International Journal of Wireless and Mobile Computing (45 papers in press)
Multicast stable path routing protocol for wireless ad-hoc networks
by K.S. Saravanan, N. Rajendran
Abstract: Wireless Ad-Hoc Networks (WANETs) enable steady communication between moving nodes through multi-hop wireless routing path. The problem identified is how to improve the lifetime of the route and reduce the need for route maintenance. This helps to save bandwidth and reduce the congestion control available in the network. This paper aims to focus on redesign and development of multicast stable path routing protocol with special features that determine long-living routes in these networks. An extensive ns-2 simulation based performance has been analysed of three widely recognised stability oriented wireless ad-hoc network routing protocols, namely are Associativity Based Routing (ABR) protocol, Flow Oriented Routing Protocol (FORP) and Lifetime Route Assessment Based Routing Protocol (LRABP). The order of ranking of the protocols in terms of packet delivery ratio, average hop count per route, end-to end delay per packet and the number of route transitions is presented.
Keywords: wireless ad-hoc networks; multicast routing protocol; wireless communication; routing protocol.
Study on carbon footprint model and its parameter optimisation of wave soldering process based on response surface method
by Renwang Li, Haixia Liu, Jiaqi Li, Jinyu Song, Rong Jie
Abstract: In order to respond to low carbon manufacturing, from energy, materials and process carbon emissions, etc., this paper constructs a carbon footprint model for the wave soldering process in the module workshop of H Company. Based on this model, the carbon footprint value of the wave soldering process is calculated. On the basis of selecting appropriate parameter factors, a parameter optimisation model of the wave soldering process life-cycle carbon footprint is constructed, and the optimum parameters combination is analysed by the response surface method, which contains surface area, clip velocity, clipping angle, flux flux, purity of solder, temperature of tin furnace, height of wave peak, etc. The the response values obtained are verified. The experimental results show that the optimised parameters are used to process and manufacture the wave soldering process, and the carbon emissions produced by wave wave welding can be controlled from above 15 kg CO2 to 12 kg CO2.
Keywords: wave soldering; carbon footprint; calculation model; response surface method; parameter optimisation.
Performance Analysis Based on Probability of False Alarm and Miss Detection in Cognitive Radio Network
by Ramkrishna Ghosh
Abstract: The rising requirement of wireless applications has set an ample of boundaries on the practice of accessible radio spectrum which is inadequate and valuable means. If examining of a radio spectrum reveals that several frequency bands in the spectrum are mostly vacant often, several other frequency bands are partly occupied and the residual frequency bands are greatly used. This directs that radio spectrum is underutilized. The underutilization of radio spectrum is reduced by the Cognitive Radio (CR). CR is a demanding technology that offers a new efficient technique to progress exploitation of available electromagnetic spectrum resourcefully. CR specifies wireless design in which a transmission scheme does not activate in a predetermined band. Spectrum Sensing (SS) assists to perceive the spectrum holes provided that high spectral resolution ability. In our paper, we have demonstrated the statistical characteristics of false alarm and miss detection probabilities.
Keywords: Cognitive Radio; Spectrum Sensing; Miss Detection; False Alarm; Wireless Communication; Spectrum.
Modelling and simulation of a thermo-fluid system with one-dimensional distributed parameters on Modelica
by Yiming Yuan, Zefei Zhu, Guojin Chen, Chang Chen
Abstract: Thermo-fluid systems are applied to industrial production and daily life widely and influence human activities deeply. This paper describes a modelling and simulation method for a thermo-fluid system with one-dimensional parameters on Modelica, aiming at providing a framework and modelling process of a one-dimensional thermo-fluid system model. The method of lines is introduced to convert the partial differential equation, which is used to describe the one-dimensional thermo-fluid system, into a group of differential-algebraic equations. Based on the Newton-Gregory polynomial, the discretisation method and step is presented and the difference expression is deduced for the first and second order spatial derivative terms in PDE. Two illustrative examples, wave equation and human body heat loss model, are presented to confirm the veracity and accuracy of the proposed method.
Keywords: thermo-fluid system; one-dimensional distributed parameter; partial differential equation; differential algebra equation; Modelica.
Semi-supervised learning of pose-specific detector for human lying-pose detection
by Xia Daoxun, Liu Haojie, Li Weian
Abstract: Under the superlow-altitude aerial image, human lying-pose detection is an important problem in object detection. This paper is mainly focused on the application study of an unmanned aerial vehicle (UAV) life detector after a disaster, and we study the problem of learning an effective pose-specific detector using weakly annotated images and a deep neural network. This typical approach 1) clusters a series of human poses for the human lying-pose and assigns an image-level label to all human lying-poses in each image and breaks them down into several categories; 2) trains multiple classifiers for each category using a deep neural network; and 3) uses the boosted semi-supervised CNN forest classifier to select a human lying-pose with high confidence scores as the positive instances for another round of training. Experiments on the XiaMen University Lying-Pose Dataset (XMULP) show that significant performance improvement can be achieved with our proposed method.
Keywords: human lying-pose detection; pose-specific detector; semi-supervised learning; object detection.
Research on pipe crack detection based on image processing algorithm
by Licheng Huang, Bo Tao, Donghai Chen, Xun Zhang, Gongfa Li
Abstract: The detection of pipe cracks based on machine vision is a new and effective technology. However, it requires high quality of the image. Moreover, images with adequately lighting, evident cracks, and clean backgrounds are difficult to obtain in practice. This paper proposes an algorithm for pipe crack detection in natural background. The algorithm performs filtering, background segmentation, edge detection, threshold segmentation, morphological contour extraction, and annotation on the image. This paper also proposes an adaptive threshold segmentation method to obtain the clear crack. By comparing the proposed algorithm with the DEE algorithm, the result shows that the proposed algorithm has certain advantages in experiments. The experiment results show that the algorithm proposed can also be used in the detection of significant pipe cracks.
Keywords: pipe cracks; noise reduce; Sobel operator; edge detection; image processing.
An optimal condition of robust low-rank matrices recovery
by Jianwen Huang, Sanfu Wang, Jianjun Wang, Feng Zhang, Hailin Wang, Jinping Jia
Abstract: In this paper we investigate the reconstruction conditions of nuclear norm minimisation for low-rank matrix recovery. We obtain a sufficient condition to guarantee the robust reconstruction or exact reconstruction of all rank matrices via nuclear norm minimisation. Furthermore, we not only show that when $t=1$, the upper bound is the same as the result of Cai and Zhang, but also demonstrate that the gained upper bounds concerning the recovery error are better. Moreover, we prove that the restricted isometry property condition is sharp. Besides, the numerical experiments are conducted to reveal the nuclear norm minimisation method is stable and robust for the recovery of low-rank matrix.
Keywords: low-rank matrix recovery; nuclear norm minimisation; restricted isometry property condition; compressed sensing; convex optimisation.
Experimental study on modal characteristics of flame tube in a can-type combustor in an aero-engine
by Guanbing Cheng, Yinsheng Chai
Abstract: The flame tube is a key component in the combustor of gas turbines. The fuel air mixtures burn efficiently in the tube and control the temperature distribution in both the radial and circumferential directions by adjusting the air entrance from the various geometrical holes in the liner. Thus, understanding the dynamic characteristics of the flame tube structure is one of the key problems in understanding performance of the combustor and turbine. The present paper, from both FEM and experimental aspects, studied the vibrating modal characteristics of a flame tube in a can-type combustor. First, the tube FEM model was constructed by Solidworks and analysed in ANSYS Workbench. Then, its first six orders modal parameters, such as frequency and mode, were obtained. Afterwards, the modal experiments were effectuated by the classical hammering method and the resonant frequencies and modes of the tube were identified. Finally, we compared the calculated frequencies in FEM with the experimental ones. The results show that in the calculated modal of the flame tube, the flame tube vibrates along the x and y directions. The periodic tangential vibration with several circumferential waves and few horizontal half waves was observed. The tube's lower order resonant frequencies varied from 145 Hz to 600 Hz, and its higher order frequencies are on the order of 1000 Hz. In the tubes experimental modal, the first order frequency is about 100 Hz, its second and third order vibrating frequencies are about 400 Hz. The last three orders frequencies vary around 1000 Hz. The damping ratio is higher in the first order case than in the other orders. In the FEM and experimental methods, the relative maximum amplitude of the flame tube still occurs at its rear part. Finally, the first three orders frequencies of the tube in the experiment are lower by about 30% than those by the FEM method. This difference probably results from the constraint condition of the swirl section. The calculated last three orders frequencies are consistent with the experimental ones.
Keywords: flame tube; can-type combustor; vibration mode; resonant frequency; FEM; experimental modal analysis method.
New transformation method in continuous particle swarm optimisation for feature selection
by Kangshun Li, Dunmin Chen, Zhaolian Zeng, Guang Chen, James Tin-Yau Kwok
Abstract: Feature selection is a very important task in many real-world problems. Because of its powerful search ability, particle swarm optimisation (PSO) is widely applied to feature selection. However, PSO was originally designed for continuous problems, and therefore, the transformation between continuous particles and binary solutions is needed. This paper proposes a new transformation methods-based PSO (PSOS) in which the related feature subset of a particle is decided by a sine function rather than comparing with a single threshold. To further upgrade the performance of the proposed method, an extra increment generated by the Gaussian distribution is added to the marginal positions (PSOSI). The experimental results show that PSOS and PSOSI can select smaller feature subsets with higher classification accuracy than all the other algorithms compared in this paper. Furthermore, in most cases, the performance of the second method is better than the first one.
Keywords: particle swarm optimisation; feature selection; classification; sine function; Gaussian distribution; transformation method.
Research of small fabric defects detection method based on deep learning network
by Siqing You, Kexin Fu, Peiran Peng, Ying Wang
Abstract: For quality improvement of textile products, fabric defects detection is significant. In this paper, the detection capacity of SSD for small defects was studied. The loss of feature information was reduced through the reduction of layers of SSD network; then the size of the default box was adjusted based on the K-means clustering algorithm, and the adaptive histogram equalisation algorithm was applied to enhance the defect features and effectively improve the detection accuracy. The improved SSD network model was tested to verify the fabric defects dataset, which further improved the accuracy of detection. In addition, the two-stage algorithm was compared to find the optimal algorithm for small object detection. According to the test results, the subsequent improvement method for small object detection with SSD was proposed.
Keywords: fabric defects detection; default box; feature enhancement; SSD; faster RCNN.
Field theory trusted measurement model for IoT transactions
by Meng Xu, Bei Gong, Wei Wang
Abstract: The Internet of Things (IoT) allows the concept of connecting billions of tiny devices to retrieve and share information regarding numerous applications, such as healthcare, environment, and industries. Trusted measurement technology is crucial for the security of the sensing layer of the IoT, especially the trusted measurement technology oriented to transaction IoT nodes. In the traditional trust management system, historical behaviour data are considered to predict the trust value of the network entity, while the nodes' trust between network entities is rarely considered. This paper proposes a novel field theory trusted measurement model of the sensing layer network, which can well adapt to the transaction scenarios of the IoT.
Keywords: field theory; internet of things; trust measurement; transaction scenario.
Effects of fuel pool on temperature profiles of fire in one engine nacelle
by Yicun Chen, Guanbing Cheng, Shuming LI
Abstract: The pool geometries, configuration and position have significant influence on the pool fire behaviour in an enclosed compartment. An attempt is made in the present paper to investigate the effects of a fuel pool on the fire temperature in an engine nacelle. We established the nacelle physical model based on classical turbofan engine CFM56 by AutoCAD, then introduced the Pyrosim to construct its numerical model. Three pool areas, A0404, A0303 and A0202, three pool shapes, S0404, S0208 and S0802, and three pool positions, front, middle and rear locations, were considered. The slice and four detectors were installed in the middle plane vertical to the nacelle longitudinal direction in order to obtain the temperature evolution and cloud charts at the left-right sides and top-bottom of the nacelle. The results indicate that for the pools with different areas, shapes and positions the temperature evolution divides into both an increasing stage within a few seconds and a steady stage, with an oscillation around an average temperature. For the pools with different areas, the increase of the pool areas contributes to the temperature augmentation at the left-right sides or the top and bottom of the nacelle. This change reveals that more fuel participates in the chemical reaction. More combustion heat boosts the fluid temperature in the nacelle by convection and radiation. For the pools with different shapes, the temperature at the left and top sides of the nacelle is higher for the pools farther away from the fire source. But it is always slightly higher at the right and bottom of the nacelle for the pools closer to the inner cylinder. For the pools with different positions, the temperature at the nacelle left and top sides is higher for the front and middle pools. However, it is always higher at the right and bottom of the nacelle for the front and rear pools.
Keywords: engine nacelle; pool area; pool shape; pool position; fire temperature; FDS.
Emotion analysis method for elderly living alone based on CNN-BGRU neural network
by Qingqing Wang, Jianglin Luo, Jianwen Song
Abstract: China has entered a serious ageing society. The psychological needs of the elderly who live alone and need to be accompanied are a common concern of society. On the basis of affective computing technology and deep learning, this paper proposes an emotional analysis method for the elderly who are alone. In the big data environment, their daily emotional changes are analysed and forewarned remotely. However, there are some problems in text classification, such as difficult to extract semantic key features and poor classification effect. Therefore, this paper proposes a hybrid neural network model based on CNN-BGRU to solve the problem of accurate classification. In this algorithm, firstly, the convolution neural network is used to extract the local features of the input text vector, and then BGRU is used to obtain the information before and after this layer, and then the global features are obtained. Finally, the emotion classification results are obtained by Softmax classifier. The experimental results show that the accuracy of the proposed algorithm is 92.8%, the lowest loss rate is 0.2, and the trend is stable. It can be seen that this model can not only obtain more semantic information between texts, but also better capture the dependence of specific emotions in the whole text, so as to more effectively identify the emotional polarity in different aspects of the text.
Keywords: elderly alone; aged-care at home; convolutional neural network; BGRU; emotional analysis; deep learning.
Electromagnetic pulse response prediction of intelligent wireless sensors based on NARX
by Cui Hao, Wenbai Chen, Hao Wu, Changjian Jiang
Abstract: the artificial neural network algorithm can represent all functions at any accuracy through learning the observed data and training parameters. Compared with conventional methods such as analytical methods, which could be limited in accuracy, or numerical modelling methods, which could be time-consuming, the artificial neural network algorithm is attractive for providing fast and accurate answer in the modelling of electromagnetic pulse response prediction of intelligent wireless sensors. According to the characteristics of input and output, nonlinear autoregressive with external input (NARX) neural network was chosen in this paper. It can obtain the current output value depends on its own previous output values and the input values. In order to verify the accuracy of the model, the electromagnetic pulse experiments of intelligent wireless sensors with protection circuit and without protection circuit were done. The results showed that the input-output curve estimated by the NARX neural network model is in good agreement with the experiment results. After two groups of simulation, the NARX model has high fitting ability, which suggests that the NARX model has good generalisation ability.
Keywords: electromagnetic pulse; intelligent wireless sensor; NARX neural network; signal line; transient voltage suppressor.
Flow field simulation and structural parameter optimisation of vacuum adsorption system for textiles fabrics
by Shunqi Mei, Qiao Xu, Zhenghui Wang, Yichuang Gu, Quan Zheng
Abstract: Vacuum adsorption and grabbing for textile fabrics is one key technology for intelligent garment processing. Owing to the softness and air permeability of textile fabrics, the design of the vacuum adsorption grab device has lacked an effective method. In this paper, the standard k-epsilon turbulence model is used to analyse the flow field in the suction cavity of vacuum adsorption device for textile fabrics, the optimization design model of structural parameters is established and solved by the Fluent software, and the verification experiment is carried out. The experimental results show that the suction mechanism with optimised parameters can effectively absorb and grasp the fabric, and the negative pressure required is minimum. The research results show that the structure parameters, such as the thickness of the suction cup cavity, the diameter of the suction hole, and the depth of the suction hole, affect the adsorption performance of the vacuum adsorption device.
Keywords: textile fabric; vacuum adsorption; Fluent simulation; parameter optimisation.
Dynamic time warping-based evolutionary robotic vision for gesture recognition in physical exercises
by Quan Wei, Kubota Naoyuki, Ahmad Lotfi
Abstract: In this paper, we propose a three-dimensional posture evaluating system from two-dimensional images, which can be implemented in physical exercises for elderly people. In this system, two-dimensional coordinates of human joints are first captured and calculated, then our proposed Dynamic Time Warping Steady State Genetic algorithm (DTW-based SSGA) is used for the evaluation of three-dimensional rotational variables from RGB images for the human arm. Finally, these predicted rotational variables would be compared with the template of sample posture by Dynamic Time Warping (DTW) to check the complement of physical exercises. The experimental result shows that our proposed DTW-based SSGA performs with higher accuracy than other evolutionary algorithms, such as standard Steady State Genetic Algorithm (SSGA) and Particle Swarm Optimisation (PSO) when evaluating human joint variables with templates, especially in the physical exercises for rehabilitation.
Keywords: gesture recognition; forward kinematics; evolutionary computing; dynamic time warping.
Research on trusted SDN network construction technology
by Fazhi Qi, Zhihui Sun, Yongli Yang
Abstract: In this paper, we combine trusted computing with SDN. By active measurement of the SDN controller when it is starting and running, we can guarantee the trust of the SDN controller. By actively measuring the behaviour of the SDN data transponder in the domain, we can guarantee trust of the SDN data transponder. When the cross-domain data interaction is involved, by trusted network connection mechanism, we can guarantee the trust of the transmission of data in different domains so as to build a trusted SDN network as a whole.
Keywords: trusted computing; SDN; active measurement.
A research framework for constructing the knowledge database of public security information
by Han Zhong, Shiqiang Zhang, Jianli Liu
Abstract: At present, the public security organs in China have accumulated a great deal of public security data. These data have broad sources, complex structures, and large and increasing scales. How to effectively integrate, manage and mine these data has become a new problem faced by all public security organs. This paper proposes a research framework for constructing the knowledge database of public security information. Based on this multi-dimension framework, data features can be effectively extracted and modelled for improving the management and use of public safety data.
Keywords: public security information; feature extraction; knowledge architecture.
Bi-GRU model based on pooling and attention for text classification
by Hu Yu-lan, Qin-Shan Zhao
Abstract: Aiming at the problems that most of the text classification models based on neural network are easy to overfit and ignore keywords in sentences in the training process, an improved text classification model is proposed. This text classification model is a bi-direction gated recurrent unit (Bi-GRU) model based on pooling and attention mechanism. The model solves the above problems in the following ways. First, the bidirectional gated recurrent unit is used as the hidden layer to learn the deep semantic representation. Second, max-pooling is adopted to extract text features and the self-attention mechanism is adopted to obtain information about the influence of words and sentences for text classification. Third, the model uses the splicing results of the two to classify texts. The experiment chooses two common Chinese datasets, which are Fudan Set and THUCNews, on Pytorch deep learning framework. The experimental results show that the proposed model is better than the Text-CNN model and Bi-GRU_CNN model, such as precision, recall rate and Fscore. Compared with the optimal model, the precision, recall rate and F-score are respectively increased by 5.9%, 5.8%, and 4.6% for Fudan Set, which is the longer Chinese text dataset.
Keywords: text classification; bi-direction gated recurrent unit; max pooling; self-attention mechanism.
A method of spatial place representation based on visual place cell firing
by Naigong Yu, Hui Feng
Abstract: Constructing a model of visual place cells (VPCs), which produce sensitive firing to visual information, is of great significance for studying bionic positioning and bionic navigation. Based on the physiological research of place cells and the analysis of existing VPC generation models, a firing model of VPCs based on the distance perception of landmarks by the agent is proposed in the paper. Based on the firing activity of VPCs, a spatial place representation method is proposed. The method mainly includes exploring the environment and detecting landmarks, calculating the firing rate of VPCs, adding VPCs and constructing the map of VPCs. Through simulation experiments, the reliability of the positioning performance of the proposed method is verified, and the influence of various parameters in the model on the accuracy of spatial representation of the VPCs map is analysed.
Keywords: visual place cell; spatial representation; bionic positioning; bionic navigation.
Five-dimensional model research of complex product assembly driven by Digital Twin
by YuJin Zou, Renwang Li, Xiang Zhang, Jinyu Song
Abstract: This paper describes an analysis of the connotation of process optimization driven by Digital Twin (DT) and puts forward the framework design of a five-dimensional assembly system driven by DT. Based on the assembly framework, the DT technology is constructed based on key features from the exploration of physical space and virtual space. A method for optimising the assembly process is put forward through assembly hierarchy division, process preparation and information collection, and process execution process feedback.
Keywords: Digital Twin; product assembly; information model; process optimisation.
Research and analysis of psychological data based on machine learning methods
by Guangshun Chen, Wei Lv, Junwei Ma, Yanchun Liang
Abstract: The integration of psychology and computer science has become a mainstream contemporary research method on psychological data. Weibo, Chinas largest open platform for communication and information sharing between users, has many emotional contents hidden in its data. According to the current trend, the Weibo data are segmented by machine learning to obtain a psychological portrait of Weibo users. This design uses long- and short-term memory networks (LSTMs) and convolutional neural networks (CNNs) to perform sentiment classification on Weibo data. The classification results are analysed using word frequency analysis and the latent Dirichlet allocation model (LDA) to obtain portraits of Weibo users sentiment and an analysis of the results. The results are displayed in the form of word clouds. According to the clustering results of the word clouds, the main factors affecting different polar emotions can be analysed.
Keywords: recurrent neural network; short-term memory network; convolutional neural network; emotion analysis; LDA.
Point cloud registration algorithm based on 3D-NDT algorithm and ICP algorithm
by Jiangge Huang, Bo Tao, Fei Zeng
Abstract: The purpose of point cloud registration is to minimise the difference of spatial position between point clouds. In addition, the point cloud registration process needs to be performed with high efficiency and accuracy. This paper combines the high efficiency of the 3D normal distribution transformation (3D-NDT) algorithm with the high precision of the iterative nearest point (ICP) algorithm, and proposes a fusion registration algorithm. At the same time, the fusion algorithm can still keep high efficiency and high precision registration. First, the 3D-NDT algorithm is used to select appropriate parameters, so that the point cloud to be registered is closer to the target. It provides an excellent initial position for the ICP algorithm to complete coarse registration. Secondly, in order to improve the efficiency of solving transformation matrix in ICP algorithm, kd-tree is introduced for acceleration. The experimental results show that the fusion point cloud registration algorithm proposed in this paper is better than the 3D-NDT algorithm and the ICP algorithm in efficiency and accuracy. The method proposed in this paper has more obvious advantages in dealing with larger point clouds.
Keywords: 3D-NDT algorithm; ICP algorithm; point cloud registration; point cloud search.
Community-based 3-SAT formulas with a predefined solution
by Yamin Hu, Wenjian Luo, Junteng Wang
Abstract: It is crucial to generate crafted SAT formulas with predefined solutions for the testing and development of SAT solvers because many SAT formulas from real-world applications have solutions. Although some generating algorithms have been proposed to generate SAT formulas with predefined solutions, community structures of SAT formulas are not considered in these algorithms. Consequently, we propose a 3-SAT formula generating algorithm that not only guarantees the existence of a predefined solution, but also simultaneously considers community structures and clause distributions. The proposed 3-SAT formula generating algorithm controls the quality of community structures through controlling (1) the number of clauses whose variables have a common community, which we call intra-community clauses, and (2) the number of variables that belong to only one community, which we call intra-community variables. For a SAT formula, the more intra-community clauses and intra-community variables, the higher the quality of community structures. To study the combined effect of community structures and clause distributions on the hardness of SAT formulas, we measure solving runtimes of two solvers, gluHack (a leading CDCL solver) and CPSparrow (a leading SLS solver), on the generated SAT formulas under different groups of parameter settings. Through extensive experiments, we obtain some noteworthy observations on the SAT formulas generated by the proposed algorithm: (1) The community structure has little or no effect on the hardness of SAT formulas with regard to CPSparrow but a strong effect with regard to gluHack. (2) Only when the proportion of true literals in a SAT formula in terms of the predefined solution is 0.5, SAT formulas are hard-to-solve with regard to gluHack; when this proportion is below 0.5, SAT formulas are hard-to-solve with regard to CPSparrow. (3) When the ratio of the number of clauses to that of variables is around 4.25, the SAT formulas are hard-to-solve with regard to both gluHack and CPSparrow.
Keywords: SAT generator; community structure; predefined solution.
Research on key technologies of intelligent working face in coal
by Peng Chen
Abstract: Mining of a fully mechanised intelligent working face with large mining height (MFMIWF-LMH) is a technology that allows the mining of 3.5~8.8 m thick coal seams at one time using a whole set of fully mechanised mining intelligent equipment. The MFMIWF-LMH is characterised by a complex production process, numerous types of equipment and severe mine pressure. In terms of mine pressure appearance, MFMIWF-LMH is likely to undergo a significant increase in the failure strength of surrounding rock, an obvious rise of the abutment pressure and peak value, and rib spalling and roof fall of coal wall.
Keywords: large mining height; fully mechanised mining; intelligent working face; intelligent mining; unmanned.
Fighting behaviour detection in video using convolutional neural network
by Ying Huang, Ling Lai
Abstract: With the rapid development of computers, networks, camera equipment and image processing and transmission technologies, video surveillance technology has developed rapidly and tends to be intelligent. This paper hopes to use deep learning technology to achieve abnormal detection of fighting behaviour in specific video surveillance scenarios, and then analyse and process more complex human gesture behaviour recognition problems. Through the collection of video data, data preprocessing and model selection, construction, training, improvement, parameter adjustment, testing and other operations, the improved simple convolutional neural network model accuracy rate reached 92.53%, while using migration technology to quote classic. The convolutional neural network structure VGG16 and GoogleNet model can reach 98.71% and 99.60% accuracy.
Keywords: fighting behaviour detection; deep learning; convolutional neural network; transfer learning; video analysis.
Fruit target detection method based on faster R-CNN
by Guanghui Yin, Yuanmin Xie, Juntong Yun, Ying Liu, Nannan Sun, Yongcheng Cao
Abstract: With the rapid development of agricultural modernisation, fruit picking is becoming more and more automatic. The detection of fruit targets by machine vision technology is the key to realise fruit automatic picking. In recent years, with the development of deep learning technology, target detection algorithms based on deep learning have gradually become a hot research topic, and the detection accuracy has been greatly improved. However, the shape and size of fruits in their natural environment are different, and the light intensity changes at any time, which affects the detection accuracy to a certain extent. In this paper, aiming at the problem of fruit detection and location in the natural environment, based on fast R-CNN target detection model, a fruit detection and location method combining image processing and deep learning is proposed. The experimental results show that the combination of image processing and deep learning can achieve high detection accuracy and speed.
Keywords: fruit picking; image processing; faster R-CNN; target detection.
A robot navigation algorithm based on the cognitive mechanism of the hippocampus
by NaiGong Yu, YiShen Liao, YaQian Wei, JianJun Yu, ChunLei Yin
Abstract: The hippocampus formation in the animal brain is the core brain region to realise spatial cognition. Aiming at the problems of inaccurate loop closure detection and lack of effective obstacle perception mechanism in the current bionic navigation model, this paper proposes a robot navigation algorithm based on the hippocampus cognitive mechanism. Firstly, according to the information transmission mechanism of hippocampus formation and the activation theory of hippocampus spatial cells, we construct calculation models of grid cells, border cells, place cells and view cells. Then, the location information, visual information and obstacle information encoded by the spatial cells are fused to build a cognitive map. Finally, according to the obtained cognitive map, we use the A* algorithm to obtain a relatively shorter path and complete the goal-oriented navigation task. The method is verified by simulation experiments, proving its effectiveness and reliability.
Keywords: spatial cells; cognitive map; hippocampus formation; path planning; goal-oriented navigation.
Nonlinear dynamic characteristics analysis of multilayer electrostatic micro-beam structure
by Xiangjuan Bian, Jinlai Qi, Youping Gong, Huipeng Chen
Abstract: The dynamic analysis of the multi-layer micro beam structure is of great significance to the structural design of MEMS devices. Firstly, the dynamic equation mathematical model of electrostatic multi-layer cantilever beam was established based on the energy principle and fluid film damping effect; second, the dynamic model was transformed into the single-layer beam solution model by equivalent parameter method. Thirdly, the Galerkin mapping reduction algorithm was used to solve this model. The theoretical computing results and simulation results can provide strong support for the study of the working motion state of MEMS devices.
Keywords: MEMS; multi-layer cantilever beam; model reduction; dynamic characteristics analysis.
Simulation analysis on the public opinion factors and public panic degree under the background of spreading sudden disaster information by new media
by Zixia Chen, Shiwen Wu, Zelin Chen, Bingqian Lv
Abstract: Nowadays, it is popular to use various chatting software such as WeChat, Twitter, QQ, Micro-Blogs, and other different social media means to communicate with each other in real-time. However, if a sudden public disaster is not managed properly on the social media platforms in time, the negative energy and information can be easily spread out quickly, resulting in massive public panics, such as the public panic that occurred during the COVID-19 outbreak in late 2019. Based on the analysis of the transmission and evolution mechanism of the sudden disaster risk, this paper fully considers the influence of public opinion after a disaster and uses the improved individual interaction model to transform public opinion information into public risk panic. In addition, the paper puts forward corresponding emergency rescue measures according to the panic characteristics to reduce the risk of social unrest. The Matlab-ABM simulation results show that at the early stage of a disaster, the degree of public panic risk evolves with the number of interactions between individuals and the deepening of risk awareness. Further, different disaster levels also lead to varying influences on the panic risk. Positive information and negative information will produce more significant distinctions in the degree of public risk panic. The simulation results can provide decision support for relevant government departments to restrain the fermentation of sudden public disasters.
Keywords: new media communication; sudden disaster; public opinion factors; the evolution of public panic; numerical simulation.
Application of deep learning in network security fault diagnosis and prediction
by Jing Wang, Fangfang Liu, Hongyan Liu, Qingqing Wang
Abstract: At present, deep learning method has been successfully applied in many application directions, but few researchers try to apply deep learning to network security fault diagnosis. This paper summarises the deep learning methods applied to network security fault diagnosis and prediction, and focuses on the attack detection using stacked automatic encoder. The network datasets are used to compare various attacks. The fault diagnosis process based on the deep learning method and the analysis and verification of the experimental results are introduced in detail. At the same time, the automatic operation time is implemented in order to monitor and predict the network application characteristics and deep learning mechanism, intrusion detection system can be used to monitor network applications and send out an alarm when an attack is detected.
Keywords: deep learning; network security; fault diagnosis; automatic encoder.
Determination of SSC and TA content of pears by Vis-NIR spectroscopy combined CARS and RF algorithm
by Baishao Zhan, Xu Xiao, Fan Pan, Wei Luo, Wentao Dong, Peng Tian, Hailiang Zhang
Abstract: Soluble solid content (SSC) and total acid (TA) are the indicators of fruit maturity and taste, which impact pear fruit quality. Therefore, it is of great significance for pear quality grading to quickly and accurately detect soluble solids and total acids. This work focuses on the visible and near-infrared spectroscopy measurement model. Savitzky-Golay (SG) smoothing, standard normal variable (SNV), and multiplicative scatter correction (MSC) were used to eliminate the error effects. Competitive adaptive weighted sampling (CARS) and random frog (RF) algorithms were used to select the characteristic wavelength spectrum to eliminate redundant information and improve measurement speed and accuracy. Partial least squares regression (PLS) model and multiple linear regression (MLR) models were built to verify the preprocessing method's performance and prediction model. The results show that SG smoothing had the most significant effect on the error elimination of the original spectra, the CARS-PLS model has the best prediction effect on SSC, R2 is 0.9012, CARS-MLR model is the best predictive performance of TA, and R2 is 0.8557. Research shows that Vis-NIR spectroscopy as a method to detect SSC content and TA in pear fruit has potential application value.
Keywords: visible and near-infrared spectroscopy; soluble solids; total acid; pears; competitive adaptive reweighted sampling; random frog.
Research on collaborative optimisation of urban agricultural product distribution centre location and routing based on improved adaptive large-scale neighborhood search algorithm
by Jingru Huang, Wei Zhang
Abstract: This paper establishes an optimisation model of limited fresh agricultural products distribution centre location considering fuel consumption, exhaust emission and multiple constraints, in order to minimise the total cost of distribution centre location. In the paper, an improved adaptive large-scale neighborhood search is designed to solve this collaborative optimisation problem, and the model simulation test is compared with the traditional CW saving algorithm under different calculation examples. The results clearly show that the proposed model is closer to the reality of agricultural products distribution, and the proposed adaptive large-scale neighborhood search algorithm is effective.
Keywords: location-routing collaborative optimisation problems; adaptive large neighborhood search; agricultural products.
Research on model predictive trajectory following control of automatic vehicle considering prediction error
by Xingyu Ye, Shaopeng Zhu, Sen Chen
Abstract: Trajectory following control system is one of the key components of autonomous driving systems. A nonlinear model predictive control scheme considering prediction error is proposed in this paper to regulate active front steering and enable an automatic vehicle to stably follow a predefined trajectory with good lateral stability and ride comfort at high velocity. A feedback compensation strategy is used to deal with the model mismatch existing between the controlled vehicle and the nominal model. Moreover, the effectiveness of the scheme is verified through the co-simulation between the software MATLAB/Simulink and Carsim. The results reveal that the proposed model predictive control with feedback compensation can effectively regulate the front steering angle and enhance the vehicle dynamics performance at high velocity.
Keywords: trajectory following; autonomous vehicle; model predictive control.
Numerical analysis of eddy current loss of high-speed axial magnetic drive spindle
by Qiao Xu, Tao Yang, Yuchen He, Shunqi Mei, Fanhe Meng, Xuemei Tang
Abstract: Eddy currents will generate in the spindle baseplate to cause energy loss when the axial magnetic force drives the spindle to rotate. Based on the magnetic field theory, the eddy current loss of the outer spindle baseplate of the axial magnetic drive spindle is analysed and calculated in the paper, and the mathematical model is established. The eddy current loss calculation method of axial magnetic drive spindle is proposed, and the factors affecting eddy current loss are analysed by finite element method. According to the analysis results, the measures is presented to reduce the eddy current loss: under the condition of meeting the working strength, the outer spindle baseplate should be made of the materials with small conductivity and permeability, and the thickness of the outer spindle baseplate should be as small as possible. The analysis results provide theoretical support for the optimization design and energy consumption reduction of axial magnetic drive spindle.
Keywords: numerical analysis; eddy current loss; magnetic drive; spindle.
Research on leak detection and location of urban gas pipeline network based on RSSI algorithm
by Liming Wei
Abstract: To solve the leakage problem of urban gas pipelines, this paper presents a method of detecting and locating leakages based on the RSSI algorithm. This technique can analyse and calculate the signal strength received between ZigBee nodes when a pipeline leaks and ultimately obtain the location of the leak. Firstly, the algorithm model is established by using the RSSI signal strength values between the leak target point and each receiving point. Secondly, the distance between the leak point and each receiving point is obtained by the model. Lastly, the approximate coordinates of the leak point are obtained by the least squares method. The simulation results show that the proposed algorithm has high positioning accuracy and wide application prospects.
Keywords: gas pipeline network; fire early warning; least squares method; RSSI algorithm; ZigBee technology.
A self-cooperative trust scheme against black hole attacks in vehicular ad hoc networks
by E.A. Mary Anita, Lakshmi Sabarinath, J. Jenefa
Abstract: The main objective of the Vehicular Adhoc NETwork (VANET) is to provide secure communications for the vehicles in the network without fixed infrastructures. It inherits all the properties of the MANET. Achieving reliable routing to avoid various routing attacks is the major concern in the vehicular network. Routing attacks degrade the performance of the network. Black hole attack is one of the routing attacks, which drops the data packets without forwarding them to the destination vehicle. Different routing schemes have been proposed to provide security against these attacks, but they still have security issues. Hence a new self-cooperative trust scheme is proposed in this paper to detect single as well as collaborative black hole attackers in the network. Two processes, self-detection and cooperative detection, are used to detect attackers in the network. The performance of the proposed scheme is analysed. Results show that the proposed scheme has better performance in terms of throughput, PDR and delay.
Keywords: AODV; black hole attack; cooperative detection; self-detection; trust information; VANET.
Secure hybrid satellite-UWOC cooperative relaying system under malicious unmanned aerial vehicle eavesdropper threat
by Kehinde Odeyemi, Pius Owolawi
Abstract: In this paper, the secrecy performance of a hybrid satellite-underwater optical communication (UWOC) system in the presence of an unmanned aerial vehicle (UAV) eavesdropper is investigated. The satellite and eavesdropper radio frequency (RF) links are respectively subjected to Shadowed-Rician and Nakagami-m fading distributions, whereas the UWOC link experiences mixed Exponential-Gamma distribution under difference detection schemes such as heterodyne detection and intensity modulation with direct detection (IM/DD). Specifically, the equivalent cumulative distribution function (CDF) closed-form expression of the concerned system is then obtained. Based on this, the analytical closed-form expressions of the system connection outage probability (COP), secrecy outage probability (SOP), strictly positive secrecy capacity (SPSC) and average secrecy capacity (ASC) are derived. By our findings, it is found that the satellite shadowing effect, air bubbles level and temperature gradients, and eavesdropper distance significantly have impact on the system secrecy performance. The result also illustrated that the heterodyne detection outperforms the IM/DD under the system and channel conditions. Finally, the accuracy of the analytical expressions is justified through Monte-Carlo simulations.
Keywords: underwater optical communication; satellite communication; cooperative relaying technique; unmanned aerial vehicle; physical layer security.
Auction theory based resource allocation in heterogeneous cellular network for 5G
by Gitimayee Sahu
Abstract: In this research work, we propose an auction mechanism to motivate the small cells to offload the traffic for the cellular network operator (CNO). The Macro Base station (MBS) is the auctioneer (i.e. dealer) and the small base stations (SBS) are the bidders (i.e. purchaser) and the UEs are the commodities. Further, the MBS chooses the SBS which will give the highest signal to interference plus noise ratio (SINR) for offloading the MUEs. The SBSs contest among each other to offload the MUEs. The MBS declares the introductory price consenting to the offer to the SBSs. According to the price paid, the SBS decides its supply capacity. The supply includes resources offered by the SBS to offload a number of UEs from MBS to maximise the benefit. The MBS's utility is the difference between the profit earned by saving the power consumption and the total expenditure that MBS has to pay to all the SBSs. The preserving power consumption leads to a gain of energy efficiency (EE) of the network. If the supply is insufficient then the requirement, then the MBS enhances the effective price so that the SBS are encouraged to raise their supply amount. The iterative auction progresses till the total supply equals to the demand. A Pareto optimal trade-off is established between the price for offloading a unit of traffic to maximise its profit. The CNO decides the amount of traffic needs to be offloaded to maximize its utility. The difference between the satisfaction of the UEs and the price paid to the CNO for offloading need to be minimal.
Keywords: traffic offloading; small cell; auction mechanism; energy efficiency; cost efficiency; Pareto optimality.
Multi-objective workflow scheduling in the cloud environment based on NSGA-II
by Tingting Dong, Chuangbai Xiao
Abstract: The emergence of cloud computing offers a novel perspective to solve large-scale computing problems. Workflow scheduling is a major problem in the cloud environment, and parallelism and dependency are two important characteristics of tasks in a workflow, which increases the complexity of problem. Workflow scheduling is also a multi-objective scheduling problem, and task execution time and cost are the two extremely significant goals for users and providers in the cloud environment. Existing heuristic algorithms are popular, but they lack of robustness and need to be revised when the problem statement changes. Evolutionary algorithms have a complete algorithm system, which is widely used in the multi-objective scheduling problem. In this paper, Nondominated Sorting Genetic Algorithm-II (NSGA-II) is utilized to solve the workflow scheduling problem aiming at minimizing the task execution time and cost. Some real-world workflows are used to make simulation expriments, and comparative simulations with genetic algorithm are given. Results show that NSGA-II is effective for the workflow scheduling.
Keywords: cloud computing; workflow scheduling; non-dominated sorting genetic algorithm-II; multi-objective scheduling.
System modelling and empirical study on 3PL location in a multi-node supply chain network
by Zixia Chen, Zelin Chen, Yujie Geng
Abstract: The multi-node supply chain network composed of suppliers, logistics centres, transfer stations, retailers and other different types of enterprise node is the research object in this paper. Starting from the demand-drive of retailers, a mathematical model of the third-party logistics centre location problem was established in a multi-node supply chain network according to the logical design of bottom-up supply chain network planning. The improved genetic algorithm was adopted to solve the problem and the analysis of an actual case was carried out. The empirical analysis results demonstrate that the modelling and the algorithm in a multi-node supply chain network are reasonable and effective. The optimal location of different logistics nodes can be obtained, and the corresponding total system cost is the lowest. This paper has improved algorithm and technological innovation in the field of logistics location for the multi-node supply chain network.
Keywords: multi-node supply chain network; 3PL location problem; model and algorithm; empirical analysis.
Application of improved deep reinforcement learning algorithm in traffic signal control
by Wang Qiang, Song Shuaidi, Zhang Tengyun, Wang Zelin
Abstract: For a regional road network, the signal control system lights at multiple intersections belong to the core technology of intelligent innovation. The relevant personnel need to integrate and analyse the relevant status information of the intersection area based on the DQN method and strategy, and strengthen the control of the signal light system effect, to achieve fast and effective detection. In this paper, we propose a reinforcement learning DQN+ algorithm by using the improved DQN reward and punishment function. Experiments show that DQN+ has obvious advantages in terms of average queue length (AQL), average speed (AS) and average waiting time (AWT) at four intersections.
Keywords: intelligent transportation; traffic signal control; reinforcement learning; deep reinforcement learning.
Analysis of wireless sensor networks based on non-exhaustive M/G/1 queueing model
by Zhanyou Ma, Xiangran Yu, Shanshan Guo, Jiaqi Fan
Abstract: In order to reduce the energy consumption of wireless sensor networks (WSNs) and control the workload of necessary topology maintenance, the sleep mode in the energy saving strategy is considered. Combining with practice, factors such as environmental interferences, physical damages, signal differences of packets and possible errors in the transmission of data packets are considered. A non-exhaustive M/G/1 repairable queueing model with negative customers, preemption, feedback and J optional random vacations is established. Sufficient and necessary condition for the existence of steady state of the system is derived by constructing Markov process with absorption state. Using the supplementary variable method, the expressions of the average number of the data packets and other performance indicators are obtained. Then, using MATLAB software for numerical analysis, the influence of system parameters on the average delay and other performance indicators is analysed. Finally, the social optimal parameters are found by constructing benefit functions. The WSNs model proposed in this paper is also compared with the normal model, and it found that the model with Type j optional sleep has a good energy-saving effect.
Keywords: wireless sensor networks; fault repair; sleep mode; queueing model; supplementary variable method.
Research on ceramic tile defect detection based on YOLOv3
by Gongfa Li, Xin Liu, Bo Tao, Du Jiang, Fei Zeng, Shuang Xu
Abstract: Artificial intelligence is a technology that studies, simulates and expands human intelligence theory and related methods. It is the direction of modern and future science and technology development. The teaching methods of artificial intelligence courses are supposed to be different from the traditional teaching methods, but the actual investigation finds that there are still some problems in these courses, such as the single teaching mode, the low enthusiasm of students for studying, and the poor practical ability of students. In order to solve these issues, this paper applies project teaching methods to an artificial intelligence course, through a specific tile defect detection project to analyse. YOLOv3 algorithm is used to detect six kinds of tile defects, and the experimental results are analysed.
Keywords: defect detection; artificial intelligence; YOLOv3 algorithm; project-based teaching.
Application of an improved physarum polycephalum algorithm to a QoS routing problem
by Zhang Yi, Yang Zhengquan
Abstract: With the development of internet technology, network information dissemination has higher requirements for Quality of Services (QoS). Traditional computing methods can no longer meet the requirements of real networks, and ordinary swarm intelligence algorithms need a large number of training sets to find the optimal parameter combination. In this paper, we propose a novel Physarum Polycephalum Model of QoS (PPMQ). We change destination nodes in the PPMQ set. The corresponding flow matrix and the source nodes flow are improved. The improved algorithm does not need to adjust parameters compared with other swarm intelligence algorithms. The simulation results show that PPMQ is feasible to calculate the Quality of Services Routing Problem (QOSRP) without finding the optimal parameter combination.
Keywords: QoS; QoS routing problem; physarum polycephalum.