Methods and Applications for Modeling and Simulation of Complex Systems

With the rapid development of military simulation technology, simulation systems for different levels of strategy, campaign, tactics, and technology have emerged. This manuscript intends to use model mapping transformation and machine learning technology to automatically reconstruct the low-resolution simulation model from the high-resolution simulation model data. The premise of learning is to have rich and effective learning data. Therefore, this project uses the existing data cultivation technology/learning data generation technology based on a large sample simulation experiment to generate the simulation data for machine learning through the continuous iterative process of scene design, high-resolution simulation scenario construction, simulation experiment design, and sample generation, efficient parallel simulation, data analysis, and application.

When conducting complex simulation experiments, experimental factor screening is of great significance to improve the effectiveness and efficiency of the simulation. Aiming at the problems of unsatisfactory accuracy and efficiency of existing experimental factor screening methods, a simulation experiment factor screening method based on combinatorial optimization (EFS-CO) was designed. EFS-CO abstracts the experimental factor screening problem into a combinatorial optimization problem, designs the experimental factor evaluation criteria based on the support vector machine model, uses it as the fitness function of the genetic algorithm, and finally obtains the significant experimental factor. At the same time, to eliminate the influence of the randomness of the genetic algorithm on the screening results, the screening results of the experimental factors were determined based on the results of multiple experiments. Finally, the effectiveness of the method is verified by example analysis and comparing the results with the traditional experimental factor screening method.

Mesh models from various numerical simulation programs have great diversity in the data formats and structures, leading to difficulties in sharing and reusing of mesh data for collaborative simulations between commonly used commercial programs and in-house codes. To break the barrier against the free exchange of mesh models, in this work a unified mesh model with full coverage of simulation-related data is deliberately designed. Based on such mesh model and the open source mesh data management framework MOAB (Mesh-Oriented data-base), a generic mesh model converter is developed, enabling the high-efficient conversions of mesh models between various simulation programs. The converter could be easily extended for conversion between new pair of mesh models by integration of the corresponding IO interfaces. The correctness and scalability of the converter is validated through a series of examples.

Model validation is one of the core issues in modeling and simulation, which aims to check whether the model output correctly reflects the output of the corresponding real system. Model validation is also a complicated process, which requires powerful software to achieve its purpose. In this paper, we describe a simulation model validation tool (shortly SMVT) that we developed recently for the validation purpose of complex simulation models. We first talk about the functions and architectural design of SMVT and then employ a missile simulation model to demonstrate the application of SMVT. This tool has so far been widely used in many scenarios.

Compared with traditional control system modeling, directed graph, as one of the graphical model representation methods, can be introduced into the control system modeling process to effectively improve the modeling efficiency. Based on this, a graphical modeling and simulation method of control system based on directed graph is proposed. Specifically, a control system modeling method based on hierarchical model elements and directed graph description is proposed. Secondly, a control system parsing method oriented to graphical description is designed and implemented, and an integrated environment of graphical modeling and simulation execution is realized. On this basis, control systems based on directed acyclic graphs and directed cyclic graph descriptions are constructed, and the effectiveness of the proposed method in control system modeling and simulation is demonstrated by cases of first-order system, second-order system, follower systems and hydraulic circuits.

In this paper, the defect mode of one-dimensional photonic crystal was studied. For photonic crystal with one defect layer, the effects of refractive index and thickness of defect layer, photon period number and incident angle on defect mode were discussed. Photonic crystal structure with multiple defect layers was constructed, and multiple defect modes were obtained in the forbidden band. Based on the photonic crystal structure with two defect layers, a dual-channel narrow-band filter operating at 480 nm and 532 nm was designed. Particle swarm optimization algorithm was introduced to optimize the position and transmittance of defect modes, which mainly depended on the thickness of defect layer. The calculation results showed that the transmittance of the two defect modes reached 97.52% and 97.55%, respectively, while the full width at half maximum was only 1.0 nm and 1.8 nm. The simulation of the electric field distribution further proved the correctness of the design results. The research results had important value in the field of underwater laser communication.

To meet the demand that analyzing the impact of architecture and equipment performance on capability and cost of equipment system-of-systems, an evaluation method is investigated based on linear constant discrete system. Model of equipment system-of-systems is established using state space equation of linear time invariant discrete system, which describes equipment composition, connection relation, operation process and equipment performance of the equipment system-of-systems. Operation characteristic of the equipment system-of-systems is analyzed using state motion equation of linear time invariant discrete system. Then, evaluation models for adaptation capability, survivability capability, task capability and response capability are established, as well as a cost evaluation model for equipment system-of-systems are given. The method proposed in this paper is demonstrated with a hypothetical air defense equipment system-of-systems. Evaluation results of the above capabilities and cost of the air defense equipment system-of-systems are derived. These results can support optimization of the equipment system-of-systems architecture and equipment performance.

When the state-space model is black-box, it is difficult to identify the system based on the input and observation. In predictive control and other fields, the state and model need to be updated in real time, and online identification becomes very important. However, compared with offline learning, online learning of black-box model is more difficult. This paper proposes an online Bayesian inference and learning method for state-space models with missing observations. When the state-space model is black-box, we expressed it as basis function expansions. Through the connection to the Gaussian processes (GPs), the state and basis function coefficients are updated online. The problems of missing observations caused by the sensor failure are often encountered in practical engineering and are taken into consideration in this paper. In order to keep the online algorithm from being interrupted by missing observations, we update the states and unknown parameters according to whether the observation is missing at the current time. This conservative strategy makes the online learning continuous when the observation is missing, and makes full use of the available statistics in the past. Numerical examples show that the proposed method is robust to missing data and can make full use of the available observations.

We consider a nonlinear state-space model with unknown state transition function and process noise. The state transition function is modeled as a Gaussian process, besides, it is also expressed as basis function expansion. Using the connection to the Gaussian process, the prior of the coefficients of basis function can be obtained. The posterior of the state and unknown coefficients can be obtained through Bayesian inference. Sequential Monte Carlo (Particle Gibbs with ancestor sampling) is used to estimate the states. The coefficients are modeled as random variables related to state statistics. The Markov Chain Monte Carlo (MCMC) method is used to repeated iteratively sample from parameter posterior and state posterior. The problems of missing observations due to sensor failure are often encountered in practical engineering and are taken into consideration in this paper. We propose a learning method for nonlinear dynamical system with missing observations. According to whether the observation data is missing or not, the state and parameters are updated respectively. The proposed nonlinear system identification method is robust to incomplete dataset. A numerical example is used to demonstrate the effectiveness of the proposed method.

Aiming at the difficulty of optimizing the parameter estimation of nonlinear models, a new method for parameter identification of nonlinear system models based on improved differential evolution algorithm based on diversity evaluation index is proposed. By establishing a population reconstruction mechanism, based on the population diversity index, the concept of population similarity is proposed to guide the selection of evolutionary strategies and the adaptive adjustment of process parameters, thereby balancing the global and local search functions at different stages. The population size decreasing strategy effectively reduces the amount of computation and improves the convergence speed and algorithm efficiency. In order to verify the performance of the algorithm, several types of standard functions with typical complex mathematical characteristics are simulated and applied to the identification of a type of thermal system model parameters. The results show that the improved algorithm has high model parameter identification accuracy and faster convergence speed, which effectively improves the accuracy and efficiency of model establishment, and provides a feasible way to solve the model parameter identification problem in practical systems.

Aiming at the problem of integrating open tools in a user development environment in complex product design, an integration strategy of model engineering environment based on OpenMBEE is studied. It is based on the technologies of user identity unified authentication and user interaction data collaboration. Specifically, in the solution of unified identity authentication on cross-platform, the extended controller method is provided based on the message queue information flow mechanism. Then, the separation of business logic, data information, and interactive view of user unified authentication are realized, on the basis, that the function of unified identity authentication is realized by using the Software Development Kit (SDK) of OpenMBEE. Besides, in the study of user interaction data collaboration technology, three-layer collaboration mechanisms are proposed, they are the presentation layer collaboration, the business layer collaboration, and the persistence layer collaboration. When the jQuery programming framework is integrated into the bootstrap front-end framework, combing with the OpenMBEE back-end server interface, the collaboration of user interaction data is realized. Through experiments and project practice analysis, the user unified authentication strategy and user interactive collaboration strategy are tested respectively for proving the effectiveness and feasibility of the proposed method.

The digital twins of the whole life cycle of complex products have been widely used. The abstract concept of digital twins provides principles for the modeling of digital twins, but it cannot effectively guide the modeling and application of digital twins. In this paper, a method to construct the digital twins of complex products is proposed by studying the components, relationships, construction processes and carrier forms of the digital twins of complex products.

Based on the Digital Radio-Frequency Memory (DRFM) technology, a design scheme of portable radar jamming simulation system is proposed in this paper. Here we mainly discuss the general design framework with detailed index, general compositions, and the jamming signal generation method, of which the DRFM module is analyzed in detail, and a digital simulation is made for some typical jamming signal forms. This portable scheme can satisfy the design requirements of the flight mounted target missile, and can make a fine developing example for such flight mounted applications of radar simulators.

To solve the coupling problem of multivariable systems, this paper proposed a decoupling design method of compensated active disturbance rejection control that can detect the disturbance among coupling loops with the extended state observer and design a control law to eliminate it. Besides, a compensation element is added to the traditional active disturbance rejection controller so that the observer can estimate the state of the system and the influence of disturbance accurately. The comparative simulation results show that the decoupling effect of the proposed method is good. The Monte Carlo experiment validates its robustness and good application potential.

This paper proposes a layered framework for combining virtual and real simulation. The framework combines a physical system based on ROS with a simulation system based on a simulation engine, builds a multi-level agent middleware between virtual and real heterogeneous systems, and analyzes the latency and packet loss rate metrics of multi-level agent communication through experiments. When sending messages at 50 Hz as frequency, reducing the number of agent levels leads to a significant reduction in the average value of latency and jitter. More stable communication performance and increased bandwidth reduce the packet loss rate. The variation of communication performance with frequency for the most complex three-level agent is analyzed with emphasis. Finally, the conclusions of the experiments are given, and future work has prospected.

In order to improve the visual perception of color images under low illumination conditions, an adaptive enhancement method is proposed to enhance the contrast and brightness, enrich the color and avoid over-enhancement. Firstly, the original low illumination image is converted from RGB color space to L*a*b* color space, and a novel gray level transformation function namely piecewise sine function is proposed to improve the brightness of L* channel image. In addition, dragonfly algorithm is utilized to optimize the parameters in piecewise sine function to achieve the best brightness adjustment effect. Then a novel saturation enhancement method is proposed to enrich color information. Subsequently, a fitness function that takes into account both the degree of overall brightness enhancement and the suppression of information loss is applied as the objective function of dragonfly algorithm. Ultimately, the processed L*a*b* color space is transformed back to RGB color space to get the enhanced image. Experimental results verify the effectiveness of the proposed method.

Aiming at the difficulty of target recognition in low and medium resolution remote sensing images of maritime mobile ships, in order to accurately match and identify ship types, this paper proposes a ship Kelvin wake extraction method based on Hough transform for matching and identifying ship models. The average distance of Kelvin wake spike wave of different types of ships is obtained through simulation, which can effectively identify the ship attributes in real remote sensing images, and the comparative analysis with the actual image data proves that the algorithm has certain feasibility. Therefore, the Kelvin wake extraction method based on Hough transform can basically achieve the recognition of target ship type under certain conditions and improve the recognition accuracy of maneuvering ship targets at sea by low and medium resolution imaging reconnaissance satellites.

Analysis of vortices is very important from the aspects of marine environment and disaster prevention. With the development of supercomputers, three-dimensional simulations of large-scale objects such as the ocean have been conducted. In ocean current simulations, multi-scale unstructured grids have been employed to precisely reproduce the topography of the seafloor and coastal areas. However, calculating vortices from unstructured grid data using the derivative of the flow velocity requires time-consuming data interpolation. Therefore, in this study, we calculate the covariance matrix using the velocity of each triangular column in the local region. The eigenvalues are used to define vorticity. In order to visualize the obtained vortices, the unstructured grid data are converted into point cloud data. In addition, the point density is adjusted according to the obtained vortices, and the opacity is changed. With the proposed method, we could visualize vortex regions with ambiguous boundaries correctly. During the visualization, we analyzed the causal relationship between the vortices and physical quantities such as flow velocity and salinity obtained from simulation by merging them into the visualization.

Metaverse expands the cyberspace with more emphasis on human-in-loop interaction, value definition of digital assets and real-virtual reflection, which facilitates the organic fusion of man, machine and material in both physical industry and digital factory. The concept of Industrial Metaverse is proposed as a new man-in-loop digital twin system of the real industrial economy which is capable of man-machine natural interaction, industrial process simulation and industrial value transaction. With the comparison with Metaverse and Digital Twin, the key features of Industrial Metaverse are summarized, which are man-in-loop, real-virtual interaction, process asserts and social network. Key technologies of Industrial Metaverse are surveyed including natural interaction, industrial process simulation, industrial value transaction and large-scale information processing and transmission technologies, etc. Potential application modes of Industrial Metaverse are given at the end as well as the challenges from technology, industry and application.

An improved maximum power point tracking (MPPT) control algorithm is proposed based on a doubly-fed induction generator wind energy conversion system. The control objective below rated wind speed is to maximize the extracted energy from the wind while reducing mechanical loads. The existing MPPT control method do rarely not take into account the influence of control method and control parameters on the mechanical parts of wind turbines such as transmission shafts. In order to bring some improvement the applicability of the control method, an improved method of MPPT control is proposed, which superimposes an additional torsional vibration suppression command on the basis of the original method, and suppresses the torsional vibration of the drive shaft of the wind power generation system while ensuring the maximum power tracking. Test studies substantiate the claims made and demonstrate the application of the methodology.

To solve the problem that a large number of services such as video, voice and multimedia need to be transmitted on the network, and these services are extremely sensitive to bandwidth, jitter and delay, the traffic integer is proposed. The traffic integer is used to classify different types of data, cache the data, and periodically send the data in the cache. This ensures that the high-speed data can be stably output on the link to avoid data loss, and ensures data continuity without data mutation. It adopts an open-loop control mode, according to different types of data system will be given the corresponding input, there is no interaction between input and output, can improve the utilization rate of bandwidth, reduce transmission delay, improve processing speed. Zynq7000 is used as the main control chip to realize data receiving and data processing on a single chip, which can improve the stability of work. Compared with the implementation of traffic integer in software, implementation in hardware has a faster processing speed.

Disturbance rejection control of underwater UAV is a key technology. First of all, In this paper, the model of underwater UAV will be established, and the closed-loop control system will be constructed according to the virtual control quantity. Then, sliding mode controllers and extended state observer are designed to realize the attitude and position control. Finally, the simulation is performed to show the robustness of UAV. It can be proved that under the influence of inhomogeneous media, UAV can still have good robustness and compensate for interference.

The take-off and landing of shipborne unmanned helicopters are the most dangerous parts in the process of its dispatch and recovery. By comparing and analyzing the advantages and disadvantages of surface aviation support system for helicopter among the navies of different countries, the key technologies of the take-off and landing process of shipborne unmanned helicopter are summarized. In order to solve problems such as the difficulties of autonomous take-off and landing of shipborne UAVs, a three-stage strategy in landing process is proposed for a 300-tons ship and a 350 kg-class unmanned helicopter, and a real-time forecasting algorithm is used to predict the motion of the deck. Simulation results verify the feasibility of the proposed strategy and the effectiveness of the forecasting algorithm based on FlightGear software, which lays a theoretical foundation to carry out the ship-aircraft adaptability test of shipborne unmanned helicopter and dramatically reducing the risk of the test.

In order to provide accurate UAV state estimation information for UAV monitoring and control, the UAV state estimation method based on variable structure multi model is studied. Firstly, the state model set is established for the motion forms of UAV, such as smooth flight, lateral turning maneuver and longitudinal jumping maneuver, and the measurement model is established based on the radar measurement principle; Then, based on the variable structure multi model framework, a model set adaptive strategy is proposed, which can solve the problems of too many threshold parameters and complex adaptive strategy; Finally, the simulation scenarios of two radar tracking UAVs are built, and the superiority of the proposed method in the accuracy of state estimation is verified by mathematical simulation.

The JAXA/ISAS SLIM project aims to land a small unmanned spacecraft on the Moon with pinpoint accuracy at its destination. This research aims to realize a method for estimating the flight position of a lunar explorer using image matching technology. The flight position is estimated by high-precision image matching between the lunar surface image taken by the probe and the lunar surface map image. However, the lunar surface image taken by the spacecraft is assumed to be a degraded low-resolution image due to various disturbances. This causes positional errors in the corresponding points between the captured image and the map image, which hinders the improvement of the accuracy of the transformation matrix. Therefore, an optimization method of transformation matrices based on the Gauss-Newton method is used to improve the accuracy of spacecraft flight position estimation. Besides, this method cannot eliminate the position error of the corresponding points themselves, thus limiting the improvement of position estimation accuracy. Then this study proposes a method for selecting corresponding points with small position error by utilizing the transformation matrix optimized by the Gauss-Newton method. Compared to the conventional method, the proposed method improved the number of successful estimates and the estimated average error from the true value. In particular, the proposed method was found to be effective for brightness, contrast, and noise disturbances, for which the conventional method had low position estimation accuracy.

Sailing speed and ocean wave environment have obvious effects on the motion characteristics of seaplane on waves. Using the numerical simulation method, based on the VOF method and the dynamic overlapping grid technology, the dynamic characteristics of the seaplane sailing on the wave surface are simulated, and the effects of the sailing speed, wavelength and wave height on the resistance characteristics, heave and pitch motion are investigated respectively. The numerical simulation results show that with the increase of navigation speed, the interaction between the body and the wave is stronger, and the peak resistance, the double amplitude values of heave and pitch are increasing; With the increase of wavelength, the wave steepness decreases, the peak value and average value of resistance curve decrease gradually, and the motion characteristics of heave and pitch and acceleration also have the same trend; With the increase of wave height, the wave steepness of the wave increases, the wave energy encountered by amphibious aircraft increases, the peak value and average value of the resistance curve of the aircraft in the wave gradually increase, and the heave, pitch motion characteristics and acceleration characteristics also have the same trend.

In view of the target tracking control and state estimation of unmanned aerial vehicle (UAV), a fixed wing unmanned aerial vehicle with a front fixed camera is used as the research object. The target is estimated by visual measurement and extended Kalman filter, and the 3D local coordinate system research guidance law. The dynamic equation of UAV and the state equation and observation equation of the target are set up, and the Kalman filter is designed to estimation the state of the ground moving target. The simulation results show that motion estimation based on EKF can achieve effective tracking of targets.

Quadcopter UAVs are always exposed to time-varying winds in the flight process, which makes it much more challenging to be detected and recognized by radar systems. Conventional solutions are to estimate and reject the wind aided adverse impacts, while in this paper we regard the wind aided aerodynamic characteristics as prior environmental information rather than disturbance. In particular, we establish a Matlab Simulink based quadcopter UAV control model in wind environment which is composed of wind shear, turbulence and gusts, and investigate the wind aided velocity and Doppler frequency fluctuations in a modeling and simulation way since the wind aided aerodynamic characteristics of the quadcopter UAV are difficult to be described in mathematical expressions. The simulation results reveal that: (1) the wind gusts are the primary components for the velocity and Doppler frequency fluctuations; (2) the impacts of wind turbulence are negligible due to the quad-copter UAVs’ wind resistance capability; (3) the impacts of the wind shear can be transformed between horizontal and vertical dimensions as a consequence of the coupling effect.

When a thrust anomaly occurs in the flight, launch vehicles should have the ability to deal with faults and deviations autonomously, predict trajectory reachable range and re-plan a new flight trajectory rapidly. The entry orbit elements are determined by flight status, and reachable range of aircraft is predicted. According to state equations and performance indexes, Hamilton function of the problem is constructed. Adaptive dynamic adjustment and control command re-planning can be adopted to ensure orbit entry. On the other hand, original problem can be re-described, and degradation scheme can be planned. According to mission partition, the optimal control problem is constructed and solved via sequence convex programming. The simulation results show that this method can re-plan trajectories rapidly in different situations and remedy abnormal states to a certain extent.

In this paper, the design and optimization of multi-aircraft cooperative guidance law with multi-constraint is studied for high-speed and large maneuvering targets. Firstly, the motion of the aircraft and the target are analyzed in the inertial coordinate system. The relative motion model and the corresponding state equation are established in a plane. Then, based on the sliding mode control method, the multi-constrained ballistic forming guidance law of the lead aircraft is designed to make it hit the target in multiple dimensions. Finally, the aircraft-to-target range and line-of-sight angle of the aircraft are selected as the state variables to be controlled, and the multi-constrained cooperative guidance law is designed for the follow aircraft. So that the motion of follow aircraft can track the lead aircraft accurately, the controlled state variables converges quickly, and hit the target at the same time until the end. The simulation results show that the leader-follower cooperative guidance law designed in this paper is effective for intercepting high-speed and large maneuvering targets.

By learning behavioral characteristics and biological phenomena in nature, such as birds, ants, and fireflies, intelligent optimization algorithms (IOA) is proposed. IOA shows feasibility in solving complex optimization problems in reality. Pigeon-inspired optimization (PIO) algorithm, which belongs to intelligent optimization algorithms, is proposed by the pigeons homing navigation behavior inspired. PIO is superior to other algorithms in dealing with many optimization problems. However, the performance of PIO processing large-scale complex optimization problems is poor and the execution time is long. Population-based optimization algorithms (such as PIO) can be optimized by parallel processing, which enables PIO to be implemented in hardware for improving execution times. This paper proposes a hardware modeling method of PIO based on FPGA. The method focuses on the parallelism of multi-individuals and multi-dimensions in pigeon population. For further acceleration, this work uses parallel bubble sort algorithm and multiply-and-accumulator (MAC) pipeline design. The simulation result shows that the implementation of PIO based on FPGA can effectively improve the computing capability of PIO and deal with complex practical problems.

In this paper, a low cost real-time monitoring and automatic detection system for PCB defect detection is studied. Two types of algorithms (single-stage algorithm and two-stage algorithm) TDD-NET algorithm and YOLO v5s algorithm in the field of deep learning object detection are selected. First, the research looks for data enhancement methods for problems with small scale of existing public data sets. Then, among at the problem of tiny target detection, feature fusion is used to improve the detection rate of tiny targets. Then, the performance of the dataset on the two algorithms is tested, and the performance indexes of the two algorithms are compared, and the real-time monitoring using YOLO v5s is realized. Finally, YOLO v5s is deployed in NVIDIA embedded hardware Jetson Nano. In addition, TensorRT and DeepStream are used to accelerate the model while invoking the CSI camera for real-time detection. After the hardware deployment, the algorithm can not only obtain the image data from the video and send it to the model for reasoning, but also maintain high accuracy and reasoning speed.

Transformation electromagnetics, which provides a new way to manipulate electromagnetic waves with coordinate transformation methodology, has become an important inspiration in wave propagation theories. In this paper, a two-dimensional cylindrical electromagnetic wave expander is discussed with its two basic variables. The variations of each constitutive parameter extremum are analyzed in detail with these two variables respectively. The different curves that vary with the basic variables can provide some guidance for the production of wave expanders. The scattering minifying functions of the wave expander are further studied through numerical simulations of two-dimensional embedded aircraft models and the quantitative calculations of their scattering width values, along with which the scattering width of the expander itself is also studied. A conclusion is drawn that the minification degree of the scattering cross-section of the embedded object inside the wave expander depends on the reduction factor of the wave expander.

The main function of the intake system of the compressor test rig is to ensure that the air flow into a compressor is uniform and consistent. Firstly, the dimensionless total pressure on the experimental measurement section from the numerical is almost agreed with that from the experimental, except that there is only 5% deviation at 10% channel height. Therefore, the numerical simulation method is correct and the result is reliable. Secondly, the characteristics of pressure field, velocity field and boundary layer thickness before the variable inlet guide vanes (VIGV) and the first stage rotor (R1) are studied. The results show that the closed area formed by the intake stabilizing chamber and the inlet bell mouth of the test article has not affected the flow around the bell-mouth, especially the evolution and development of boundary layer. Two vortices directly behind the support plate wake are observed at the hub, which is caused by the flow channel falls. These two vortices have opposite rotational directions and axial vorticity direction. Throttle does not affect the dimensionless velocity and dimensionless pressure distribution at the inlet of the test article, nor does it affect the thickness of the boundary layer. With the increase of rotational speed and mass flow, the radial distribution becomes increasingly uniform. Meanwhile, the thickness of boundary layer at the hub and shroud decreases slightly, but the range of variation is within 1% of the channel height.

To satisfy an increasing demand to reconstruct an athlete's motion for performance analysis, this paper proposes a new method for reconstructing the position, acceleration, velocity and angle of skis in the context of ski jumping trajectories. Therefore, a real-time Measurement System was used. The system consisted of wearable devices attached to the athletes and fixed ultra-wide band (UWB) antennas next to the jumping hill. To determine the accuracy and precision of the method, six athletes of the China A or B National Team performed 25 measured ski jumps. The method was used to measure the trajectory, velocity, acceleration and skis angles during the jump. The measurements are compared with camera measurements of a Markerless Human Movement Automatic Capture System to assess their accuracy. The test results demonstrate that the method has sufficient accuracy and reliability for ski jumping. Thus, the system can be used as a tracking system during training and competitions for coaches and sports scientists.

Taking the core engine test air intake system as the research object, through the numerical simulation calculation of the test intake device and the compressor inlet of a core engine, the influence of the selection of the test section of the intake device on the accuracy of the air flow field test is explored, and the influence factors of the compressor air inlet struts and probe on the test accuracy are analyzed. The results show that: the selection of different test sections of the air intake device, the uniformity of static pressure distribution, the unevenness of static pressure (no more than 1%) and flow coefficient are different, but the inlet flow pipe, the inlet probe of test section I and the inlet support plate have a certain impact on the inlet flow field of the compressor. The test uniformity should be considered when arranging the circumferential positions of the probes of the main test sections of the compressor, and the influence of the wake area of the probe of section I and the air inlet struts should be considered in space.

With the advent of the era of big data, distributed data centers with advantages in computing, storage, security, etc. have become the trend of future development. In the civil field, when distributed data centers are abnormal, problems such as service interruption, data leakage and capital loss will be caused. Therefore, it is necessary to predict failure for distributed data centers. In the military field, when the LVC system is applied to distributed data centers, the live assets will be lost if the simulation system fails, and the loss is very costly. Hence, failure prediction is more necessary for the simulation system. In summary, first, the failure prediction in distributed data centers is systematically reviewed from four aspects: overall architecture, data types, mainstream prediction methods, and existing difficulties and challenges, to analyze the current research status at home and abroad. Subsequently, from LVC system perspective, the importance and challenges of failure prediction for the simulation application layer in distributed data centers are analyzed.

Recognizing military coordination relationships among adversarial operations is essential in task planning and decision making. Due to the complexity of modern informationized war, it is challenging to analyze the diverse relations between entities in the war situation. In this study, we propose a novel framework based on knowledge graph to predict operational coordinations. We first construct a novel large scale knowledge graph that consits of 29313 nodes and 191542 edges from Wargame Competition dataset. The embedding method jointly considers information from node attributes, local situations and global structure, and then combine the three parts with a self-attention mechanism. Experiments compared with baselines demonstrate that the proposed model is more accurate and robust than existing methods.

To solve the problem that the COPERT model cannot be applied to heavy-duty diesel vehicle OBD monitoring data to obtain accurate NOx emission factors, we propose a ResNet amendment model based on historical driving time-frequency wavelet features. First, for multiple consecutive driving segments of the actual collected diesel vehicle single-vehicle trip data, the attributes with high correlation with the actual OBD emission factors were obtained using Spearman rank correlation analysis based on the data volume. Then, the highly correlated attributes and the emission factors predicted by the COPERT model were combined by constructing the historical information matrix and using the continuous wavelet transform for the time-frequency representation; Finally, the time-frequency representation was used as the input to complete the amendment of the COPERT model on the ResNet50 network. The experimental results show that the proposed method can effectively amend the COPERT model so that the COPERT model can be applied to OBD data to achieve accurate prediction of NOx emission factors.

Statistics show that about 70% of the surface mount quality problems are in the solder paste printing process, so it is particularly important to optimize the solder paste printing process parameters. The traditional parameter adjustment method still relies on the engineer’s own experience to manually adjust, which is inefficient. In recent years, artificial intelligence technology has achieved excellent performance in various cities, providing a new idea for the optimization of key process parameters of SMT production lines. Therefore, this paper proposes a method for optimizing process parameters of solder paste printing based on machine learning. It mainly includes two parts, one is the construction of the solder paste prediction model, and the other is the optimization algorithm based on the model prediction parameters. We choose the SVR model as the regressor of parameter prediction, and use the value predicted by the model as the evaluation standard, and use the PSO algorithm to optimize the search in the search space of process parameters. Finally, after experimental tests, the process parameters optimized by this method can well meet the actual production requirements and effectively improve the production efficiency.

Few-shot learning methods are studied for the problem of insufficient samples in neural network tasks. Taking advantage of the excellent feature extraction capabilities of neural networks, meta-learning was proposed and became the mainstream for few-shot learning. Among the few shot learning methods, the metric-based method has the characteristics of simple mode and no need for iterative training in the inference process, which is more in line with the original intention of few-shot learning task and has been widely studied. The metric-based method with Euclidean distance has established a relatively complete training and inference system, but the accuracy has gradually stagnated. The Earth Mover’s Distance has gradually emerged in the field of few-shot learning, and greatly surpassed Euclidean distance methods in terms of accuracy. Aiming at the problem that the current EMD method can only be trained with 1-shot learning and cannot be trained with multi-shot, we proposed a SeparateEMD that can train in multi-shot mode. In order to make up for the lack of global information of the SeparateEMD, Hierarchy Attention Module for proto is constructed to increase the intra-category correlation and intra-category global information of the support samples, and also improve the sample discrimination. We build the Half Pyramid Merge Module for query to increase the correlation between local information and global information of a single query sample. The Global Sampling is constructed on the basis of the Random Sampling of the EMD, which can not only increase the randomness of the samples, but also ensure the stability of the information. Experiments show that our proposed method can achieve 5-shot training of EMD-like method and outperform the current state-of-the-art few-shot learning methods on popular benchmark datasets.

The objective of this study is to improve the robustness of homography estimation using deep learning for images superimposed with various types of disturbances. In conventional deep learning homography estimation, the original image and the image with perturbations and disturbances are simultaneously input to the model for estimation. The disadvantages of this approach are that the original image is affected by noise and the model itself is unclear. In this study, features are extracted separately for each of the two images, and a model is constructed based on ResNet using these features as input. In addition, when extracting the features of the perturbed and disturbed images, WaveCNet, which integrates the discrete wavelet transform into the CNN, is used to add pinpoint tolerance to the disturbance. The estimation accuracy of the homography matrix in the method proposed in this study shows improved accuracy in various noises. These results suggest that the proposed method is effective in reducing the effect of disturbance by extracting features robust to disturbance for each image.

Tire quality plays an important role in traffic safety. Among the categories of defects that often occur in the actual production process of tires, the tire shoulder belt layer cords joint opening defect is the most common and serious defect. In this paper, a tire shoulder belt layer cords joint opening defect detection algorithm based on grayscale feature statistics and threading method is proposed based on the machine vision nondestructive testing technology. It first combines the periodic texture grayscale feature to accurately pre-locate the defect position, followed by performing a series of pre-processing operations on the target area to accurately determine the tire X-ray image. Through comparative experimental analysis, the detection algorithm has high recognition and accuracy. The detection speed of the algorithm has also reached a satisfactory level.

To reconstruct the 3D model of specific targets in real time, a multi-sensor data fusion-based 3D reconstruction algorithm is proposed in this paper. This network-based algorithm takes the camera image and lidar point-cloud data as inputs, employing RGB channel and lidar channel to process each type of data separately, and finally obtains the targets’ dense depth map by fusion. In RGB channel, the transformer network rather than CNN (convolutional neural network) is used to obtain multi-scale image features with global receptive field and high resolution, and generate monocular depth, guidance map and semantic segmentation. In the lidar channel, the sparse lidar data is fused with the guidance map to generate the final prediction of dense depth. In the test, our algorithm achieved a high ranking on the leaderboard. In application, under the condition of equal reconstruction quality, a five times faster speed is obtained in 3D reconstruction comparing to the traditional image-based method.

3D point cloud is an important expression of three-dimensional environment information. The registration of point cloud is the basis of realizing the functions such as localization, map construction and target detection based on 3D point cloud. For better registration, a method based on feature points and their spatial structure properties is proposed, which is divided into two stages: coarse registration and fine registration. Firstly, a feature point extraction network based on PointNet++ and Probabilistic Chamfer Loss is designed to extract robust and highly repetitive feature points. On the extracted feature points, the Super 4PCS method based on the spatial structure features of point cloud is used for global coarse registration. Then, taking the result obtained by coarse registration as the initial solution, the fine registration method based on NDT is used to register the point cloud more accurately. Finally, accurate registration results are obtained. This method combines the advantages of deep learning method and traditional method based on structure of point cloud, and makes full use of the spatial distribution properties of feature points. Experiments show that this method can achieve good registration results in a variety of application scenarios. And the method also shows the application potential of global location for initialization purpose in small-scale environments.

The application of deep reinforcement learning (DRL) for autonomous navigation of unmanned ground vehicle (UGV) has the problem of sparse rewards, which makes the trained algorithm model difficult to converge and cannot be transferred to real vehicles. In this regard, this paper proposes an effective exploratory learning autonomous navigation algorithm Double I-PPO, which designs pre-training behaviors based on imitation learning (IL) to guide UGV to try positive states, and introduces the intrinsic curiosity module (ICM) to generate intrinsic reward signals to encourage exploratory learning strategies. Build the training scene in Unity to evaluate the performance of the algorithm, and integrate the algorithm strategy into the motion planning stack of the ROS vehicle, so as to extend to the actual scene for testing. Experiments show that in the environment of random obstacles, the method does not need to rely on prior map information. Compared with similar DRL algorithms, the convergence speed is faster and the navigation success rate can reach more than 85%.

Digital archiving is the activity of digitizing cultural heritage to preserve and utilize it. Visualizing these data helps us to understand the complicated internal structure of the cultural heritage. Transparent stereoscopic visualization based on 3D point clouds is effective for understanding the complex structures of cultural heritage. However, the position and depth information often become unclear when three-dimensional data are rendered transparently. We examined whether perceived depth of transparently and stereoscopically visualized objects can be improved by highlighting 3D edges of the structure. We also investigated the effect of edge opacity on perceived depth. We tested two types of figures: vertically and horizontally combined shapes and diagonally combined shapes, which is extracted from the point cloud data of cultural heritage. We conducted psychophysical experiments, and the results suggest that edge highlighting improves the accuracy of perceived depth. Moreover, the effect of edge highlighting was more significant when binocular disparity and motion parallax were available.

There are few data sources for fault diagnosis of stepper motor motion control systems, conventional voltage, current, vibration signals cannot well characterize stepper motor faults, and the fault diagnosis capabilities of stepper motor controllers on the market are limited. In view of the problem that the stepper motor data source is limited, this paper uses the stepper motor load value to provide an effective data source for fault diagnosis and condition monitoring. The time series decomposition of motor load value data is carried out, the time series data is analyzed by GRU recurrent neural network algorithm, fault classification is performed, and a GRU stepper motor fault diagnosis model is constructed to conduct online self-diagnosis of intermittent vibration, instantaneous overload and connection loosening fault, and fault prediction is carried out on motor overload stall fault.