.
 Select Chinese short text classification method based on word2vec embedding Mingxia GAO,Jingwei LI Journal of Shandong University(Engineering Science)    2019, 49 (2): 34-41.   DOI: 10.6040/j.issn.1672-3961.0.2018.197 Abstract （1831）   HTML （46）    PDF（pc） （2868KB）（257）       Save In the short text classification process, the weak feature expression of the limitation of the number of words restricted the classification effect. To solve this problem, a Chinese short text classification method based on embedding trained by word2vec from Wikipedia (CSTC-EWW) was proposed, and a series of experiments for short texts with 4 topics from the iask.com website were finished. This method firstly trained the embedding by word2vec from Wikipedia corpus. the feature of short text based on the embedding was established. Naive Bayes and SVM was used to classify short text. The experimental results showed the following conclusions: CSTC-EWW could effectively classify short texts and the best F-value could reach 81.8%; Comparing the text feature expression of BOW model weighted by TF-IDF and the method of extending feature from Wikipedia, the classification results of CSTC-EWW were significantly better and F-measure of CSTC-EWW on car could be increased by 45.2%.
 Select Review of capsule network Jucheng YANG,Shujie HAN,Lei MAO,Xiangzi DAI,Yarui CHEN Journal of Shandong University(Engineering Science)    2019, 49 (6): 1-10.   DOI: 10.6040/j.issn.1672-3961.0.2019.312 Abstract （1438）   HTML （103）    PDF（pc） （4331KB）（877）       Save Recently capsule network with dynamic routing was the new neural network model which was considered a significant model in next generation. In recent years, much research evidenced capsule network exceptional ability to fit features. But the high computational overhead made it unable to fit complex and large datasets. Consequently, reducing computational became a research hotspot. There were two methods, including optimized capsule and optimized routing, to solve the issue. Optimized capsule was usually driven by application purpose which was designed as a model of specific classification tasks. And optimized routing was the way to improve the performance of the model from an algorithmic perspective.
 Select Images auto-encoding algorithm based on deep convolution neural network Yijiang HE,Junping DU,Feifei KOU,Meiyu LIANG,Wei WANG,Ang LUO Journal of Shandong University(Engineering Science)    2019, 49 (2): 61-66.   DOI: 10.6040/j.issn.1672-3961.0.2017.432 Abstract （1413）   HTML （34）    PDF（pc） （1646KB）（457）       Save At present, image coding research was focused on information lossless, but it did not reflect the social network image differentiation. A novel social network images auto-encoding algorithm based on deep convolution neural network was proposed. The algorithm obtained good performance on image auto-encoding, which combined the feature extraction ability of deep convolutional neural network and characteristics of images in social networks. It combined the characteristics of the social network image with the clustering algorithm to cluster social network image and got the distance information, next the deep convolutional neural network was used to learn the distance information of these images, then it extracted the fully connected layer in the deep convolution neural network as the image coding, repeated the above steps and got the image coding finally. The experimental results showed that the proposed algorithm performed better than other algorithms of image search, and was more adaptive in the social network image search than that of the other algorithms mentioned.
 Select A microblog rumor events detection method based on C-GRU Lizhao LI,Guoyong CAI,Jiao PAN Journal of Shandong University(Engineering Science)    2019, 49 (2): 102-106, 115.   DOI: 10.6040/j.issn.1672-3961.0.2018.189 Abstract （1314）   HTML （37）    PDF（pc） （1249KB）（460）       Save A microblog rumor events detection model based on convolution-gated recurrent unit(C-GRU) was proposed. Combining the advantages of CNN and GRU, the microblog event′s posts was vectorized. By learning the features representation of the microblog windows through the convolution layer of CNN, the features of microblog windows was spliced into a sequence of window feature according to the time order, and the sequence of window feature was put into the GRU to learn feature representation of sequence for rumor events detection. Experimental results from real data sets showed that this model had better ability to rumor detection than other models based on traditional machine learning, CNN or RNN.
 Select An automatic reading method for pointer meter Yanghao ZHOU,Yifan LIU,Li LI Journal of Shandong University(Engineering Science)    2019, 49 (4): 1-7.   DOI: 10.6040/j.issn.1672-3961.0.2018.275 Abstract （1107）   HTML （149）    PDF（pc） （4120KB）（625）       Save An automatic reading method for automatically monitoring pointer meter in substation was proposed based on the machine learning and image processing algorithms, which was consisted of two stages: meter detection and pointer recognition. The position of the meter in the input image was detected by using the fully convolutional networks, and then the patch of the meter was extracted. The interference of illumination and shadow on the pointer recognition was reduced by using histogram equalization, median filtering and bilateral filtering, and the tilt of shooting was rectified by using the affine transformation. The position of the pointer was detected via the improved Hough transform. The reading was obtained by computing the angle of the pointer. The results showed that the method could detect the pointer meter and recognize the reading accurately for the pointer instrument in the substation. The method showed good robustness to the disturbances such as illumination and shadow, which could significantly reduce the substation inspection personnel workload and improve the work efficiency.
 Select Cross-domain text sentiment classification based on domain-adversarialnetwork and BERT Guoyong CAI,Qiang LIN,Kaiqi REN Journal of Shandong University(Engineering Science)    2020, 50 (1): 1-7,20.   DOI: 10.6040/j.issn.1672-3961.0.2019.293 Abstract （997）   HTML （38）    PDF（pc） （1549KB）（635）       Save In order to capture more sentence semantic information from the extracted shared sentiment features for cross-domain sentiment analysis, a deep network model based on domain adversarial mechanism and BERT (bidirectional encoder representations from transformers) was proposed. The model firstly used BERT to obtain the semantic representation vectors of sentences, and then extracted the local features of sentences with a convolutional neural network. A domain adversarial neural network was designed to make the representations of features extracted from different domains to be as indistinguishable as possible, that was, the features extracted from source domain and target domain had much more similarities; and a sentiment classifier was trained on the source domain dataset with sentiment labels, and it was expected that the trained sentiment classifier would have good classification performance in the source domain, and in the target domain. The experimental results on Amazon product reviews dataset showed that the proposed method achieved the expectation and was competent for achieving cross-domain text sentiment classification.
 Select Image denoising based on 3D shearlet transform and BM4D Shengnan ZHANG,Lei WANG,Chunhong CHANG,Benli HAO Journal of Shandong University(Engineering Science)    2020, 50 (2): 83-90.   DOI: 10.6040/j.issn.1672-3961.0.2019.262 Abstract （962）   HTML （10）    PDF（pc） （12734KB）（207）       Save Aimed at the disadvantage that the traditional block matching denoising method could only deal with two-dimensional images, an image denoising method based on 3D shearlet transform and BM4D(block-matching and 4D filtering) was proposed. This method used 3D shearlet transform to obtain transform domain coefficients, and realized joint filtering in transform domain through hard threshold and Wiener filtering stage. The 3D shearlet transformation was localized through two filtering stages: multi-scale decomposition and directional decomposition. The hard threshold and Wiener filtering were performed, which include grouping, collaborative filtering and aggregation. The 4D transformation of the cubes was based on the local correlationandon-local correlation cubes. The estimated values of each grouped cube were obtained by inverse transformation of 3D shearlet transform, and self-adaptive aggregation was performed at their original positions. PSNR(peak signal to noise ratio) and SSIM(structural similarity) were used as evaluation criteria. The results showed that this method could effectively remove image noise in high noise environment, and effectively improved the visual effect of the image with high accuracy.
 Select Advanced collaborative filtering recommendation model based on sentiment analysis of online review Chunlin QIAN,Xingfang ZHANG,Lihua SUN Journal of Shandong University(Engineering Science)    2019, 49 (1): 47-54.   DOI: 10.6040/j.issn.1672-3961.0.2017.485 Abstract （893）   HTML （19）    PDF（pc） （1086KB）（274）       Save Aiming at the uncertainty of users' subject opinions in online Chinese review, a sentiment analysis model was proposed based on uncertainty theory. An individual recommendation algorithm was designed on the basis of the proposed sentiment analysis model. Firstly, the tokenizers of ICTCLAS and IKAnalyzer were used to preprocess online Chinese review to generate characteristic words, and the point mutual information value of characteristic words accounting for the sentiment direction were computed based on sentiment dictionary (HowNet). Then, the sentiment analysis model was established via uncertainty theory of uncertain variable and uncertain set. In addition, the new similarity formula based on the proposed model was used to search the nearest neighbors. Finally, the recommendation lists were given. The experiments were carried out on two real datasets. The results showed that the proposed method could effectively improve the accuracy of recommendation and alleviate the sparse data problem.
 Select Review on smart highways critical technology Jianqing WU,Xiuguang SONG Journal of Shandong University(Engineering Science)    2020, 50 (4): 52-69.   DOI: 10.6040/j.issn.1672-3961.0.2020.149 Abstract （887）   HTML （478）    PDF（pc） （2579KB）（990）       Save Giving highway engineering "wisdom" and establishing new generation five-in-one system of "Internet+" design, construction, management, monitoring and operation, namely, the smart highway, was the hot issue of the interdisciplinary study of civil engineering, control engineering, mechanical engineering, transportation engineering, and computer science. To comprehensively understand the smart highway, this review focused on the critical technology in the integrated system in full life-cycle of the smart highway as well as systematically investigated the relevant previous efforts, critical common technologies, and future scopes on multi-function pavement material, smart construction, smart detection, autonomous vehicles, connected vehicles, and internet of things technology.
 Select Design of robot cloud service platform based on microservice Lei YIN,Fengyu ZHOU,Ming LI,Yugang WANG,Yinbo GUO,Ke CHEN Journal of Shandong University(Engineering Science)    2019, 49 (6): 55-62, 80.   DOI: 10.6040/j.issn.1672-3961.0.2019.251 Abstract （883）   HTML （10）    PDF（pc） （4187KB）（488）       Save To improve the intelligent level of service robots, a cloud service design method for service robots based on micro-service was proposed in this paper. The framework of cloud service based on micro-service was designed based on virtual model of service robot. Kinematics model, sensor model and environment model were proposed to map the robot parameters to cloud center. The interactive interface for cloud services was proposed. The adaptive matching of heterogeneous protocols was realized by using the protocol response of robots. Cloud service development methods were proposed in detail. The experiment result were done to demonstrate the cloud service results, real time and fine-grained quality of service of the proposed cloud service design for service robots.
 Select Review of energy consumption and demand forecasting methods Ming YANG,Pingjing DU,Fengquan LIU,Xupeng HAO,Yifan BO Journal of Shandong University(Engineering Science)    2020, 50 (1): 56-62,71.   DOI: 10.6040/j.issn.1672-3961.0.2019.180 Abstract （878）   HTML （937）    PDF（pc） （2290KB）（793）       Save In view of the increasing dependence of energy planning on energy demand forecasting and the difficulty of energy demand forecasting, this paper analyzed various energy forecasting methods and discussed the direction of energy development. The article analyzed the current demand situation of energy development methods from the direction of global energy demand development in recent years. The existing main energy forecasting methods were summarized and compared. The advantages and disadvantages of the existing research methods and applicable occasions were summarized. Combined with the new direction of energy development, the future development prospects of energy forecasting were given. Furthermore, this paper applied the LEAP model to predict the energy demand of the African region, and analyzed the regional energy complementation effect and the role of "electricity substitution" in the development of energy demand.
 Select An ensemble learning algorithm for unbalanced data classification Zongtang ZHANG,Sen WANG,Shilin SUN Journal of Shandong University(Engineering Science)    2019, 49 (4): 8-13.   DOI: 10.6040/j.issn.1672-3961.0.2019.050 Abstract （870）   HTML （9）    PDF（pc） （1159KB）（315）       Save For unbalanced data classification problem in underwater acoustic target recognition, a random subspace AdaBoost algorithm called RSBoost was proposed. Subtraining sample set was extracted by random subspace method in different underwater acoustic feature space and base classifier was trained in every subtraining sample set. The base classifier with the maximum margin mean of minority class was chosen as the base classifier of this round, the final ensemble classifier was formed iteratively. The experiment was carried out on the measured data, the performance of RSBoost and AdaBoost in different feature space was evaluated by F-measure and G-mean. The results showed that, compared with AdaBoost, the F-measure of RSBoost improved from 0.07 to 0.22 and the G-mean improved from 0.18 to 0.25, which showed that RSBoost was superior to AdaBoost in underwater acoustic unbalanced classification problem.
 Select Slope deformation and parameters sensitivity in the design of double-row supporting piles Yang GAO,Haokai SUN,Richeng LIU Journal of Shandong University(Engineering Science)    2019, 49 (3): 86-94.   DOI: 10.6040/j.issn.1672-3961.0.2019.019 Abstract （867）   HTML （11）    PDF（pc） （3228KB）（385）       Save This manuscript took the design of double-row pile support for a foundation pit project in Huaiyin District of Jinan City as an example. Elastoplastic solid elements and linear pile elements in finite difference numerical software was used to simulate the whole process of excavation-support-construction. The pile-soil interaction was taken into account and three-dimensional dynamic analysis carried out. In this paper, the design parameters of soil mechanics parameters, pile diameter, pile length, pile spacing, and coupling beam used to compare the sensitivity of foundation pit deformation. The calculation results for the "entity" and "structure" unit piles used for comparison. The research showed that the simulation process could better show the force mechanism of the double-row supporting pile construction process, and the calculation accuracy was high. The cohesive force, friction angle of the soil and the length of the pile and the distance of the pile in the controllable parameters had a great influence on the support effect. The calculation results could provide a reference for the selection of design parameters of double-row support piles.
 Select Entity alignment method based on adaptive attribute selection Jialin SU,Yuanzhuo WANG,Xiaolong JIN,Xueqi CHENG Journal of Shandong University(Engineering Science)    2020, 50 (1): 14-20.   DOI: 10.6040/j.issn.1672-3961.0.2019.415 Abstract （864）   HTML （20）    PDF（pc） （1167KB）（477）       Save Most existing entity alignment methods typically relied on external information and required expensive manual feature construction to complete alignment. Knowledge graph-based methods used only semantic information and failed to use structural information. Therefore, this paper proposed a new entity alignment method based on adaptive attribute selection, training an entity alignment model based on the joint embedding of the two knowledge graphs, which combined the semantic and structural information. Also, this paper proposed the use of strong attribute constraint based on adaptive attribute selection, which could adaptively generate the most effective attribute category and weight, to improve the performance of entity alignment. Experiments on two realistic datasets showed that, compared with traditional methods, the precision of the proposed method was improved by 11%.
 Select A method of multi-sensor data fusion under the complicated environment Mingming TIAN,Jihua YE,Shimin WANG,Yejing WAN Journal of Shandong University(Engineering Science)    2019, 49 (3): 32-38.   DOI: 10.6040/j.issn.1672-3961.0.2017.426 Abstract （860）   HTML （10）    PDF（pc） （1038KB）（185）       Save The existing methods do not use the reliability information of the source of evidence data collection, a multi-sensor data fusion algorithm based on the temperature data in complex environment was proposed. It proceeded from the sensor data source, analyzed the evidence source information, evaluated the confidence measure, and revised the conflict evidence with the confidence measure. In the evidence fusion moment, we used iterative fusion method to revise and fusion evidence until the fusion result convergence. Compared with other fusion methods, this method was effective and had better results in the question of evidence conflict.
 Select Air quality prediction approach based on integrating forecasting dataset Minghe GAO,Ying ZHANG,Rongrong ZHANG,Zihao HUANG,Linyan HUANG,Fanyu LI,Xin ZHANG,Yanhao WANG Journal of Shandong University(Engineering Science)    2020, 50 (2): 91-99.   DOI: 10.6040/j.issn.1672-3961.0.2019.404 Abstract （843）   HTML （15）    PDF（pc） （4733KB）（566）       Save Towarding the air quality prediction research problem, LightGBM was employed to propose and design a predictive feature-based air quality prediction approach, which could effectively predict the PM2.5 concentration, i.e., the key indicator reflecting air quality, in the upcoming 24-hour within Beijing. During constructing the prediction solution, the features of the training data set was analyzed to execute data cleansing, and the methods of random forest and linear interpolation were used to solve the problem of high data loss and noise interference. The predictive data features were integrated into the dataset, and meanwhile the corresponding statistical features were designed to imiprove the prediction accurancy. The sliding window mechanism was used to mine high-dimensional time features and increase the quantity of data features. The performance and result of the proposed approach were analyzed in details through comparing with the basedline models. The experimental results showed that compared with other model methods, the proposed LightGBM-based prediction approach with integrating forecasting data had higher prediction accuracy.
 Select Design for autonomous charging system of family companion robot Fengyu ZHOU,Fang WAN,Jiancheng JIAO,Junjian BIAN Journal of Shandong University(Engineering Science)    2019, 49 (1): 55-65, 74.   DOI: 10.6040/j.issn.1672-3961.0.2018.301 Abstract （835）   HTML （21）    PDF（pc） （7686KB）（378）       Save To address the limited battery capacity and discontinuous work ring of the family companion robot, a based autonomous charging and docking system was designed for the family companion robot using ROS. The Calman filtering algorithm is used to fuse the encoder data and IMU data. Meanwhile, the environment 2D grid map was constructed by the laser ranger data combined with the Rao-Blackwellized particle filter SLAM algorithm. The global path planning and the local path planning were carried out with the A* approach and the DWA algorithm to control robot reach the neighborhood of the charging station.The dual priority based infrared navigation and docking algorithm was used to guide the robot to the charging station to accurately docking with the charging station. The experimental results showed that the proposed system effectively solved the problem of limited charging distance compared with the traditional method, and had high docking efficiency, success rate, accuracy and generalization ability. Therefore, the system fully satisfied the charge demand of the family companion robot, which could be widely used to address real-world problems.
 Select Prediction method of wind power and PV ramp event based on deep learning Zhixiang LIANG,Xiaoming LIU,Ying MU,Yutian LIU Journal of Shandong University(Engineering Science)    2019, 49 (5): 24-28.   DOI: 10.6040/j.issn.1672-3961.0.2019.132 Abstract （828）   HTML （17）    PDF（pc） （1178KB）（382）       Save With the gradual increase of the renewable energy penetration rate, the ramp event that caused the unbalanced active power occured sometimes, and even a large load loss. Due to the insufficient accuracy of wind power and photovoltaic prediction, there were many operational scenarios to be considered. The time domain simulation could not meet the online assessment requirements. A method based on deep learning was proposed in this paper. Considering the generation unit and tie line adjustment ability, the stacked denoising autoencoder was used to extract each layer feature to train support vector machine. The wind power, photovoltaic and load forecast data, and the power of the tie line at the previous moment were taken as inputs, and whether the ramp event occured as an output. The vector machine was used to quickly predict whether a ramp event occured. The simulation results of practical power grid showed that the proposed method was fast and accurate. It could effectively identify ramp events.
 Select A multi-microcontroller communication method based on UART asynchronous serial communication protocol Jinping MA Journal of Shandong University(Engineering Science)    2020, 50 (3): 24-30.   DOI: 10.6040/j.issn.1672-3961.0.2019.705 Abstract （817）   HTML （12）    PDF（pc） （3677KB）（160）       Save To reduce the additional communication modules, complexity and costs of the different Internet of Things(IoT) devices communication, a multi-microcontroller communication method based on UART asynchronous serial ports was proposed. Based on the universal asynchronous receiver/transmitter(UART) serial communication method of the microcontroller, the control line was utilized to control the usage of the communication lines by the communication device, and a method of occupying the signal line by multiple machines in a time-sharing manner was realized. The master-slave control strategy was used to set the communication protocol. The master implements signal forwarding and identification, and the slave got signals from the master to achieve reliable and stable communication among multiple machines. By transplanting the μC/OS-Ⅱ operating system to the STM32 microcontroller, and using the real-time multitasking characteristics of μC/OS-Ⅱ, the signal reception, transmission and identification were designed into tasks of different priorities, and the master and the slave were realized. The functions of information receiving, sending and identification and the characteristics of multi-slave expansion were achieved through the communication protocol, solving the problem of multi-microcontroller communication that the traditional UART method could not achieve. The feasibility of the proposed method was verified through experiments, which provided a new solution for multi-microcontroller communication of edge devices in the Internet of Things.
 Select Research status and development trend of autonomous cognition and learning of robot manipulation skills Wei WANG,Feng WU,Fengyu ZHOU Journal of Shandong University(Engineering Science)    2019, 49 (6): 11-24.   DOI: 10.6040/j.issn.1672-3961.0.2019.229 Abstract （808）   HTML （23）    PDF（pc） （2077KB）（399）       Save Autonomous cognition and learning of manipulation skills, being one of the most important skills for robots, has been one of the hot issues in the field of robotics research. Combining with the authors' work in the field of robotics, this paper's focus is placed on giving a comprehensive overview of the mainstream modes, methods, algorithms, as well as advantages and disadvantages of different methods in terms of robots' manipulation skill learning. It concludes the challenges faced by autonomous learning and the key issues that need to be addressed for the individual cloud robots learning manipulation skills in the knowledge sharing mode. At the end, a potential solution for the above issues is given, and that is to integrate individual learning mode and shared learning model for the purpose of enhancing autonomous cognition and learning ability for robots.
 Select Review on application of artificial intelligence in power system restoration Yutian LIU, Runjia SUN, Hongtao WANG, Xueping GU Journal of Shandong University(Engineering Science)    2019, 49 (5): 1-8.   DOI: 10.6040/j.issn.1672-3961.0.2019.122 Abstract （798）   HTML （54）    PDF（pc） （1128KB）（859）       Save The research results of expert system, fuzzy mathematics, evolutionary algorithm and machine learning in power system restoration were summarized. It was pointed out that the existing researches were mainly offline restoration method and the researches about online decision-making were in the initial stage. In addition, the application potential of artificial intelligence technology in system restoration was prospected.
 Select An improved algorithm of maximal information coefficient based on dynamic equipartition Yanxia MENG,Yuchen GUO,Li WANG Journal of Shandong University(Engineering Science)    2019, 49 (5): 105-111.   DOI: 10.6040/j.issn.1672-3961.0.2018.209 Abstract （793）   HTML （6）    PDF（pc） （2337KB）（226）       Save In order to solve the problem of high computational time complexity of the maximal information coefficient algorithm, an improved algorithm of the maximal information coefficient(MIC) based on dynamic equipartition was presented. The scattered points shown in the grid were iterated and optimized by using dynamic mean division pairs of variables. The obtained mutual information entropy was regularized to obtain the optimal MIC value, and the multi-thread computation of the data set was carried out by using the POSIX parallel strategy, which made the computation more efficient in the computation of large data sets. Compared with the RapidMIC method on multiple data sets, the DE-MIC algorithm was faster and more efficient under the premise of preserving the generality and equitability of the maximal information coefficient algorithm.
 Select Visualization of interactive ThemeRiver based on time-series data Xindi CHEN,Tianrui LI,Huanhuan YANG Journal of Shandong University(Engineering Science)    2019, 49 (4): 29-35, 43.   DOI: 10.6040/j.issn.1672-3961.0.2017.400 Abstract （789）   HTML （14）    PDF（pc） （3445KB）（318）       Save ThemeRiver was a widely used technique when visualizing time-series data, which showed the whole trend efficiently. But the user often encountered ambiguity when observing a single river due to the influence of the lower river. However, the current visualization technique couldn't solve the issue. The following improvements of ThemeRiver were proposed upon the above issue: The river fluctuation was evaluated by the coefficient of variance instead of standard deviation, and the rivers with smaller fluctuation were arranged in lower position in order to reduce the impact on the upper rivers; The concept of ambiguity point was proposed, and the user could set fluctuation and max contribution as filtering parameters in the user interface so that the ambiguity points would be dynamically displayed by the system; The ambiguity point was visualized as bellow(The peak and valley of a river were represented by triangle and inverted triangle, respectively; The triangles would be filled with the color of the river with max contribution; A reorderable interactive ThemeRiver was proposed, in which the user was able to reorder the rivers by dragging them according to the information given by the ambiguity points). The proposed method was applied on the data containing 20 themes of 2015 from Zhihu and the results showed that the system could efficiently eliminate the ambiguity of ThemeRiver when presenting the trend of a single river and provided a flexible and personalized visualization.
 Select Survey of human-robot interaction control for autonomous driving Qijie ZOU,Haoyu LI,Rubo ZHANG,Tengda PEI,Yan LIU Journal of Shandong University(Engineering Science)    2019, 49 (2): 23-33.   DOI: 10.6040/j.issn.1672-3961.0.2017.503 Abstract （789）   HTML （9）    PDF（pc） （1268KB）（573）       Save This article summarizes the machine learning methods of human-robot interaction in autonomy vehicles. By introducing the value and significance of human-robot interaction, the relationship between the human-robot interaction problem definition and machine learning were identified, the human-robot interactions team framework was built. The frameworks of human-robot interaction and the research methods of autonomous driving system were reviewed, the general structure for solving human-robot interaction problems was presented. Furthermore, its machine learning algorithm from the two aspects of autonomous control system and driver modeling was introduced. The prospects of the future research direction were summarized.
 Select Key frame extraction based on ViBe algorithm for motion feature extraction Qiuling LI,Baomin SHAO,Lei ZHAO,Zhen WANG,Xue JIANG Journal of Shandong University(Engineering Science)    2020, 50 (1): 8-13.   DOI: 10.6040/j.issn.1672-3961.0.2019.276 Abstract （776）   HTML （14）    PDF（pc） （4029KB）（418）       Save Aiming at the fact that the background was dominant in the key frame extraction algorithm, in which the foreground target was too small and it was not easy to extract the features of moving targets in sports video, a key frame extraction algorithm for foreground moving target feature extraction based on background modeling algorithm was proposed, which was called visual background extractor (ViBe) algoritm. The foreground target detection of video sequence was firstly carried out using ViBe algorithm, afterwards the scale-invariant feature transformation (SIFT) features of the foreground moving target were extracted. Based on the similarity calculated from video frame series, the key frames of video were output according to the key frame discrimination method. The experimental results showed that the proposed algorithm could solve the problem of missed selection and misselection in traditional key frame extraction. Compared with the algorithm based on SIFT distribution histogram, the F1 score was well improved. The algorithm based on ViBe could effectively identify key frames in sports video.
 Select The vulnerability mining method for KWP2000 protocol based on deep learning and fuzzing Chengbin ZHANG,Hui ZHAO,Zongyu CAO Journal of Shandong University(Engineering Science)    2019, 49 (2): 17-22.   DOI: 10.6040/j.issn.1672-3961.0.2018.340 Abstract （764）   HTML （17）    PDF（pc） （1273KB）（277）       Save A kind of vehicle-onboard diagnosis Protocol standard, keyword protocol 2000 (KWP2000) KWP2000, was investigated in details. KWP2000 was widely used in the automobile industry and the loophole of possible communication Protocol. We analyzed the current situations of the fuzzing, and based on this, we proposed a generative adversarial networks (GAN) by deep learning neural network for automobile body network KWP2000 protocol hole mining method. The forward feedback network was closeted as the generation model, and the support vector machine was used as the discriminant model. We used the neural network model to train the test case data of the KWP2000 protocol data, the fuzzing of KWP2000 was carried out by using these test case data. Through experiments, we found that the target protocol KWP2000 had long loopholes, coding errors and other vulnerabilities. Experimental results showed that this fuzzing method was efficient and safe.
 Select Research onfeature selection technology in bearing fault diagnosis Jiachen WANG,Xianghong TANG,Jianguang LU Journal of Shandong University(Engineering Science)    2019, 49 (2): 80-87, 95.   DOI: 10.6040/j.issn.1672-3961.0.2018.268 Abstract （761）   HTML （9）    PDF（pc） （3981KB）（219）       Save A new method based on feature selection (FS) was proposed to select efficient features to promote the classification accuracy in bearing fault diagnosis. First, the outstanding features whose classification accuracy were higher than the threshold were directly selected by diagnosis model from a big feature set. Then the significant combinations of features which had less dimensions and higher classification accuracy were selected in the candidate feature set by a distinctive feature-oriented manner. Experiments showed that the proposed method had advantages in selecting efficient features, reducing the model parameters, decreasing the demand of samples and enhancing the model classification accuracy. As a result, it provided a new idea for feature selection and improved the efficiency of bearing fault diagnosis.
 Select End-to-end security encryption scheme of NB-IoT for smart grid based on physical unclonable function Donglan LIU,Xin LIU,Jianfei CHEN,Wenting WANG,Hao ZHANG,Lei MA,Dong LI Journal of Shandong University(Engineering Science)    2020, 50 (1): 63-71.   DOI: 10.6040/j.issn.1672-3961.0.2019.034 Abstract （751）   HTML （18）    PDF（pc） （3224KB）（189）       Save In order to improve the high security of power network data transmission, an end-to-end security encryption scheme of NB-IoT (narrow band internet of things, NB-IoT) for smart grid based on physical unclonable function and domestic cipher algorithm SM3 was proposed in this paper. A self-controllable NB-IoT application layer security architecture was designed by introducing the SM3, extending the existing key derivation structure of LTE, and combining the physical unclonable function to ensure the generation of encryption keys between NB-IoT terminals and power grid business platforms. Analysis and experiment showed that the proposed scheme realized secure data transmission and bidirectional identity authentication between IoT devices and terminals. Its features included high compatibility, low communication costs, lightweight and flexible key update. In addition, the scheme also supported terminal authentication during key agreement, which furtherly enhanced the security of business systems in smart grid.
 Select Review of developments in titanium-based coagulants Baoyu GAO,Xin HUANG,Guangping YAO,Qinyan YUE Journal of Shandong University(Engineering Science)    2020, 50 (1): 109-114.   DOI: 10.6040/j.issn.1672-3961.0.2019.359 Abstract （741）   HTML （45）    PDF（pc） （1129KB）（579）       Save The development process and its applications of titanium-based coagulant were reviewed, including the mono-titanium coagulants and poly-titanium coagulants, and the single titanium coagulants and composite titanium coagulants, and the recent research process and the future development of titanium coagulants, which could provide guidances and references for the research and development of titanium-based coagulants.
 Select Laminar flow field characteristics in the stirred vessel equipped with an eccentric-shaft impeller Meiting LI,Wei LI,Xiaoguang LI,Fengling YANG Journal of Shandong University(Engineering Science)    2019, 49 (4): 93-98, 107.   DOI: 10.6040/j.issn.1672-3961.0.2018.530 Abstract （728）   HTML （5）    PDF（pc） （2868KB）（559）       Save For the purpose of improving mixing efficiency of the stirred tank with high viscosity fluid laminar flow condition, an eccentric-shaft agitation method was proposed. With a purity of 99% glycerol as medium, and the traditional 2-flat-blade impeller as the research object, the laminar flow fields were numerically studied. The inner diameter of the stirred vessel was 0.3m and the agitated fluid was glycerol. The modeling reliability and simulation methods of the concentric agitation were validated by experimental results. In comparison with the concentric agitation, flow fields generated by eccentric and eccentric-shaft agitation were asymmetrical, and theoverall volume weitghted average velocity of the groove fluid generated by eccentric-shaft agitation could be raised by 68%. And furthermore, when operated under the same speed, the power consumption of eccentric-shaft agitation increased 15.3% than that of concentric agitation. However, the eccentric shaft agitation increased the speed in the tank and expanded the range of disturbance. Accordingly, the superiority of eccentric-shaft agitation of improving the mixing efficiency in the tank was apparent.
 Select Image attribute annotation based on extreme gradient boosting algorithm Hongbin ZHANG,Diedie QIU,Renzhong WU,Tao ZHU,Jin HUA,Donghong JI Journal of Shandong University(Engineering Science)    2019, 49 (2): 8-16.   DOI: 10.6040/j.issn.1672-3961.0.2018.271 Abstract （721）   HTML （6）    PDF（pc） （4657KB）（311）       Save To improve annotation performance, a novel image attribute annotation model based on eXtreme gradient boosting (XGBoost) algorithm was proposed: image features i.e. local binary patterns (LBP), Gist, scale invariant feature transform (SIFT), and visual geometry group (VGG) were extracted respectively to better characterize the key visual content of images. Then the state-of-the-art boosting algorithm called XGBoost was used to design a strong classifier by integrating a group of weaker classifiers. Based on the strong classifier, image attribute annotation was implemented. A lot of valuable deep semantic implied by image attribute was mined in turn to create a novel hierarchical attribute representation mechanism, which was closer to human's objective cognition. Finally, transfer learning strategy was designed to further improve annotation performance. Experimental results showed that the key visual content of images was truly characterized by the Gist feature. Compared to the best competitor before transfer learning, the accuracy of basic transfer (BT) learning strategy was improved about 8.69%. Compared to the best competitor of BT, the accuracy of hybrid transfer (HT) learning strategy was improved about 17.55%. The annotation accuracy was improved by the presented model.
 Select Hydraulic turbine operation status detection based on LSTM network prediction Chang CHEN,Xiaolei LI,Weiyu CUI Journal of Shandong University(Engineering Science)    2019, 49 (3): 39-46.   DOI: 10.6040/j.issn.1672-3961.0.2018.240 Abstract （719）   HTML （13）    PDF（pc） （3204KB）（308）       Save Long short-term memory (LSTM) networks was adapted to make accurate prediction of the unit's operation status. The streaming monitoring data of the turbine unit was standardized, and the sliding window technology was used to convert the data into the training data set and test data set for LSTM network training. The LSTM prediction model structure was given, and the structure of LSTM prediction model was fine-tuned, such as the number of network layers and the number of hidden layer neurons. The time series data prediction model of the hydro turbine unit was established. The experimental analysis proved that the multi-measurement-based LSTM network prediction model had higher prediction accuracy than other models, which calculated the health deviation based on the improved radar image analysis method and successfully detected the abnormality of the No. 5 hydraulic turbine unit of a hydropower plant at the end of May, and verified the validity of the model.
 Select A force-based method for robot hole-searching and assembly Yujun WU,Wei WU,Yu GUO,Jian GUO Journal of Shandong University(Engineering Science)    2019, 49 (5): 119-126.   DOI: 10.6040/j.issn.1672-3961.0.2018.396 Abstract （715）   HTML （12）    PDF（pc） （7332KB）（309）       Save In the process of replacing arresters for live working robots, a method based on force control for robot to search hole on insulated cross-arms and assembling arresters was proposed for accurate assembly. In the control process, a force sensor was installed at the end of the robot for the special shape of the arrester's lower end and the scene of the pole cross arm, and the converted force was used as the controlled amount. A peg-in-hole assembly strategy was designed, which included four processes of touching, hunting, evolving and inserting. The robot could insert the arrester's lower end into the mounting hole on the cross arm accurately. The experimental results verified the effectiveness of the proposed method.
 Select Vehicle classification and tracking for complex scenes based on improved YOLOv3 Shiqi SONG,Yan PIAO,Zexin JIANG Journal of Shandong University(Engineering Science)    2020, 50 (2): 27-33.   DOI: 10.6040/j.issn.1672-3961.0.2019.412 Abstract （698）   HTML （16）    PDF（pc） （5481KB）（292）       Save Aiming at the influence of weather conditions and mutual occlusion of vehicles on vehicle classification and tracking accuracy and stability, a hybrid model based on improved YOLOv3 and matching tracking was proposed. The improved YOLOv3 network refered to DenseNet′s design idea, replaced the residual layer in the network with a dense convolution block and changed the design structure of the network. The fused features of dense convolution blocks and convolution layers were classified by using Softmax classifier. According to the detection result of single frame image, the target matching function was designed to solve the vehicle tracking problem in video sequence. In the KITTI dataset test, the improved algorithm achieved an average precision of 93.01%, the number of frames per second reached 48.98, and the average recognition rate in the self-built dataset was 95.79%. The experimental results showed that the proposed method could effectively distinguish the types of vehicles in complex scenes with higher accuracy. At the same time, the method had higher accuracy and robustness in vehicle tracking.
 Select TBM comprehensive advanced geological prediction in a tunnel andits application in Yangling Tunnel Jun LIU,Lijun HAN,Qingbin MENG Journal of Shandong University(Engineering Science)    2019, 49 (4): 51-60.   DOI: 10.6040/j.issn.1672-3961.0.2018.093 Abstract （690）   HTML （5）    PDF（pc） （5462KB）（354）       Save Advanced geological prediction was used to play an irreplaceable role in the aspect of informative tunnel construction, effective disaster prevention, economical construction support in tunnel engineering. The effectiveness of the advanced geological prediction method in the drilling and blasting method under the tunnel boring machine (TBM) specific construction environment was compared and analyzed. The comprehensive advanced geological prediction system under TBM tunnel construction with "long and short distance, geophysical and drilling exploration, sound wave reflection and electromagnetic method, key and normal" was established. Based on the horizontal sonic profiling (HSP) sound wave reflection, drilling process monitoring (DPM) advanced drilling, transient electromagnetic instrument, the multiple solutions and error superposition of single advanced geological prediction method were suppressed effectively. And then the 3D forward theory calculation was considered as an appropriate way for enhancing the accuracy and reliability. Based on the law of electromagnetic response and the study of removing electromagnetic interference, the results summarized from the paper were used to guide the application of transient electromagnetic method in Yixing Yangling Tunnel and to provide a reference for future projects.
 Select Phase change characteristics of paraffin in rectangular storage unit Huilin ZHOU,Yan QIU Journal of Shandong University(Engineering Science)    2019, 49 (4): 99-107.   DOI: 10.6040/j.issn.1672-3961.0.2019.018 Abstract （684）   HTML （8）    PDF（pc） （3235KB）（694）       Save To study the heat transfer characteristics of phase change material (PCM) in a storage unit and improve the heat transfer efficiency of phase change heat exchanger, the enthalpy-porous media model and FLUENT program were used to carry out a numerical investigation on the heat transfer process of paraffin in a rectangular heat storage unit. The element liquid fraction β and dimensionless Fo, Ste and Ra were introduced to analyze the influences of different positions outside the tube and inlet temperatures of heat transfer fluid on the melting and solidification process of paraffin. The results showed that the paraffin outside the tube was melted in order from the upper to the left/right part, then the lower part. The total melting time of paraffin in upper part was shorten by at least 20% compared with other parts. Conversely, in the heat release process, the paraffin was solidified in order from the lower, the left/right part and the upper part. The heat transfer mechanism in the unit changed gradually from heat conduction to natural convection in the thermal storage process. The efficiency of heat storage and release could be improved significantly by increasing the temperature difference between heat transfer fluid and paraffin. The criterion of β was obtained by polynomial fitting.
 Select Features analysis for Chinese irony detection Rongxiang ZHOU,Xiuyi JIA Journal of Shandong University(Engineering Science)    2019, 49 (1): 41-46.   DOI: 10.6040/j.issn.1672-3961.0.2018.341 Abstract （683）   HTML （6）    PDF（pc） （1104KB）（247）       Save The research object was data in microblog. The features of irony detection were studied. In view of the characteristics of microblog and irony detection, a variety of features were constructed, such as emotional phrases, emoticons and so on. The experiments showed that the proposed irony features improved 0.34% on recognition accuracy, 0.74% on recall and 0.18% on F-measure, compared with the existing ones for the imbalanced datasets. The proposed irony features also improved 0.44% on recognition accuracy, 2.54% on recall and 0.14% on F-measure, compared with the existing ones for the balanced datasets.
 Select The interlaminar shear rule between flexible base and semi-rigid base Zhizhong ZHAO,Hongzhe LIU,Guiqiang LIU,Zhenyu YANG Journal of Shandong University(Engineering Science)    2019, 49 (3): 57-62.   DOI: 10.6040/j.issn.1672-3961.0.2017.436 Abstract （681）   HTML （9）    PDF（pc） （2628KB）（189）       Save ATB-25 asphalt mixture and cement stabilized macadam were taken to shape composite specimen, the amount of asphalt and chip in the composite specimen layer was controlled. The develop shear apparatus was used to carry out shear strength test, the influence factors and change rules of interlaminar shear strength were analyzed based on the test results. Different test schemes were worked out by setting up the test temperature and the amount of chip and asphalt. The test results showed that the temperature was the most influencing factor on the shear strength of the layer, when the test temperature increased from 5 ℃ to 35 ℃, the maximum shear stress decreased from 1.36 MPa to 0.16 MPa. The second influencing factor was the amount of asphalt, the aggregate content had the least influence, and the optimum asphalt content was different due to the different temperature. Therefore, in practical engineering, the amount of asphalt and gravel should be considered in combination with the location. The asphalt content in high temperature area should be slightly less than in the low temperature area. The aggregate content should be strictly controlled to improve the quality of interlayer connection and define the laws of interlaminar shear strength and connection.
 Select Flow characteristics of rough rock fractures under wide range of Reynolds numbers Jie LIU,Zhechao WANG,Yupeng ZHANG,Huayang SUN Journal of Shandong University(Engineering Science)    2019, 49 (4): 70-77, 85.   DOI: 10.6040/j.issn.1672-3961.0.2018.533 Abstract （676）   HTML （11）    PDF（pc） （6754KB）（339）       Save Through experiments and theoretical analysis, the flow characteristics of rock rough fractures with different geometric parameters were studied, including non-Darcy coefficient β, critical Reynolds number Rec and non-Darcy effect factor E. Testing equipment for fluid flow in fractures was developed and 9 rough single fracture models with different apertures and Joint Roughness Coefficients (JRC) were prepared. Laboratory tests on flow in rough fractures with Reynolds numbers, i.e., from 2 to 5000 were performed. The characteristics of flow in single fractures with different roughness (JRC=2-20) under wide range of Reynolds numbers were obtained. It was shown that fracture roughness had a significant impact on the nonlinear flow characteristics of the fracture. Based on the Forchheimer equation, the effects of fracture roughness on equation parameters were quantitatively studied. The larger fracture roughness was, the more likely it was to cause the nonlinearity of the fracture flow, the smaller critical Reynolds number was, and the stronger non-linear effect would be.
 Select Crack initiation theory and experimental verification of pre-existing plane cracks with seepage pressure Yajuan XIE,Song YU,Bangxiang LI,Xiang XU,Weishen ZHU Journal of Shandong University(Engineering Science)    2019, 49 (4): 36-43.   DOI: 10.6040/j.issn.1672-3961.0.2018.140 Abstract （670）   HTML （4）    PDF（pc） （3494KB）（149）       Save In order to study the crack initiation characteristics of pre-existing plane cracks with seepage pressure, cracks were divided into two types: open type and closed type. Theoretical analysis and experimental verification were carried out. The theoretical derivation of crack stress intensity factor, crack initiation angle and crack initiation stress was expounded in detail, and the crack initiation strength formula for closed crack under compression was given. Precast concrete mortar specimens with open-type crack were produced and subjected to biaxial compression experiments. The variation of stress intensity factor and crack initiation angle with crack inclination angle, hydraulic pressure and crack thickness were investigated through experimental verification and analysis. The relationship of crack initiation strength with hydraulic pressure and lateral pressure were also investigated. The results showed that type Ⅰ stress intensity factor of open-type crack with seepage pressure increased with the increase of crack angle. The type Ⅰ crack was most easily initiate when the crack angle was 75°. KⅠ decreased with the increase of water pressure and crack thickness. Type Ⅱ stress intensity factor of shear crack was distributed symmetrically with the crack inclination angle, and the maximum value was obtained at the crack inclination angle of 45°, irrespective of hydraulic pressure and crack thickness. Crack initiation angle of open-type crack decreased with the increase of crack dip, hydraulic pressure and crack thickness; crack initiation strength was inversely proportional to hydraulic pressure and proportional to lateral pressure. The research results could provide reference for crack initiation and crack propagation with seepage pressure in geotechnical engineering.
 Select Real-time traffic prediction based on MGU for large-scale IP backbone networks Fang GUO,Lei CHEN,Ziwen YANG Journal of Shandong University(Engineering Science)    2019, 49 (2): 88-95.   DOI: 10.6040/j.issn.1672-3961.0.2018.342 Abstract （665）   HTML （9）    PDF（pc） （4768KB）（280）       Save In order to overcome the shortcomings of long short-term memory (LSTM) computing cost, a real-time traffic prediction method based on minimum gated unit (MGU) for large-scale IP backbone networks was proposed. The experimental results showed that compared with the LSTM-based traffic prediction method, the proposed method achieved fairly or even better traffic prediction performance with less model training time, meanwhile it outperformed the most advanced feed forward neural network (FFNN), LSTM and gated recurrent unit(GRU) in terms of prediction accuracy and real-time performance.
 Select LDA-based topic feature representation method for symbolic sequences Chao FENG,Kunpeng XU,Lifei CHEN Journal of Shandong University(Engineering Science)    2020, 50 (2): 60-65.   DOI: 10.6040/j.issn.1672-3961.0.2019.760 Abstract （650）   HTML （6）    PDF（pc） （1403KB）（247）       Save To address the problems of high feature dimensionality and high algorithm time complexity in the existing methods, a topic feature representation method was proposed to transform the symbolic sequences into a set of topic probability vectors, based on the topic model latent Dirichlet allocation (LDA) commonly used in text mining. In the new method, each short sequence gram was considered as the shallow feature (word) of the sequence, and the topics with their probability distributions were extracted as the deep features of the sequences using the LDA model learning algorithm.Experiments were carried out on six real-world sequence sets, and compared with the existing grams-based and Markov model-based methods. The results showed that the new method improved the learning efficiency of the representation model while reducing the feature dimensionality, and achieved better accuracy in the application of symbolic sequence classification.
 Select Abnormal sound detection of washing machines based on deep learning Chunyang LI,Nan LI,Tao FENG,Zhuhe WANG,Jingkai MA Journal of Shandong University(Engineering Science)    2020, 50 (2): 108-117.   DOI: 10.6040/j.issn.1672-3961.0.2019.419 Abstract （645）   HTML （15）    PDF（pc） （5582KB）（594）       Save Based on the convolutional neural network (CNN) framework, a model for abnormal sounds recognition of washing machine was proposed. According to the remarkable feature extraction ability and translation invariance of convolutional neural network, the abnormal sound features of washing machines were learned, so as to achieve the purpose of the automatic intelligent recognition of abnormal sounds for washing machines in production line. This method provided a complete process to solve the problems of training datasets establishment and data imbalance. A network model for data augmentation called advanced deep convolution generated adversarial network (ADCGAN)was proposed to solve the problem of training data scarcity. The traditional deep convolution generated adversarial network (DCGAN) model was improved to better adapt to the generation of industrial sounds. This model could be used to extend the original data and generate the abnormal sound augmented datasets of washing machine. The augmented datasets was used to train the convolutional neural network, and the test accuracy reached 0.999. The generalization ability of abnormal sounds recognition model for washing machine network was tested by using the data set with background noise signal added. The correct recognition rate reached 0.902, which indicated that this network had good robustness in recognizing abnormal noises of washing machines.
 Select Impact analysis on construction of large section and small spacing road tunnel Mingcai LIU Journal of Shandong University(Engineering Science)    2019, 49 (4): 78-85.   DOI: 10.6040/j.issn.1672-3961.0.2018.055 Abstract （644）   HTML （4）    PDF（pc） （3052KB）（255）       Save The construction project of Fo Yangling Tunnel was taken as the research background, the construction was carried out by using benching tunneling construction method. The fast Lagrangian analysis of continua 3D (FLAC3D) finite element software was used as the calculation platform to establish the finite element model of tunnel with large section and small clear distance. The tunnel model was numerically simulated and subsidence, stress of surrounding rock and lining structural were analyzed. And the corresponding monitoring and reinforcement measures were put forward. The results of numerical analysis showed that the settlement of the upper part of the advance tunnel was larger than the subsequent tunnel. The construction of the subsequent tunnel increased the vertical displacement of the advance tunnel, which caused the increase of maximum principal stress and the minimum principal stress. The stress of tunnel lining structure was concentrated in the area from arch waist to arch foot. The support structure of advance tunnel played the role of stabilizing rock mass in the excavation of subsequent tunnel, and the excavation of the subsequent tunnel would make the stress concentration of the middle rock. In the construction, attention should be payed to the surrounding rock and reinforcement measures should be taken. The analysis results could provide a scientific basis for the design and construction of large section and small spacing road tunnel.
 Select Improvement of bandwidth model for high speed optical communicationlaser and its optimization by parallel computing Si YANG,Sitong LI,Jindong ZHANG,Yu BAI Journal of Shandong University(Engineering Science)    2019, 49 (1): 17-22, 29.   DOI: 10.6040/j.issn.1672-3961.0.2018.196 Abstract （642）   HTML （4）    PDF（pc） （3794KB）（231）       Save In order to adapt to high-speed network, features of optical fiber communication laser need to be optimized. Laser thermal sensing model was improved by measuring the data exported by vertical cavity surface emitting laser (VCSEL). In the modified model, the quantum rate equation was developed to describe features of laser VCSEL. Considering the relationship between the gain constant, the transparent current carriers and the temperature, the high order fitting method was introduced to optimize the laser bandwidth system, which maked it closer to the measured data. Meanwhile, based on the parallel data mining method, the buffer helped to speed up the solving process to satisfy the demand of the next-generation high-speed network. The results showed that the output light power appeared to decrease in thermal saturation under the high temperature and large injection current. The model after optimization was closer to the actual measured data, genetic algorithm had a strong adaptability towards scale change of system, and the route planned by algorithm was reliable. The calculation speed was increased by 9.15% based on parallel data processing method. This modified model met the demand of the new-generation high-speed fiber optic communication, and considered the thermal limitation.
 Select Study on a low-speed direct drive transverse-flux switched reluctance motor Zhenwei ZHAO,Zhigang DONG,Yongbin LI Journal of Shandong University(Engineering Science)    2019, 49 (6): 81-85.   DOI: 10.6040/j.issn.1672-3961.0.2019.339 Abstract （641）   HTML （9）    PDF（pc） （2985KB）（203）       Save In order to improve the torque density of switched reluctance motor (SRM), achieve low speed and large torque output, and meet the requirements of direct electrical drive systems such as servo and electric vehicle, a new type of transverse magnetic flux switch reluctance motor (TFSRM) was proposed. The principle of transverse flux motor was applied to switch reluctance motor, which simplified the complexity of the motor structure and improved the torque density of the motor. The working principle and structural characteristics were introduced in detail. The magnetic field distribution of the stator and rotor in different relative positions were analyzed by the three-dimensional equivalent magnetic network method, and the torque angle characteristics of the motor were calculated under constant current control. A closed-loop current control system based on digital signal processor was designed, and the prototype was tested. The experimental results were consistent with the theoretical analysis, which verified the feasibility and validity of the theoretical analysis and design method.
 Select Face recognition based on improved prameter-free supervised localitypreserving projections Jun FAN,Qiaolin YE,Ning YE Journal of Shandong University(Engineering Science)    2019, 49 (1): 10-16.   DOI: 10.6040/j.issn.1672-3961.0.2017.419 Abstract （638）   HTML （7）    PDF（pc） （3094KB）（248）       Save To solve the problem of unsupervised and complexity of parameter selection of the locality preserving projection algorithm, an improved supervised dice parameter-free locality preserving projection algorithm (SdPLPP)was proposed. SdPLPP constructed affinity matrix by using generalized Dice coefficient and extract features of data under the supervised mode, which could avoid the problems of parameters selection and adjustment of locality preserving projection (LPP) algorithm. The proposed algorithm performed experiment of image visualization based on the Iris dataset, analyzed the relationship between the value of the distance of sample data and the performance of the algorithm. To verifying the effectiveness and performance of algorithm, SdPLPP carried out the feature extraction experiments based on three kinds of human face databases, such as ORL, Yale and FERET, and used nearest neighbor classifier to get correct recognition rate. The experimental results showed that the SdPLPP algorithm was superior to PCA, ULDA, LPP, SPLPP and EP-SLPP algorithm in face recognition, and it was better than other algorithms of supervised parameter-free locality preserving projections.