Loading...

Table of Content

    15 August 2023, Volume 37 Issue 7 Previous Issue    Next Issue
    Vehicle engineering

    Overview of the application of artificial intelligence technology in intelligent networked new energy vehicles

    2023, 37 (7):  1-15. 
    Abstract ( 711 )   PDF (2282KB) ( 872 )   Save
    The development of intelligent networked new energy vehicles is a long-term national strategy in China. This paper focuses on the application of artificial intelligence technology in intelligent networked new energy vehicles. Firstly, the definition of intelligent networked new energy vehicles is given, and their intelligent application is pointed out as an important feature. Then, the representative intelligent applications of the existing artificial intelligence technologies applied in automotive intelligent, network connection and new energy systems are summarized, compared and analyzed. Besides, the technical challenges faced in the existing automotive intelligent applications are pointed out. Finally, the trend of technology development in this field is discussed.
    Related Articles | Metrics

    Design and testing of the detection system of off-hand automatic steering wheels

    2023, 37 (7):  16-24. 
    Abstract ( 275 )   PDF (3760KB) ( 327 )   Save
    Unmanned driving represents the development goal of automotive technology in the future, but the current level of automated driving cannot support the driver to completely take off. In order to ensure the safety of driving, it is necessary to detect the state of a driver holding the steering wheel in real time. Therefore, according to the automatic driving standards of various countries and industries, this paper proposes a steering wheel off-hand detection method based on tactile sensing for the off-hand steering wheel detection function in the field of human-computer interaction and risk management and control of intelligent connected vehicles. Flexible micro-nano bionic 3D array pressure sensors, high-frequency signal acquisition and multi-signal fusion control technology are adopted, and a set of hands-off detection and testing system is developed, which integrates signal acquisition, processing, calculation, result display and driving simulation. The off-hand detection system is tested under the simulated driving state based on the simulated driver. The results show that the off-hand detection system can realize high-efficiency and high-precision recognition of the two states of the driver’s holding the steering wheel and off-hand driving in multiple driving scenarios and with complex steering wheel holding postures.
    Related Articles | Metrics

    Research on multi-objective signal timing optimization for improved artificial fish swarm algorithm

    2023, 37 (7):  25-33. 
    Abstract ( 105 )   PDF (2910KB) ( 119 )   Save
    Due to the impact of signal timing schemes on the traffic efficiency of intersections, in order to improve the operation status of intersections, this paper constructs an intersection signal timing optimization model with the optimization objectives of vehicle delay, number of stops and delay imbalance, and uses an improved artificial fish swarm algorithm to solve the problem. It introduces attenuation functions in the improved algorithm to obtain variable step size and variable field of view. By combining the crossover and mutation operations genetic algorithm in artificial fish movement strategy, a cloud model is used to generate new crossover and mutation probabilities to optimize crossover and mutation operations. Finally, a certain intersection in Chongqing is selected as a case study, and VISSIM software is used for experimental simulation to compare and analyze the impact of the current timing scheme and the optimized scheme on the operational efficiency of the intersection. The results show that the application of this method can reduce vehicle delay, parking times and queue length, effectively improving the traffic efficiency of road intersections.
    Related Articles | Metrics

    Lane potential field-based lane keeping strategy with model predictive control

    2023, 37 (7):  34-43. 
    Abstract ( 133 )   PDF (3787KB) ( 187 )   Save
    As an important part of advanced assisted (ADAS) driving, lane keeping control is significant to reduce driving fatigue and improve driving safety. Based on lane line potential field design and model predictive control, this paper proposes a model predictive lane keeping control method that integrates lane line potential field to optimize driving stability and safety during lane keeping. The co-simulation of the proposed algorithm is carried out under Carsim&Simulink. The simulation results show that, compared with PID control, the lane keeping algorithm based on model predictive control has obvious advantages in both tracking accuracy and vehicle stability, and the lateral tracking accuracy and vehicle stability are significantly improved. Furthermore, compared with the lane keeping MPC without the fusion of the lane line potential field, the proposed method has better vehicle stability and traffic efficiency (27.8 km/h vs 26.6 km/h), and its maximum yaw rate and maximum lateral acceleration decrease by almost 10%. The simulation results show the effectiveness and superiority of the proposed algorithm, which has good engineering guiding significance.
    Related Articles | Metrics

    Research on lane depth perception of autonomous vehicles with dual attention mechanism

    2023, 37 (7):  44-50. 
    Abstract ( 160 )   PDF (3280KB) ( 169 )   Save
    In order to solve the problems caused by lane segmentation, such as heavy computation, weak fusion effect, occlusion, loss and misrecognition, this paper designs a lightweight convolutional neural network structure based on semantic segmentation to introduce the channel attention mechanism, as well as row and column attention mechanism into the network. A lightweight training network ResNet-18 is frstly used to rapidly downsample the input images to generate multi-stage feature maps. Then, the channel attention mechanism is applied to higher-order feature maps to extract higher-order semantic information. The row and column attention mechanism is applied to the low-order feature map to extract the spatial information of the lane lines. Furthermore, the feature fusion mechanism FFM is used to sample the high-order feature map and get fused with the low-order feature map to improve the segmentation accuracy of lane lines. A three-layer fully connected network is constructed to predict the categories of the segmented pixels, which replaces the traditional clustering method, classifies the background and the lane lines, and enables the whole network to get end-to-end training and output. The lightweight codec network model is trained and tested on Tusimple data set for lane detection, and later compared with previous research models. The results show that the designed deep convolutional network can still accurately and quickly identify lane lines in the case of lane lines with occlusion, blurring, shadow interference and exposure. Compared with the existing lane detection model, the segmentation accuracy and detection speed are improved, which can meet the requirements of real-time detection of automatic driving.
    Related Articles | Metrics

    Trajectory tracking of unmanned vehicles based on improved active disturbance rejection compensation

    2023, 37 (7):  51-61. 
    Abstract ( 157 )   PDF (3310KB) ( 130 )   Save
    Aiming at the problem that the tracking accuracy of unmanned vehicles in trajectory tracking control decreases due to imprecision of the model and internal and external interference, this paper proposes a control algorithm combining model predictive control (MPC) and improved active disturbance rejection control compensation. Firstly, an MPC controller is established. In order to improve the real-time performance of the calculation, a nonlinear model predictive control is transformed into linear time-varying model predictive control. Due to trajectory deviations caused attributed to the inaccuracy of the model and internal and external disturbance, a smoother nonlinear function is constructed, and then the improved active disturbance rejection control is designed. With the lateral deviation and the angular deviation tending to be zero as the goal to compensate the front wheel angles of the vehicle, this paper enhances the unmanned vehicle trajectory tracking performance and anti-jamming performance. The simulation results under different working conditions show that this method effectively completes trajectory tracking with small tracking deviations, and has strong robustness to external interference and vehicle parameter change.
    Related Articles | Metrics

    Research on path planning for front parking based on spline theory

    2023, 37 (7):  62-69. 
    Abstract ( 121 )   PDF (2440KB) ( 102 )   Save

    With the continuous increase in the number of electric vehicles in China and the increasingly narrow parking spaces in cities, how to park safely has become one of the problems that troubles the people. Traffic accidents caused by parking errors are also common in traffic accidents. As an important component of autonomous driving, automatic parking has always had excellent market prospects and is also a hot research topic in the field of autonomous driving and automotive intelligence technology.

    At present, automatic parking has relatively complete technological development, but it still has the following obvious drawbacks. Firstly, the existing automatic parking systems are all based on reversing, while electric vehicles in China have been in a high-speed development stage in recent years. For many electric vehicles, the charging ports are installed at the front of the vehicle, which will lead to the problem of charging cables being too short to charge when the driver operates reverse parking. Even if the charging cable is long enough, there is a risk of safety hazards caused by excessive stretching of the cable. Therefore, the existing parking methods based on reversing are not suitable for some electric vehicles. It is necessary to develop a parking method based on front parking. Secondly, the existing automatic parking systems do not comprehensively consider vehicle steering, braking constraints and initial dynamic posture conditions in the path planning process. This can lead to the need for in situ steering in the planned path, which increases the wear of the tires and the steering equipment. Additionally, there are potential problems in the planned reference path, such as curvature discontinuity and inconsistency with the initial posture characteristics of real vehicles. These potential issues may all affect the tracking performance of actual vehicles.

    Aiming at the problems in the existing parking system, this paper designs a new parking method. The main contents are as follows. Aiming at the characteristics of the charging ports of most new electric vehicles installed in the front of the vehicle, a new parking method based on the constraint optimization model of polynomials and the B-spline curve is proposed to optimize the front parking of a vehicle. This paper analyzes the trajectory curves of the forward and backward stages during the parking process, and calculates the parking point and attitude of the car during the transition between the two stages. Based on the possible collision points during the parking process, as well as the vehicle parameter constraints such as steering and braking constraints, a polynomial multi-constraint equation for the parking planning path is established, effectively changing potential issues such as a mismatch of the initial pose characteristics of the vehicle with the pose characteristics of the starting point of the planning path. In order to ensure that the planned path can meet the continuous change of the vehicle steering wheel speed, it is necessary to optimize the parking path through four fifth order B-spline curves to solve the curvature discontinuity problem of the parking path, reduce the parking time and improve the parking efficiency. This paper proposes to use Matlab software to obtain trajectory control points using nonlinear constraint optimization functions, and obtain trajectory curves that meet the constraint conditions. The experimental results show that the path planned by combining spline theory and multi-constraint equations can achieve collision free parking and positioning of vehicles, meet the requirements of curvature continuity, and avoid problems such as turning in place.

    Related Articles | Metrics

    Obstacle avoidance path planning for intelligent vehicle sampling area optimization

    2023, 37 (7):  70-79. 
    Abstract ( 118 )   PDF (4965KB) ( 116 )   Save
    Aiming at obstacle avoidance path planning for intelligent vehicles on structured roads, this paper proposes an obstacle avoidance path planning method based on sampling area optimization. Fully considering the road environment and obstacle vehicle information, it establishes an obstacle vehicle expansion ellipse layer model to divide the risk of the road environment. The obstacle avoidance process is divided into three stages: changing lanes to avoid obstacles, going straight after changing lanes, and returning to the global reference route. In each stage, low-risk sampling points are selected to generate candidate paths through the expansion elliptic layer model. Considering path comfort and global path tracking abilities, this paper designs the comfort and offset cost functions, and combines the expansion ellipse layer model to establish the safety cost function and select the optimal path. In order to test the effectiveness of the method, the obstacle avoidance simulation and real vehicle verification are carried out by constructing straight roads and curve scenes. The research results show that under different scenarios, the proposed method can effectively avoid collision with static and dynamic obstacle vehicles, and efficiently plan safe and smooth driving paths.
    Related Articles | Metrics

    Vehicle object detection methods based on improved YOLOv5s

    2023, 37 (7):  80-89. 
    Abstract ( 248 )   PDF (4473KB) ( 248 )   Save
    Aiming at the problem of missing detection of small target vehicles in autonomous driving, this paper proposes an improved vehicle detection algorithm based on YOLOv5s. The algorithm adopts the weighted bidirectional feature pyramid network (BiFPN) fusion method, which can enhance the fusion of different levels of information while preserving more shallow semantic information. It also introduces multiple self-attention mechanisms into the backbone network to improve the feature extraction capability. The experimental results show that, compared with the unimproved YOLOv5s model, the mean average precision (mAP) of the improved network model increases by 1.01%. Its detection speed meets the real-time requirements, and it can effectively detect small target vehicles under different lighting conditions.
    Related Articles | Metrics

    Autonomous overtaking trajectory planning and tracking control on two-way two-lane roads

    2023, 37 (7):  90-100. 
    Abstract ( 121 )   PDF (2843KB) ( 163 )   Save
    This paper studies overtaking trajectory planning and tracking control of autonomous ground vehicles on a scene of two-way two-lane roads. Firstly, an objective function is established to balance the stability and efficiency of lane changing, and the optimal quintic polynomial lane changing trajectory is obtained. Then, the overtaking process is divided into three stages: lane changing, overtaking and lane merging. A safety distance model is established based on the critical collision conditions of the overtaking vehicle and other vehicles so as to judge the feasibility of the overtaking, obtain the duration of the overtaking, and plan the complete overtaking trajectory. After that, a model predictive controller and a discrete sliding mode controller are designed to track the lateral path and longitudinal speed respectively, and fuzzy control is used to compensate for the front steering angle output by the model predictive controller. Finally, overtaking verification is carried out under the joint simulation environment of Matlab and CarSim. The results show that the proposed overtaking trajectory planning and tracking method enable the vehicle to achieve autonomous overtaking maneuvers and maintain good tracking accuracy and stability.
    Related Articles | Metrics
    “Precision Engineering Measuring Technology and Instrument” Special Column

    Intelligent diagnosis of complex bending and torsional coupling faults of unbalanced rotor systems

    2023, 37 (7):  101-109. 
    Abstract ( 108 )   PDF (3662KB) ( 72 )   Save
    The coupling of bending vibration and torsional vibration often exists in the actual operation of rotating machinery. This paper considers the bending and torsional coupling of unbalanced rotors under different complex conditions, and utilizes the advantages of deep learning technology to construct a diagnosis model based on one-dimensional convolutional neural networks. An intelligent fault diagnosis method for handling the bending, torsion and bending torsional coupling vibration of the unbalanced rotors is proposed. The influence of data input type and L2 regularization on the diagnosis is analyzed, and the diagnosis model is optimized to improve the diagnosis accuracy. The research results indicate that this method can realize intelligent diagnosis of single or multiple composite faults when bending torsional coupling vibration occurs at different speeds, and achieve better diagnostic results than other methods.
    Related Articles | Metrics
    "Intelligent robot perception, Planning and Application Technology" special column

    Deep down-sampling methods for local fault diagnosis of rolling bearings

    2023, 37 (7):  110-119. 
    Abstract ( 118 )   PDF (5025KB) ( 88 )   Save
    The healthy state of bearings is very important for the normal operation of rotating machinery such as radar driving structure and helicopter transmission mechanism. Aiming at the characteristics of complex working conditions, noise, and insufficient and unbalanced samples of the fault labels of the vibration signals of rolling bearings, this paper proposes an improved one-dimensional convolution neural network fault diagnosis method for rolling bearings based on deformable convolution of the disturbance training samples and depth residual block structure. The deformable convolution is set to improve the ability of extracting local fault features, and the improved depth residual block is introduced to improve the generalization ability and sensitivity of the model to the training data. When the training data are fed, the training disturbance layer is set to add disturbance samples to improve the robustness of the model. The Case Western Reserve University bearing data set is used as the experimental data set to divide the training set and the test set. The experimental results prove the effectiveness of the proposed method. TD-DCCNN algorithm can still achieve an average accuracy of 90.35% when the signal-to-noise ratio is 0, which has certain advantages compared with other diagnostic algorithms.
    Related Articles | Metrics

    Error analysis and optimization of multi-field coupling absolute time-grating angular displacement sensors

    2023, 37 (7):  120-128. 
    Abstract ( 84 )   PDF (5925KB) ( 207 )   Save
    In response to the problem that traditional absolute angular displacement sensors overly rely on encoding technology and ultra-precision machining methods, this paper proposes a multi-loop angular displacement sensor with coupled multiple physical fields by using a combination of “positioning+measurement”. At the same time, the error characteristics of the sensor are analyzed through theoretical error derivation, intra-pole simulation error analysis and experimental error research. The sensor structure optimization plan is also proposed. Two sets of prototypes before and after the sensor optimization are made by using PCB technology and the experimental research is carried out. The results show that the measurement accuracy of the optimized sensor system improves from ±16.5″ to ±6.2″. This structure can fully utilize the advantages of the strong anti-interference ability of time-grating magnetic field and high accuracy of time-grating electric field, and has important application value. The sensor achieves the measurement of spatial quantities by measuring time quantities, and has excellent performance, especially suitable for use under complex working conditions.
    Related Articles | Metrics

    Disturbance observation and compensation design of SOTM antenna

    2023, 37 (7):  129-134. 
    Abstract ( 85 )   PDF (1892KB) ( 55 )   Save
    The difficulty of SOTM antenna is to suppress the disturbance of antenna pointing accuracy caused by carrier movement. Compensation for disturbance can significantly improve the dynamic performance of antenna. The traditional control method realizes the measurement of the disturbance angular velocity with a rate gyro, and it’s difficult to accurately obtain the disturbance on the antenna rotation axis. In this way, limited improvement effect occurs, as only a certain proportion is compensated into the speed loop input. Therefore, this paper proposes a disturbance compensation method based on a state observer. A disturbance model is obtained by collecting the angular velocity of the inertial navigation system when the antenna is working normally on a typical road surface, and approximating the disturbance spectrum by using a shaped filter. A disturbance observer model is designed to estimate the angular velocity of the disturbance, and the test is carried out on the vehicle-mounted mobile antenna. The test results show that, compared with the traditional disturbance feedforward compensation, the proposed method reduces the angular error peak by 13% and RMS by 43%, which can better adapt to the tracking road conditions.
    Related Articles | Metrics

    Fault diagnosis methods of rolling bearings based on TD-DCCNN

    2023, 37 (7):  135-143. 
    Abstract ( 93 )   PDF (2840KB) ( 107 )   Save

    The healthy state of bearings is very important for the normal operation of rotating machinery such as radar driving structure and helicopter transmission mechanism. Aiming at the characteristics of complex working conditions, noise, and insufficient and unbalanced samples of the fault labels of the vibration signals of rolling bearings, this paper proposes an improved one-dimensional convolution neural network fault diagnosis method for rolling bearings based on deformable convolution of the disturbance training samples and depth residual block structure. The deformable convolution is set to improve the ability of extracting local fault features, and the improved depth residual block is introduced to improve the generalization ability and sensitivity of the model to the training data. When the training data are fed, the training disturbance layer is set to add disturbance samples to improve the robustness of the model. The Case Western Reserve University bearing data set is used as the experimental data set to divide the training set and the test set. The experimental results prove the effectiveness of the proposed method. TD-DCCNN algorithm can still achieve an average accuracy of 90.35% when the signal-to-noise ratio is 0, which has certain advantages compared with other diagnostic algorithms.


    Related Articles | Metrics

    Fault diagnosis of gearboxes based on multi-sensor signal fusion and residual neural network

    2023, 37 (7):  144-152. 
    Abstract ( 135 )   PDF (4564KB) ( 120 )   Save
    In order to solve the problems that the fault signal is easy to be flooded by strong noise, the collected signal is not comprehensive and the training network is complicated, this paper introduces the residual neural network with multi-sensor signal fusion and attention mechanism into gearbox fault diagnosis. Firstly, the signals collected by multiple sensors are fused based on the variance contribution rate of the vibration signals to obtain more comprehensive fault information of a gearbox. Then, a time-frequency diagram of the signal is obtained by wavelet transform, and the two-dimensional time-frequency information of the fault signal is constructed. Finally, the residual neural network (ResNet) with local cross-channel interaction strategy (ECA module) is used to learn and classify different fault states. After the global channel-level average pooling without reducing dimension, the classification effect is obviously improved. Through the identification and analysis of the gearbox fault signals under different fault types, different signal-to-noise ratios and different working conditions, and compared with different diagnosis methods, it is proved that the proposed method is feasible and a fast recognition rate.
    Related Articles | Metrics

    Design of absolute angular displacement time-grating measurement system

    2023, 37 (7):  153-160. 
    Abstract ( 74 )   PDF (4077KB) ( 81 )   Save
    Aiming at the problems of cumbersome steps and low efficiency in information processing and error analysis of the existing time-grating measurement systems, this paper proposes a time-grating measurement system based on LabVIEW. The measurement system reads and analyzes the bottom position information of a sensor to achieve data acquisition, real-time positioning, spectrum analysis and error processing. The experimental results show that, as the measurement error analysis of the sensor is accurate, the system runs stably, and multiple functional modules are integrated into a system platform, which greatly improves the measurement efficiency.
    Related Articles | Metrics

    An electric field time-grating signal processing method for constructing homologous digital reference signals

    2023, 37 (7):  161-168. 
    Abstract ( 80 )   PDF (4246KB) ( 96 )   Save
    This paper proposes a new method for measuring the phase change of electric field time-grating signals. This method uses the characteristics that the frequency of the time-grating induction signal is fixed and the signal returns to the initial phase point after the phase shift of 2π. The counter based on the high frequency pulse is used as the time measuring ruler, and the full range measurement of the ruler is equal to the periodic change time of the induction signal. The time scale is then subdivided by multiple phase clock sampling. When the phase of the signal moves periodically, the corresponding reading on the ruler changes periodically. Based on the inherent characteristics of the time-grating induction signal, this method avoids the measurement errors caused by the inconsistency of the time reference, signal skew and edge jitter in the construction of the entity reference signal, and also avoids the limitation of the sampling performance of the multi-phase clock, which further improves the measurement stability and resolution. By designing a time-grating signal processing system based on FPGA, a feasibility test of the optimized scheme is carried out on a 120-pole electric field circular time-grating sensor. The stability of the sensor reaches 0.405′′, which is nearly 3 times higher than before. The theoretical resolution and actual resolution are 0.135′′ and 0.54′′, which are about 2 times higher than before.
    Related Articles | Metrics
    Machinery and materials

    Research on improved Debevec-YOLOv5 for surface defect detection methods of high-reflective metals

    2023, 37 (7):  169-176. 
    Abstract ( 148 )   PDF (2000KB) ( 247 )   Save
    High-reflective parts have extremely strong reflectivity. When machine vision systems are used to detect such parts, the captured images contain high brightness interference factors, making it difficult to accurately detect surface defects on the parts. Therefore, based on high dynamic range imaging technology, this paper proposes a method for surface defect recognition by combining the improved Debevec algorithm with YOLOv5. The camera response curve algorithm and image synthesis algorithm of the Debevec algorithm are improved using particle swarm optimization algorithm, and YOLOv5 is used for defect recognition on the synthesized images. Objective evaluation metrics such as information entropy are calculated for the synthesized images, and the results show that the improved algorithm has a better image synthesis quality for reflective parts than the Debevec algorithm and Mertens algorithm do. The false detection rate and missing detection rate of the improved algorithm combined with YOLOv5 to synthesize images are lower than those of the Debevec algorithm and Mertens algorithm do, indicating practical value.
    Related Articles | Metrics

    Study on properties of petroleum asphalt and mixtures on Road 90 A in Tianzhize, Xinjiang

    2023, 37 (7):  177-183. 
    Abstract ( 103 )   PDF (2333KB) ( 78 )   Save
    In order to study the basic properties of petroleum asphalt and the pavement performance of mixtures on Road 90 A in Tianzhize Xinjiang, this paper adopts tests of the basic properties, high temperature rutting, low temperature bending, immersion Marshall and freeze-thaw splitting. The results show that all the basic property indexes of the petroleum asphalt on 90 A meet the technical requirements. The petroleum asphalt mixtures on 90 A have good high temperature stability, low temperature crack resistance and water damage resistance. The dynamic stability value of the mixtures is far greater than the standard value. The flexural tensile strength and flexural stiffness modulus are 8.5 MPa and 2 608 MPa respectively, and the flexural tensile strain is 3 259 με. The residual stability of the immersion is 86.6% and meets 75% of the standard requirements. The freeze-thaw splitting tensile strength ratio is 82.1%, which meets not less than 70% of the specification requirements. The petroleum asphalt on 90 A road has good high temperature, low temperature and anti-aging properties, which can effectively prolong the service life of the asphalt pavements in Xinjiang.
    Related Articles | Metrics

    Experimental and simulation research on thermal characteristics of square lithium battery modules

    2023, 37 (7):  184-191. 
    Abstract ( 150 )   PDF (3505KB) ( 187 )   Save
    For efficient and reliable operation of lithium batteries, this paper investigates the thermal characteristics of lithium battery cells and modules under different multiplier discharges. Based on a one-dimensional heat generation model and a three-dimensional heat transfer model, an electrochemical-thermal coupling model for lithium batteries is established. The simulation is compared with the experimentally measured voltage and average temperature to verify the validity of the model. Meanwhile, the change rule of the battery temperature is analyzed from the perspective of heat generation. In addition, the influence of the connecting aluminum bars and both the positive and negative poles on the temperature field distribution of the battery module is explored from whether the bars and the poles generate heat. The results show that the larger the discharge multiplier, the faster the temperature rise of the battery surface. At a small discharge rate, reversible heat is the main factor affecting the change of the battery surface temperature. The heat generation of the bars and the poles has a large effect on the peak temperature and the location of the battery module, but almost no effect on its minimum temperature as well as the location. When the discharge multiplier is 2 C, the peak temperature difference reaches 6.5 ℃.
    Related Articles | Metrics

    Comparison study on ISO and AGMA standards for gear surface damages

    2023, 37 (7):  192-200. 
    Abstract ( 104 )   PDF (1604KB) ( 107 )   Save
    Aim at the difference between ISO 6336 and AGMA 925 standards on the calculation methods of gear surface damage such as scuffing, wear, micro-pitting and so on, this paper studies the difference of contact temperature and specific lubricant film thickness ratio in detail. The standard on the capacity of gear surface damages is studied from three aspects, including principle analysis, theoretical calculation and experimental cases. The calculation of multiple examples is conducted to compare the contact temperature, lubricant film thickness and safety factors of the gear surface under two different standards. The research results indicate that there are differences in the calculation methods of gear surface load in ISO 6336 and AGMA 925 standards, and differences in the fitting of the friction coefficients, resulting in differences in the calculation results. AGMA 925 uses the bulk temperature to calculate the local lubricant viscosity, while ISO 6336-22 uses the local sliding coefficient to consider the impact of the flash temperature on the local lubricant temperature. Therefore, the oil film thickness calculated by ISO 6336-22 is lower than that by AGMA 925 standard at high flash temperature points. Due to the different risk assessment results calculated by different standards, each standard result should be compared during the design process to improve the accuracy of risk prediction.
    Related Articles | Metrics
    Information and computer science

    An improved voxelized generalized iterative closest point search algorithm

    2023, 37 (7):  201-207. 
    Abstract ( 144 )   PDF (3101KB) ( 127 )   Save
    In an outdoor environment with too many features in a large area, problems like accuracy errors and lack of robustness of a simultaneous localization and mapping (SLAM) system with multiple sensors may occur because of feature mismatch. In this view, this paper proposes an improved voxelized generalized iterative closest point (VGICP) method. Firstly, a feature credibility screening method is proposed to provide the system with an accurate initial guess by using the complementary characteristics of laser-inertial navigation-vision sensors to perceive different environments. Then, the visual feature subset is associated with the point cloud data through depth information, and the voxelized target point cloud group with high observability is screened by adding visual constraints, which makes the positioning and mapping more accurate while reducing the computational complexity. The simulation experiments show that, when the SLAM system based on multi-sensor fusion builds maps in an environment with more feature point clouds, the positioning accuracy improves by 12.335% compared with that of LVI-SAM system. When the operation linear speed exceeds 10 m/s, the robustness of the system is improved, which has strong feasibility.
    Related Articles | Metrics

    Research on intrusion detection methods for industrial control network

    2023, 37 (7):  208-216. 
    Abstract ( 170 )   PDF (1492KB) ( 150 )   Save
    Aiming at the problems of uneven distribution of data types and high dimensions in the current industrial control network environment, this paper uses the data augmentation method of auxiliary classifier generative adversarial network (ACGAN) to enhance the data set, and adopts a convolutional neural network (CNN) and extreme learning machine (ELM) hybrid model for feature extraction and classification of the data set. Through the simulation experiments on the NSL-KDD data set, the accuracy rate of the hybrid model reaches 99.26%, and the false negative rate is lower than 0.625%, which are better than traditional machine learning algorithms. At the same time, the natural gas pipeline data set of Mississippi State University is used for experimental simulation verification, with an accuracy rate of 99.18% and a false negative rate lower than 0.621%. This model is also applicable in complex industrial control environment, and broadens the research idea of industrial intrusion detection.
    Related Articles | Metrics

    Dialogue relationship extraction with dependency relation

    2023, 37 (7):  217-226. 
    Abstract ( 88 )   PDF (2624KB) ( 112 )   Save
    In order to enhance the ability to extract entity pair relationship in dialogues, this paper proposes a DEP-GAT model by introducing dependency relation into a heterogeneous graph attention network. Initially, the basic characteristics of each word are obtained through the preprocessing layer. Subsequently, in the discourse coding layer, context features are extracted and dependency information is added to further understand the speech structure. Eventually, a heterogeneous graph is constructed by utilizing the features, and an effective message passing mechanism is designed to enable the updated dialogue entity pairs to contain all the context information and grammatical features of the entire dialogue, thereby further enhancing the ability of the model to extract entity relations. The experimental results reveal that, on the DialogRE data set, the DEP-GAT model performs better than the baseline model does, with an increased F1 value of 2.9% in the development set and 1.8% in the test set respectively.
    Related Articles | Metrics

    A blockchain neighbor node optimization strategy based on Monte Carlo method

    2023, 37 (7):  227-234. 
    Abstract ( 87 )   PDF (1746KB) ( 87 )   Save
    Aiming at the problems of long block propagation time and poor blockchain network topology transmission performance in the current blockchain network, this paper designs an improved neighbor node optimization strategy based on Monte Carlo method. Firstly, the strategy calculates the score between a node and its neighbor node by the time of each round of blocks arriving at the node. Then, the strategy randomly adds new nodes from the candidate nodes into the neighbor set of the current nodes according to the elimination rate of the current neighbor nodes, and the strategy calculates all possible elimination combinations. The strategy then uses Monte Carlo method and Softmax function to obtain the probability that each combination may be eliminated. Finally, the strategy randomly selects nodes from the network to replace the current neighbor nodes according to the probability of elimination of the current neighbor nodes. The simulation experiments show that, compared with the strategy of randomly selecting neighbor nodes, the neighbor node optimization strategy can improve the propagation efficiency of blocks in the blockchain network, which reduces the average propagation time of the blocks by about 30%.
    Related Articles | Metrics

    Vulnerable node mining in directed weighted dependency software network based on community partitioning

    2023, 37 (7):  235-244. 
    Abstract ( 103 )   PDF (2647KB) ( 89 )   Save

    The structure of software systems is becoming more complex, and the possibility of software failure increases, which makes the cost of software understanding and maintenance for developers higher. Due to a lack of consideration of dependencies between classes in the existing vulnerability class mining methods, software maintenance is difficult to realize. In order to mine the vulnerable classes in the software and reduce the maintenance cost of the software, this paper designs a class dependency software network vulnerable class node mining algorithm based on community partitioning for the directed weighted class dependency network.

    Firstly, considering the dependencies and frequency between classes in the software, a directed weighted class dependency software network is constructed. The dependency frequency is used as the weight of the directed edge of the software network, and the weighted entropy of class dependence is defined. With the weighted entropy and node betweenness, class node vulnerability measurement is designed. Aiming at the deliberate attack of class nodes with high vulnerability, with the idea of BGLL algorithm, a directed weighted class dependency software network community partitioning algorithm based on modularity is proposed to divide the community of the class dependency software network. To test the performance of the community partitioning algorithm, the number of communities and weighted modularity are obtained and analyzed. The class nodes are used in the deliberate attack strategy, and, based on the analysis of community number and modularity change, the vulnerable classes in the class dependency software network are mined.

    To mine software network vulnerability classes based on the result of the community partitioning, an open source software system Jmeter3.0 is used as the standard experimental data for feasibility testing. As a Java based stress testing tool, Jmeter 3.0 includes 256 classes to support software execution. Using the software network analysis platform SNAP to parse the software source code of Jmeter3.0, the structural information of the software is obtained. Then, the directed weighted class dependency software network of Jmeter 3.0 is constructed. By using the class dependency software network vulnerable class node mining algorithm, Jmeter3.0 is divided into 32 communities. The class dependency software network of the largest community is displayed, which takes TestElement as the core. The top 15 vulnerable class nodes in Jmeter 3.0 are listed and the numbers of communities and modularity are discussed. When these vulnerable class nodes are used as the intentional attack, the numbers of communities and modularity are analyzed in comparison with the original values.

    The experiments are designed and the real open source software Jmeter3.0 is used to verify the effectiveness of the vulnerable class node mining algorithm for mining vulnerable classes. Three different class node attack strategies are used, including initial attack strategy, repeated attack strategy and random attack strategy. Compared with the random attack on 15 class nodes, the average modularity of the class dependency software network increases by 12.7% and the average community number increases by 60.12% in the intentional attack on the top 15 class nodes in the vulnerability. The proposed vulnerability class mining algorithm can effectively mine vulnerable classes in the class dependency software network.

    Related Articles | Metrics

    Micro-blog text emotion classification based on the fusion of content features and spread features

    2023, 37 (7):  245-255. 
    Abstract ( 114 )   PDF (1152KB) ( 165 )   Save

    The text vector representation method based on Word2vec does not fully consider the content features and spread features of micro-blog texts, so it is not good enough to finish the micro-blog text vector representation. Besides, a single machine learning algorithm which is applied to classify the micro-blog text through emotions can’t provide a high accuracy of emotion classification. To further improve the effect of emotion classification for the micro-blog text, this paper proposes a new text vector representation method, which is combined with the improved Stacking ensemble learning algorithm to accomplish emotion classification for micro-blog text data in this paper.

    At first, text feature vectors with rich semantic and emotional information are proposed to be constructed together by integrating text content features such as emoticons, semantic features of words, and part of speech and emotion, with the spread features such as comments, retweets and likes. Specifically, when constructing the initial text feature vector, this paper synthesizes the content features such as emoticons, word semantics, as well as part of speech and emotion. Meanwhile, it also constructs the corresponding feature vectors according to the above content features, and splices these vectors into the initial text feature based on content characteristics. Secondly, the influence of the text is constructed based on the spread features of the text, such as the number of comments, retweets and agreements. Finally, the influence of the micro-blog text is combined with the initial text feature vector to further enrich the semantic and emotional information contained in the vector representation of the micro-blog text.

    Moreover, in the improved Stacking ensemble learning algorithm, combined with the initial training data set, four classification algorithms are selected, such as AdaBoost, random forest, GBDT and XGBoost. Then, a 5 fold cross-validation method is used to generate a high-performance base classifier. More importantly, the class probability vector is used instead of the class label output from the base classifier. Different weights are set and multiplied with the class probability vector according to the performance of the base classifiers on the training data set. After that, they are multiplied by the class probability vector to get the weighted class probability vector, retaining the maximum weighted probability values, the minimum weighted probability values and the average weighted probability values of each text predicted by all base classifiers belonging to each category. A simple and stable logistic regression algorithm is selected as the meta-classifier as well. At last, the original Stacking algorithm is improved by integrating the above weighted probability values as the input data of the meta-classifier with the original text feature vector so as to accomplish emotion classification of micro-blog text.

    The experiment results on the data set of the micro-blog text show that the proposed method can better represent text vectors, and the improved Stacking ensemble learning classifier by the weight method is superior to the single emotion classifier. Compared with other emotion classification methods, the method proposed in this paper has made a performance improvement on the accuracy index from 1.75% to 4.90%, effectively improving the effect of emotion classification.

    Related Articles | Metrics
    Electrical and electronic

    Scheduling optimization of distribution network with wind power considering demand flexibility

    2023, 37 (7):  256-264. 
    Abstract ( 108 )   PDF (2200KB) ( 112 )   Save
    The inherent strong instability and randomness of wind power supply pose a great challenge to the classic distribution network scheduling and operation system. In order to deal with the deficiency of the existing studies that ignore uncertain out-of-set risks and fail to fully consider demand flexibility, this paper establishes a demand flexibility-based distribution grid scheduling optimization model with wind power based on the objective quantification of the uncertain out-of-set expected risk of wind power. In order to solve the complex mixed integer nonlinear programming model, this paper proposes a two-layer optimization algorithm based on the coordination of the improved antlion algorithm with the branch-and-bound algorithm, and makes full use of the advantages of artificial intelligence algorithms and classic mathematical algorithms to solve the complex model rapidly. The case study results show that, compared with the existing advanced scheduling methods, the proposed method can make full use of demand flexibility, effectively reduce the risk of distribution network scheduling, and reasonably weigh the economy and risks of the scheduling plan. In the actual distribution network scheduling, this research can provide important guidance for realizing more efficient and stable distribution network scheduling with wind power.
    Related Articles | Metrics

    Research on fault diagnosis of high voltage circuit breakers based on MI-ECHPO-PNN

    2023, 37 (7):  265-271. 
    Abstract ( 79 )   PDF (1734KB) ( 64 )   Save
    In order to improve the accuracy of fault diagnosis of circuit breakers and realize accurate fault identification, this paper proposes a fault diagnosis method of high voltage circuit breakers (MI-ECHPO-PNN) based on mutual information feature selection and improved prey algorithm optimized probabilistic neural network. After the vibration signal is decomposed by variable mode decomposition, the components with higher frequency is selected to extract the fault feature, and the feature is screened by the mutual information algorithm as the input of the diagnosis model. Using the improved predator algorithm to optimize the smoothing factor of the probabilistic neural network, the optimized parameters are input into the probabilistic neural network to build an ECHPO-PNN fault diagnosis model. The simulation results show that the ECHPO-PNN model has better diagnostic effect than other PNN models do, and the accuracy can reach 100 %, showing good accuracy and stability.
    Related Articles | Metrics

    Time characteristic analysis of stereo garages based on time-segment non-parametric tests

    2023, 37 (7):  272-278. 
    Abstract ( 59 )   PDF (3780KB) ( 106 )   Save
    In order to improve the operation efficiency of car parking and car departure in a stereo garage, this paper analyzes the distribution characteristics of customer arrival, customer departure and vehicle residence time in the stereo garage, and sets up an operation time model of customer car parking and car departure by using real arrive-departure data. A discrete distribution model is used to fit vehicle arrival and departure time of the stereo garage, with a continuous distribution model of vehicle residence time for fitting. A non-parametric test analysis method is used to verify and analyze the multi-period data, make sure of the fitting precision, and, at the same time, accurately describe the distribution characteristics of the arrival time and residence time of customers in different time periods. The verification results show that the arrival time of customer car parking and car departure obeys Poisson distribution and the residence time in the garage obeys Gamma distribution. The simulation compares the vehicle simulation arrival process with the actual operation arrival process, which shows that the arrival process of customer car parking and car departure the stereo garage conforms to the Poisson process. The study on the time distribution characteristics of stereo garages is helpful to obtain the passenger flow information of time-segment garages, which provides a reference for studying the operation process of this kind of stereo garages and the subsequent scheduling strategy design.
    Related Articles | Metrics

    Three-vector model predictive current control of doubly-fed wind turbines

    2023, 37 (7):  279-288. 
    Abstract ( 88 )   PDF (4290KB) ( 120 )   Save
    Double-vector model predictive current control of a DFIG improves the problems of fixed voltage vector direction amplitude, low current waveform quality and large torque ripples in traditional single-vector model predictive current control. However, there are still some current ripples and torque ripples only when q-axis current is deadbeat controlled. In this regard, the number of vectors is increased on the basis of double vectors, and a three-vector model predictive current control strategy is proposed. The desired voltage vector in this strategy is composed of two adjacent effective voltage vectors and a zero vector, and its range can cover any direction and any amplitude. The current of d axis and q axis of the side of the doubly-fed wind turbine is also deadbeat controlled, so the current pulsation of d axis and q axis reduces effectively. The simulation and experimental results show that, compared with single and double vector model predictive current control strategies, the current ripples and torque ripples under the current control of the three-vector model prediction reduce significantly, showing a better dynamic and steady-state performance.
    Related Articles | Metrics

    Research on online parameter identification of permanent magnet synchronous motors

    2023, 37 (7):  289-296. 
    Abstract ( 152 )   PDF (2506KB) ( 509 )   Save
    Aiming at the problems of long recognition time and slow convergence speed in the parameter identification process of permanent magnet synchronous motors of the basic particle swarm algorithm, this paper proposes a chaotic genetic particle swarm algorithm (CHPSO) combining chaotic mapping and information transmission to identify the parameters of permanent magnet synchronous motors online. The algorithm generates chaotic particles through chaos mapping, combines with the previous parameter identification results to generate an initialized population, and then introduces dynamic inertia weight coefficients to improve particle diversity. At the same time, step-by-step identification and cyclic updating methods are adopted to solve the problem of under-ranking parameter identification. The simulation shows that the deviations of the algorithm in identifying the motor parameters are 1.32% stator resistance, 1.08% flux linkage, 0.92% d-axis inductance and 1.16% q-axis inductance respectively. Finally, the effectiveness of the identification scheme is proved by bench experiments.
    Related Articles | Metrics

    Prediction of interfacial tension of transformer oil based on KPCA-SSA-ENN

    2023, 37 (7):  297-305. 
    Abstract ( 78 )   PDF (3539KB) ( 72 )   Save

    Aiming at the problems of long time of detection and high cost in traditional detection methods of interfacial tension of transformer oil, this paper proposes a novel prediction method of interfacial tension based on multi-frequency ultrasonic detection technology and an artificial intelligence algorithm. 175 groups of transformer oil samples are measured through the ring interfacial tension method and multi-frequency ultrasonic detection, and the correlation between amplitude-frequency response, phase-frequency response and interfacial tension of multi-frequency ultrasonic signals is analyzed. The multi-frequency ultrasonic data are preprocessed by kernel principal component analysis (KPCA), and the sample set is divided into a training set with 140 groups and a test set with 35 groups. The sparrow search algorithm (SSA) is established to optimize the interfacial tension prediction model of Elman neural network (ENN). The average percentage error of the prediction is 6.53%, and the prediction accuracy reaches 93.47%.


    Related Articles | Metrics
    Energy, power and environment

    Power consumption prediction and application of GM(1,1)-MEA-BP combined model

    2023, 37 (7):  306-314. 
    Abstract ( 72 )   PDF (2750KB) ( 82 )   Save
    Aiming at the problems of the weak generalization ability and low prediction accuracy of the traditional single model, this paper proposes a combined prediction model of GM(1,1) grey model and MEA-BP neural network. The model solves the problems of the influence of large random fluctuation and low prediction accuracy of the GM(1,1) prediction model on energy consumption prediction caused by time factors. With the advantages of MEA-BP neural network parallel computation, strong fault-tolerant force and distributed information storage, it also reduces the situation where the accuracy of the prediction results is affected by data fluctuations, and solves the problem of error feedback adjustment. The total amount of national electric energy consumption from 1985 to 2020 is selected as the modeling data, and the prediction results of linear regression, triple exponential smoothing, GM(1,1), BP neural network, MEA-BP neural network and other models are analyzed and compared. The results show that, compared with other models, GM(1,1)-MEA-BP combination model has the highest prediction accuracy and the smallest error, with MAPE value reaching 0.006 5 and RMSE value reaching 977.996 1. It is proved by an example that GM(1,1)-MEA-BP combination model has a high prediction accuracy of the electric consumption in China, which provides a basis for national macro intelligent scheduling in energy.
    Related Articles | Metrics

    Numerical study on the aerodynamic noise at the inlet of the dust removal system of a tobacco factory

    2023, 37 (7):  315-323. 
    Abstract ( 75 )   PDF (2751KB) ( 67 )   Save

    The tobacco industry is one of the important pillar industries of national economy in China, and, with the expansion of tobacco factories as well as a larger number of high-speed unit equipment into the tobacco industry, noise pollution in the industry is becoming more and more prominent, in which the noise pollution in wrapping workshops is particularly serious. The noise source in such a workshop is mainly from the dust removal system, the noise characteristics of which are usually low to medium frequency noise, with strong penetration and energy difficult to attenuate. The air inlet end of the dust removal system is the closest to the workers, and the long time exposure of the workers to the noise environment will lead to damage of hearing organs as well as tinnitus, palpitation, sleep disorders, etc. Therefore, it is necessary to analyze the noise of the air inlet of the dust removal system as well as to optimize the structure of the air inlet to reduce the noise.

    Most of tobacco factories for the optimization of the noise reduction of the dust removal system is mainly done through the experimental method, but with the development of computer technology and numerical computation methods, the digital simulation method has become one of the most effective ways to solve complex engineering problems. Compared with the analytical method which is only applicable to the noise analysis of simple structures, it can be used to solve the aerodynamic noise of large and complex structures of air inlet ducts. At present, the research on the noise of the dust removal system of a tobacco plant mainly focuses on covering the pipe wall with acoustic materials and installing mufflers at the inlet and outlet of the pipe, etc. There are few researches on the noise reduction by optimizing the internal structure of the air inlet, and the establishment of a mathematical model to optimize the internal structure of the air inlet is a brand-new attempt.

    In this paper, the air inlet of the dust removal system in a tobacco factory in Hubei province is taken as the research object. The CFD software Fluent and the acoustic software LMS Virtual.lab are used to jointly simulate and establish an air inlet model of the dust removal system with a valve. Through the established digital model, the causes of the pneumatic noise of the valve are analyzed, and the characteristics of the pneumatic noise generated by gas flowing through the valve in the pipe are discussed. The influence rule of valve opening on noise is studied and the valve opening structure is optimized. The results show that the aerodynamic noise of the inlet is mainly concentrated in the low frequency range. Valve opening affects the sound pressure level of the pneumatic noise but does not change the spectral characteristics. Under different valve opening degrees from 30° to 75°, the difference of the maximum noise pressure level drop is about 6.62 dB, 5.42 dB and 1.43 dB respectively, when the valve increases by 15°. The mean values of the sound pressure levels and the maximum sound pressure levels decrease with the number of valve opening. The mean sound pressure level with 9 holes reduces by about 25.69% compared with that without holes. The relevant results provide research ideas and practical cases for the noise reduction design of the air inlet of the dust removal system in a tobacco factory based on multi-software co-simulation.

    Related Articles | Metrics

    Study on the effects of different ventilation schemes on air distribution and air quality in hospital outpatient clinics

    2023, 37 (7):  324-335. 
    Abstract ( 110 )   PDF (4366KB) ( 139 )   Save

    A lack of ventilation and poor air quality in many buildings not only affect the learning efficiency of indoor occupants, but also cause sick building syndrome and even increase the risk of infectious diseases. As public spaces where health care workers look after patients with a complex mix of people, hospital outpatient clinics are highly crowded and are largely enclosed during consultations, so airflow and air quality requirements should be even more stringent. In order to avoid health problems caused by poor air quality, a reasonable and effective study on indoor airflow and air quality in hospital outpatient clinics is needed to ensure the quality of their indoor air environment. The existing studies on outpatient clinics mainly include the airflow in large spaces in outpatient buildings and the risk of indoor droplet exposure in outpatient clinics, without considering the evaluation of indoor personnel comfort and contaminant removal efficiency. There is a lack of multi-perspective evaluation research on indoor airflow organization, indoor air quality and personnel comfort in outpatient clinics.

    Therefore, to improve the comfort and safety of people in hospital outpatient clinics, this paper conducts indoor air quality tests for hospital outpatient clinics, and establishes a ventilation model of outpatient clinics based on the numerical simulation method. It combines the three air supply modes (upper air supply and upper exhaust, upper air supply and lower exhaust, and under floor air supply and upper exhaust) with two exchange times per hour (ACH=3/6 h-1). The effects of six ventilation schemes on the indoor air distribution and contaminant concentration are compared and studied. Draught rate, energy utilization coefficient and contaminant removal efficiency are used for evaluation and analysis. The results show that:

    (1) An experimental study of outpatient clinics in a hospital in Taiyuan reveals that the indoor air quality in consultation rooms is poor. All eight consultation rooms measured have a concentration that exceeds the standard (>1 000 ppm) during the consultation period. In seven of the consultation rooms, the time of excessive concentration accounts for more than 50% of the total time of consultation, with the highest percentage reaching 86%.

    (2) By increasing the air change per hour, the indoor CO2 concentration can be reduced from the over-standard concentration to less than 1 000 ppm (part per million), and the reduction range is more than 200 ppm.

    (3) Under the upper air supply and upper exhaust scheme, the draught rate is lower and the contaminant removal efficiency is higher for the same number of air changes, which is effective in removing pollutants while ensuring the comfort of the occupants. However, the energy utilization coefficient is the lowest, which means some heat is wasted.

    (4) The adoption of the upper air supply and lower exhaust scheme helps to improve the personal comfort and the energy utilization coefficient of the air supply. Its draught rate is the lowest and its energy utilization coefficient is the highest under the same ACH. However, it has the lowest contaminant removal efficiency and the highest CO2 concentration at the same height, which does not effectively remove pollutants but leads to a build-up of pollutants.

    (5) When the under floor air supply and upper exhaust scheme is used, the contaminant removal efficiency is the highest, with 1.24 and 1.49 for the two types of air change per hour respectively, allowing for efficient removal of contaminants in the exhalation range. However, when the air supply velocity is low, its draught rate is the highest, which affects the personal comfort directly. This phenomenon can be effectively alleviated by increasing the air change per hour.

    Related Articles | Metrics

    SOC joint estimation of lithium batteries at different temperatures

    2023, 37 (7):  336-342. 
    Abstract ( 103 )   PDF (2297KB) ( 139 )   Save
    Aiming at the problems of noise and poor robustness of extended Kalman filter algorithm, this paper carries out pulse discharge experiments and the least square off-line identification at different temperatures based on the second-order RC equivalent circuit model. Then, the double adaptive robust extended Kalman filter (DAREKF) algorithm is proposed to jointly estimate the model parameters and SOC online. The simulation results show that, compared with AREKF algorithm, the proposed algorithm can keep the SOC estimation error within 1.14% at room temperature. At a low temperature, the closer to 0°C, the smaller the error of the algorithm.
    Related Articles | Metrics