Parallel simulation test construction method based on multimode heterogeneous data cultivation

Abstract

The invention provides a parallel simulation test construction method based on multimode heterogeneous data cultivation, and belongs to the technical field of parallel simulation test construction. The method mainly comprises five technologies of task scene design, simulation system construction, simulation experiment design, cloud environment parallel simulation and capability verification analysis. The method can adapt to diversified task scenes, quick response and loop iteration can be performed on addition of additional information in a parallel test system, and feasibility, adaptability and integrity of a model operation process scheme can be reflected more accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of parallel simulation test construction, and in particular relates to a parallel simulation test construction method based on multi-mode heterogeneous data farming. Background Art

[0002] At present, the construction of parallel test system basically adopts linear process or "V"-shaped logical framework. The "V"-shaped logical framework is a design method that combines top-down decomposition with bottom-up comprehensive integration. The parallel test tasks are first decomposed, and standardized modeling is carried out on this basis. Finally, the bottom-up combination design generates the parallel test system model operation process. Through the repeated iteration of the parallel test system model operation process, designers continuously provide new concepts, strategies and technical solutions for optimization design, and finally form the best model operation plan.

[0003] The overall framework of this technology is shown in the attached Figure 1 As shown in the figure, the first step is to decompose the tasks from top to bottom, analyze the test tasks, and combine the additional information such as military strategies, combat regulations, and system establishment to decompose the activities of the parallel test business, and obtain the set of parallel test business activities. After the parallel test business activities are verified, the link modeling is carried out. Starting from the input, output, constraints, activities, applicable methods and other components of the process link, the standardized design is carried out, the basic link meta-model of the process is established, the basic links of the parallel test model operation process are extracted, and the model operation process link library is established to lay the foundation for the flexible construction of the test process. Then, the model operation process link is combined from bottom to top. Through the comprehensive integration method of system engineering theory, the various parts are gathered and combined according to certain rules to form a whole. The parallel test process construction requirements are analyzed, and the flexible link design is carried out according to the link meta-model. Finally, the parallel test model operation process is flexibly constructed. Finally, the parallel test model operation process is repeatedly iterated through verification and confirmation, and finally a parallel test model operation scheme with better effect is obtained.

[0004] The different LVC resources involved in the parallel test bring about complex and changeable command and operation environments for mission tasks. The V-shaped logical framework is essentially an iterative linear system model operation process. The operation model of this parallel test system (such as combat activity process, information activity process, system information interaction relationship, etc.) cannot adapt well to diverse mission scenarios, nor can it clearly reflect the relationship between operation efficiency and quality and external factors such as the natural environment and intelligent system compilation. These two points cannot be solved by static and linear system analysis. It is necessary to build a simulation system that meets realistic mission scenarios and conduct large-sample simulation experiments under scientific guidance to obtain quantitative data support. Summary of the invention

[0005] Based on actual needs, a parallel simulation experiment construction method based on multi-mode heterogeneous data cultivation is proposed to solve the problem that the existing parallel experiment construction technology cannot adapt to diverse task scenarios and its model operation process cannot reflect the relationship between operation efficiency and external factors.

[0006] The method includes five stages: task scenario design, simulation system construction, simulation experiment design, cloud environment parallel simulation, and capability verification analysis; among which:

[0007] In the task scenario design phase, for the information service architecture problem to be verified and analyzed, determine the benchmark scenario that can solve the problem to form a scenario design model;

[0008] In the simulation system construction stage, the architecture design model is used as input, combined with the scenario design model, and an executable simulation system is automatically constructed through the simulation system generation technology based on model fusion;

[0009] In the simulation experiment design phase, for the information service architecture problem to be verified and analyzed, the variable points in the executable simulation system are determined, and the experimental design table is generated through the experimental design strategy and converted into an executable simulation sample;

[0010] In the cloud environment parallel simulation stage, the executable simulation samples are run based on a high-performance computing environment or a distributed environment, and test data are collected and statistically analyzed to provide data support for capability verification analysis;

[0011] In the capability verification and analysis phase, based on the test data generated by parallel simulation, by applying data mining, data calculation and visualization analysis technology, the verification results of the architecture functional indicators and non-functional indicators for the information service architecture problem to be verified and analyzed are obtained;

[0012] The two stages of task scenario design and simulation system construction constitute a test definition cycle, and the three stages of simulation experiment design, cloud environment parallel simulation, and capability verification analysis constitute a multi-simulation operation cycle;

[0013] Wherein, if the verification result indicates that the scenario design model or the executable simulation system has a design defect, repeating the test definition cycle;

[0014] If the verification result indicates that the test data cannot solve the information service architecture problem to be verified and analyzed, the multiple simulation operation cycles are repeated.

[0015] In the above method, in the task scenario design stage: the benchmark scenario includes: scenario deployment, environment configuration, resource configuration, task decomposition, task planning and capability verification indicators of concern.

[0016] In the above method, during the task scenario design phase:

[0017] Collecting historical data, new data, and simulated data generated in the test definition cycle to analyze system behavior characteristics and identify key system parameters and variables;

[0018] Based on the key system parameters and variables, extract the features related to building the benchmark scenario, including environmental conditions, operating behaviors, and system status, and define the input and output of the benchmark scenario;

[0019] The features related to building the benchmark scenario are used for scenario parameterization to form a scenario design model, determine the initial state, and define the boundary conditions.

[0020] In the above method, in the simulation system construction stage: taking the architecture design model as input, combined with the scenario design model, through the simulation system generation technology based on model fusion, an initialized simulation system is automatically constructed, and the initialized simulation system is supplemented, refined and configured to form a complete executable simulation system, and the executable simulation system formed is not unique.

[0021] In the above method, during the simulation system construction phase:

[0022] When building an executable simulation system, data preprocessing is used to process and prepare the data required for simulation to ensure data quality and availability;

[0023] Based on the collected and pre-processed data, develop data-driven simulation models to simulate the behavior and evolution of real systems;

[0024] The constructed parallel simulation model is validated using historical data or experimental data to ensure that the model accurately reflects the behavior and characteristics of the system.

[0025] In the above method, in the simulation experiment design stage:

[0026] By analyzing the task scenarios and architecture design, we identify the factors related to the capability verification issues of concern, including changes in simulation system model parameters and changes in simulation system model structure;

[0027] Establish the system entity structure corresponding to the simulation model, mark each identified influencing factor in the system entity structure tree, and extract variables;

[0028] Through the design of the system entity structure and the extraction of variables, the type of each variable and the value space of the variable are determined, and the mapping relationship with the influencing factors is established;

[0029] The types of variables are divided into discrete variables and continuous variables. Discrete variables correspond to the parameters of the model or the structure of the model. The value spaces of different types of variables are marked using different methods.

[0030] In the above method, in the simulation experiment design stage:

[0031] Determine the experimental design method based on the variable data, variable type and value space, including orthogonal design, uniform design and Latin hypercube sampling; and perform:

[0032] Data seeding analysis: When designing simulation experiments, seed key test parameters and variables and analyze the relationship between data;

[0033] Parameter optimization: Based on the results of data analysis, optimize and adjust the parameters of the simulation test to ensure that the test produces positive results;

[0034] According to the determined test design method, the test variables are automatically processed to form an test design table. Through mapping, each group of variable values ​​is first mapped to a trimmed system entity structure tree, and then the system entity structure tree is mapped to a simulation model sample, thereby forming an architecture simulation sample set, which contains several executable simulation samples.

[0035] In the above method, in the cloud environment parallel simulation stage:

[0036] When performing parallel simulation in a cloud environment, large-scale simulation data is analyzed through big data processing, and the simulation process is accelerated using a distributed computing framework;

[0037] Monitor and analyze the real-time data generated during the parallel simulation process, and dynamically adjust the simulation parameters and configurations based on the analysis results.

[0038] In the above method, during the proficiency testing analysis phase:

[0039] Use data mining techniques to discover hidden information and patterns in simulation data to evaluate the performance and accuracy of simulation systems;

[0040] Use statistical analysis to evaluate and validate simulation results, including hypothesis testing and variance analysis, to determine whether the simulation system meets the expected performance requirements.

[0041] In summary, the technical solution of the present invention can adapt to various task scenarios, can quickly respond and iterate the addition of additional information in the parallel test system, and can more accurately reflect the feasibility, adaptability and integrity of the model operation process solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 Flow chart of parallel test construction technology implementation in the prior art;

[0044] Figure 2 The overall architecture diagram of the parallel simulation test construction technology based on data farming in the present invention;

[0045] Figure 3 Schematic diagram of the experimental design and sample generation process in the present invention;

[0046] Figure 4 Technical flow chart of Example 1 of the present invention;

[0047] Figure 5 A flowchart of three-dimensional scene construction in Example 1 of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] As an effective means of coordinating various types of test resources, the parallel test system can scientifically guide parallel test activities and is a reliable guarantee for the orderly and efficient implementation of the test. Aiming at the problem that the current linear and "V"-shaped parallel test system model operation technology cannot adapt to diverse task scenarios and cannot reflect the relationship between operation efficiency and external factors, the present invention proposes a parallel simulation test construction method based on data cultivation. Create multiple assumptions based on the perspective boundary and perform a "data-in-the-loop" cyclic simulation.

[0050] The technical solution proposed in this invention mainly includes five technologies: task scenario design, simulation system construction, simulation experiment design, cloud environment parallel simulation, capability verification analysis, etc. It is mainly used in military joint simulation experiments, and can build stable and accurate parallel simulation experiments to adapt to diverse task scenarios, quickly respond to and iterate the addition of additional information in the parallel experiment system, and more accurately reflect the feasibility, adaptability and integrity of the model operation process solution.

[0051] This invention is based on the idea of ​​data farming, and its basic idea is "loop in loop". Based on the uncertain causal relationship in the system, multiple scenarios are created according to the perspective boundary to perform "data in the loop" loop simulation. In a broad sense, data farming is a specific set of methodologies. Through result feedback, dynamic exploration, high-performance computing and other technologies, a large amount of data is obtained in the field, the uncertain factors in the simulation are refined, and visualization technology is used to obtain efficient, flexible and adaptable scenario solutions from the model.

[0052] The technical process of the present invention is as shown in the attached Figure 2 As shown, it mainly includes five stages: task scenario design, simulation system construction, simulation experiment design, cloud environment parallel simulation, and capability verification analysis.

[0053] In the task scenario design phase, for the information service architecture problem to be verified and analyzed, determine the benchmark scenario that can solve the problem to form a scenario design model;

[0054] In the simulation system construction phase, the architecture design model is used as input, combined with the scenario design model, and an executable simulation system is automatically constructed through the simulation system generation technology based on model fusion;

[0055] In the simulation experiment design phase, for the information service architecture problem to be verified and analyzed, the variable points in the executable simulation system are determined, and the experimental design table is generated through the experimental design strategy and converted into an executable simulation sample;

[0056] In the cloud environment parallel simulation stage, the executable simulation samples are run based on a high performance computing environment or a distributed environment, and test data are collected and statistically analyzed to provide data support for capability verification analysis;

[0057] In the capability verification and analysis phase, based on the test data generated by parallel simulation, by applying data mining, data calculation and visualization analysis technology, the verification results of the architecture functional indicators and non-functional indicators for the information service architecture problem to be verified and analyzed are obtained;

[0058] The two stages of task scenario design and simulation system construction constitute a test definition cycle, and the three stages of simulation experiment design, cloud environment parallel simulation, and capability verification analysis constitute a multi-simulation operation cycle;

[0059] Wherein, if the verification result indicates that the scenario design model or the executable simulation system has a design defect, repeating the test definition cycle;

[0060] If the verification result indicates that the test data cannot solve the information service architecture problem to be verified and analyzed, the multiple simulation operation cycles are repeated.

[0061] Specifically, solution verification is a process of continuous iterative processing, and different activities can be seamlessly linked. Through the parallel test system construction technology and task scenario design technology, a "test definition loop" is formed to construct an effective architecture simulation system for the issues of concern to the test evaluators. Through the iteration of this process, the constructed architecture simulation system can also be refined so that all participants can understand the scope and intention of the simulation model, ensuring that the constructed model can well describe the problem being studied. Then, the "multi-simulation operation loop" composed of test design and sample generation technology, cloud environment parallel simulation technology and capability verification analysis technology is entered. During the execution of this loop, the test evaluators determine which scenarios or model parameters should be checked and explored, and generate samples for large-scale simulation through test design technology. Then, by implementing parallel and efficient execution of simulation samples in the cloud environment, simulation test data is obtained, and the issues of concern are verified and analyzed through data mining, data analysis and its visualization method. If it is found that the data cannot meet the requirements, the "multi-simulation operation loop" is repeated. If problems are found in the scenario or architecture design through verification and analysis, return to the "test definition loop". The following introduces the specific process of each stage of the command and dispatch test based on the idea of ​​data cultivation.

[0062] In the above method, in the task scenario design stage: the benchmark scenario includes: scenario deployment, environment configuration, resource configuration, task decomposition, task planning and capability verification indicators of concern.

[0063] Specifically, for the information service architecture problem to be verified and analyzed, a benchmark scenario that can answer the problem is determined, including deployment, environment configuration, resource configuration, task decomposition (user types), task planning (usage process for each type of user) and capability verification indicators of concern. This benchmark scenario is one of the important bases for generating and configuring the simulation system and determining the test data collection requirements.

[0064] In the above method, during the task scenario design phase:

[0065] Collecting historical data, new data, and simulated data generated in the test definition cycle to analyze system behavior characteristics and identify key system parameters and variables;

[0066] Based on the key system parameters and variables, extract the features related to building the benchmark scenario, including environmental conditions, operating behaviors, and system status, and define the input and output of the benchmark scenario;

[0067] The features related to building the benchmark scenario are used for scenario parameterization to form a scenario design model, determine the initial state, and define the boundary conditions.

[0068] Specifically, the idea of ​​data farming is mainly reflected in:

[0069] Data collection and analysis: In the scenario design phase, historical data and new or simulated data generated in the cycle are collected to decompose the behavior and characteristics of the parallel system. By analyzing the data, the key parameters and variables of the system and the relationship between them are identified.

[0070] Feature extraction and selection: Based on the results of data analysis, features related to building parallel simulation task scenarios can be extracted, such as environmental conditions, operating behaviors, system status, etc., to help define the input and output of the simulation task scenarios.

[0071] Scenario parameterization: The features extracted from the data are used for scenario parameterization, such as determining the initial state of the simulation model, defining the boundary conditions of the simulation, etc.

[0072] In the above method, in the simulation system construction stage: taking the architecture design model as input, combined with the scenario design model, through the simulation system generation technology based on model fusion, an initialized simulation system is automatically constructed, and the initialized simulation system is supplemented, refined and configured to form a complete executable simulation system, and the executable simulation system formed is not unique.

[0073] Specifically, taking the architecture design model as input and combining it with the scenario design model, the proposed simulation system generation technology based on model fusion is used to automatically build an initialized simulation system, and the simulation system is supplemented, refined and configured as necessary to form a complete and executable simulation system. The reason for supplementation, refinement and configuration is mainly because the simulation system generated by fusion may be incomplete. Through this step, a benchmark simulation model that can carry out simulation experiments is constructed. It should be noted that according to the results of the task scenario design, the benchmark simulation model formed may not be unique.

[0074] In the above method, during the simulation system construction phase:

[0075] When building an executable simulation system, data preprocessing is used to process and prepare the data required for simulation to ensure data quality and availability;

[0076] Based on the collected and pre-processed data, develop data-driven simulation models to simulate the behavior and evolution of real systems;

[0077] The constructed parallel simulation model is validated using historical data or experimental data to ensure that the model accurately reflects the behavior and characteristics of the system.

[0078] Specifically, the idea of ​​data farming is mainly reflected in:

[0079] Data preprocessing: When building a parallel simulation system, data preprocessing is used to process and prepare the data required for simulation to ensure the quality and availability of the data.

[0080] Model development: Based on the collected and pre-processed data, data-driven simulation models are developed to simulate the behavior and evolution of the actual system.

[0081] Model validation: Use historical data or experimental data to validate the constructed parallel simulation model to ensure that the model can accurately reflect the behavior and characteristics of the system.

[0082] In the above method, in the simulation experiment design stage:

[0083] By analyzing the task scenarios and architecture design, we identify the factors related to the capability verification issues of concern, including changes in simulation system model parameters and changes in simulation system model structure;

[0084] Establish the system entity structure corresponding to the simulation model, mark each identified influencing factor in the system entity structure tree, and extract variables;

[0085] Through the design of the system entity structure and the extraction of variables, the type of each variable and the value space of the variable are determined, and the mapping relationship with the influencing factors is established;

[0086] The types of variables are divided into discrete variables and continuous variables. Discrete variables correspond to the parameters of the model or the structure of the model. The value spaces of different types of variables are marked using different methods.

[0087] Specifically, for the problem to be studied, the variable points in the simulation model are determined, and through a scientific experimental design strategy, an experimental design table for the research problem is generated and converted into an executable simulation sample set, so as to achieve the purpose of effectively supporting the analysis of the research problem and minimizing the simulation samples, thereby improving the efficiency of the architectural design test verification process.

[0088] How to use as few simulation samples as possible to improve the efficiency of simulation experiments while considering all variable point factors as much as possible is one of the keys to architecture verification. To address this problem, the present invention proposes the following Figure 3 The experimental design and sample generation process are shown.

[0089] First, through the analysis of task scenarios and architecture design, the elements related to the capability verification issues of concern are identified. These elements can be changes in the parameters of the simulation system model or changes in the structure of the simulation system model. On the basis of establishing the system entity structure (SES) corresponding to the simulation model, each key influencing factor identified is marked in the system entity structure tree, and then the variables are extracted. Through the design of SES and the extraction of variables, the type of each variable and the value space of the variable can be determined, and a mapping relationship with the influencing factors can be established. The types of variables are divided into discrete variables and continuous variables. Discrete variables can correspond to the parameters of the model or the structure of the model. The value spaces of different types of variables are marked using different methods.

[0090] In the above method, in the simulation experiment design stage:

[0091] Determine the experimental design method based on the variable data, variable type and value space, including orthogonal design, uniform design and Latin hypercube sampling; and perform:

[0092] (1) Data seeding analysis: When designing simulation experiments, key experimental parameters and variables are seeded and the relationship between the data is analyzed;

[0093] (2) Parameter optimization: Based on the results of data analysis, the parameters of the simulation test are optimized and adjusted to ensure that the test produces positive results;

[0094] According to the determined test design method, the test variables are automatically processed to form an test design table. Through mapping, each group of variable values ​​is first mapped to a trimmed system entity structure tree, and then the system entity structure tree is mapped to a simulation model sample, thereby forming an architecture simulation sample set, which contains several executable simulation samples.

[0095] Specifically, determine the appropriate experimental design method based on the variables and their value space. There are many experimental design methods at present, including orthogonal design, uniform design, Latin hypercube sampling, etc. Each design method has different applicability and effects. Therefore, it is necessary to determine a reasonable experimental design method based on the variable data, variable type and value space, so as to better verify and analyze. One of the important tasks of this project is to establish an optimized experimental design method selection strategy, which can independently decide to select a single or combined experimental design method based on the current experimental variables. The idea of ​​data farming is mainly reflected in:

[0096] Data seeding analysis: When designing simulation experiments, seed key experimental parameters and variables and analyze the relationship between the data.

[0097] Parameter optimization: Based on the results of data analysis, the parameters of the simulation test are optimized and adjusted to ensure that the test can produce meaningful results.

[0098] Finally, according to the algorithm of the determined actual experimental method, the experimental variables are automatically processed to form an experimental design table. Through mapping, each set of variable values ​​is first mapped to a trimmed system entity structure tree, and then the system entity structure tree is mapped to a simulation model sample, thereby forming an architecture simulation sample set.

[0099] In the above method, in the cloud environment parallel simulation stage:

[0100] When performing parallel simulation in a cloud environment, large-scale simulation data is analyzed through big data processing, and the simulation process is accelerated using a distributed computing framework;

[0101] Monitor and analyze the real-time data generated during the parallel simulation process, and dynamically adjust the simulation parameters and configurations based on the analysis results.

[0102] Specifically, when the research problem involves many changing factors, the number of simulation samples generated is often huge, and due to the existence of randomness in the simulation model, each simulation sample needs to be run multiple times. Therefore, it is necessary to rely on high-performance computing environments or distributed environments to achieve efficient parallel operation of simulation samples, and collect and statistically analyze test data to provide data support for verification and analysis. The idea of ​​data farming is mainly reflected in:

[0103] Big data processing: When performing parallel simulation in a cloud environment, big data processing analyzes large-scale simulation data, such as using a distributed computing framework to accelerate the simulation process.

[0104] Real-time data analysis: Monitor and analyze the data generated during the parallel simulation process, and dynamically adjust the simulation parameters and configurations based on the analysis results.

[0105] In the above method, during the proficiency testing analysis phase:

[0106] Use data mining techniques to discover hidden information and patterns in simulation data to evaluate the performance and accuracy of simulation systems;

[0107] Use statistical analysis to evaluate and validate simulation results, including hypothesis testing and variance analysis, to determine whether the simulation system meets the expected performance requirements.

[0108] Specifically, based on the test data generated by efficient parallel simulation, by applying data mining, data calculation and analysis, and visualization technology, we can obtain the evaluation results of the functional and non-functional indicators of the software architecture in different situations, and then support the verification of whether the designed airport command and dispatch architecture can meet the mission requirements, and support the analysis of the relationship between different factors. The idea of ​​data cultivation is reflected in:

[0109] Data Mining: During the capability verification analysis phase, data mining techniques are used to mine hidden information and patterns in simulation data to evaluate the performance and accuracy of the simulation system.

[0110] Statistical analysis: Use statistical analysis to evaluate and verify simulation results, such as hypothesis testing, variance analysis, etc., to determine whether the simulation system meets the expected performance requirements.

[0111] Example 1

[0112] The parallel simulation test construction method based on data farming proposed in the present invention is used to construct a parallel system model operation test of a UAV.

[0113] The flow chart of the embodiment is as follows Figure 4 As shown in the figure, firstly, task decomposition and task planning are carried out according to the specific test tasks, then the environment deployment is carried out, and finally the indicators are verified. The benchmark simulation model is generated according to the specific tasks of the UAV.

[0114] The sample set is generated by finding the variable points of the model. When a general drone conducts optical reconnaissance, it first collects video images of the flight scene area through the optical pod, and then processes the video images based on the visual perception algorithm to obtain the battlefield situation.

[0115] In the town space, each frame of the optical pod video image is usually determined by the position, posture and camera parameters of the optical pod (as shown in Table 1), where the camera parameters generally include focal length, image plane width, and resolution. The optical pod is generally fixedly mounted on the body, and its coordinates in the body coordinate system are expressed as, and its optical axis generally has two deflection degrees of freedom in azimuth and pitch relative to the body.

[0116] In addition, the UAV has a degree of freedom attitude relative to the earth, namely the coordinates of the center of mass in the earth coordinate system and the three attitude angles of heading, pitch, and roll. During simulation, the sample variable points are usually selected as the parameters listed in the table below. After a series of coordinate transformations and assignments to the input parameters of the viewport rendering program interface, each frame of the image can be obtained through graphics rendering.

[0117] Table 1

[0118]

[0119]

[0120] A three-dimensional scene generally includes terrain, vegetation, roads, water systems, sky, buildings, artificial facilities, lighting, rainfall, cloud cover, wind speed, waves and other content. Among them, the terrain can be randomly generated according to the scene editor or GIS data and its texture map can be imported, and the terrain generation tool can be used to generate the terrain grid; vegetation, sky, water system, roads and other content can be drawn through the corresponding tools of the scene editor; buildings, artificial facilities and the three-dimensional display model of the drone body are first designed through three-dimensional modeling software, and then the data is saved; environmental data such as lighting, rainfall, cloud cover, wind speed and waves are set by writing scene scripts. The construction process of the three-dimensional scene is shown in the attached figure. Figure 5 shown.

[0121] After the above design, the sample set required for simulation is generated. Through the execution of simulation samples, the simulation data is collected and stored, and the results are visualized. Then, the collected data is mined, and the obtained data is used as additional information as input to redeploy the environmental planning and update the model.

[0122] Example 2

[0123] The parallel simulation test construction device based on data farming includes five modules: task scenario design, simulation system construction, simulation experiment design, cloud environment parallel simulation, and capability verification analysis; wherein the task scenario design module and the simulation system construction module constitute an experiment definition loop; the simulation experiment design, cloud environment parallel simulation and capability verification analysis constitute a multi-simulation execution loop.

[0124] The task scenario design module is used to plan tasks; the simulation system construction module is used to automatically build an initialized simulation system; the simulation experiment design generates an experimental design table for the research problem and converts it into an executable simulation sample set; the cloud environment parallel simulation module is used for simulation sample execution and management; the capability verification analysis module is used for system capability verification and result display.

[0125] The task scenario design module specifically includes: task decomposition sub-module, task planning sub-module, deployment environment planning sub-module, and verification indicator determination sub-module.

[0126] The simulation system building modules specifically include: LVC model conversion and fusion sub-module, dedicated model development sub-module, model combination configuration sub-module, and model management sub-module.

[0127] The simulation experiment design module specifically includes: a variable point design submodule, a design strategy selection submodule, and a sample automatic generation submodule.

[0128] The cloud environment parallel simulation module specifically includes: simulation sample execution submodule, data acquisition submodule, and simulation execution control submodule.

[0129] The capability verification analysis module specifically includes: data preprocessing submodule, data mining analysis submodule, system capability verification submodule, and result display analysis submodule.

[0130] Example 3

[0131] The parallel simulation experiment construction method based on data farming includes the following steps:

[0132] Step 1: For the information service architecture problem to be verified and analyzed, determine the benchmark scenario that can answer the problem, including deployment, environment configuration, resource configuration, task decomposition (user types), task planning (user usage process for each type of user) and capability verification indicators of concern;

[0133] Step 2: Taking the architecture design model as input and combining it with the scenario design model, the initialized simulation system model is automatically constructed through the proposed real system generation technology based on model fusion;

[0134] Step 3: Define loops through experiments and continuously improve the simulation system model;

[0135] Step 4: Construct a path from the experiment definition loop to the multi-simulation execution loop through model transfer;

[0136] Step 5: Determine the variable points in the simulation model for the problem to be studied, generate an experimental design table for the research problem through a scientific experimental design strategy, and convert it into an executable simulation sample set;

[0137] Step 6: Rely on high-performance computing environment or distributed environment to realize efficient parallel operation of simulation samples and collect and count test data;

[0138] Step 7: Based on the test data generated by efficient parallel simulation, by applying data mining, data calculation and analysis, and visualization technology, the evaluation results of the functional indicators and non-functional indicators of the software architecture concerned in different situations are obtained;

[0139] Step 8: Continuously improve simulation capabilities and results through multiple simulation execution loops;

[0140] Step 9: Construct a path from the multi-simulation execution loop to the experimental definition loop through simulation result feedback.

[0141] In summary, the technical solution proposed in the present invention can adapt to various task scenarios, and can quickly respond to and iterate the addition of additional information in the operation of the parallel test system model, without having to verify and iterate after the complete model construction process is executed as in the traditional linear process. At the same time, the idea of ​​data farming can reflect the relationship between operating efficiency and external factors, and can more accurately reflect the feasibility, adaptability and integrity of the system model operation process plan, and build an efficient and stable parallel test.

[0142] Please note that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several variations and improvements can be made, which all belong to the scope of protection of the present application. Therefore, the scope of protection of the patent in this application shall be based on the attached claims.

Claims

1. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming, characterized in that: The method includes five stages: task scenario design, simulation system construction, simulation experiment design, cloud environment parallel simulation, and capability verification analysis; among which: In the task scenario design phase, for the information service architecture problem to be verified and analyzed, determine the benchmark scenario that can solve the problem to form a scenario design model; In the simulation system construction phase, the architecture design model is used as input, combined with the scenario design model, and an executable simulation system is automatically constructed through the simulation system generation technology based on model fusion; In the simulation experiment design phase, for the information service architecture problem to be verified and analyzed, the variable points in the executable simulation system are determined, and the experimental design table is generated through the experimental design strategy and converted into an executable simulation sample; In the cloud environment parallel simulation stage, the executable simulation samples are run based on a high performance computing environment or a distributed environment, and test data are collected and statistically analyzed to provide data support for capability verification analysis; In the capability verification and analysis phase, based on the test data generated by parallel simulation, by applying data mining, data calculation and visualization analysis technology, the verification results of the architecture functional indicators and non-functional indicators for the information service architecture problem to be verified and analyzed are obtained; The two stages of task scenario design and simulation system construction constitute a test definition cycle, and the three stages of simulation experiment design, cloud environment parallel simulation, and capability verification analysis constitute a multi-simulation operation cycle; Wherein, if the verification result indicates that the scenario design model or the executable simulation system has a design defect, repeating the test definition cycle; If the verification result indicates that the test data cannot solve the information service architecture problem to be verified and analyzed, the multiple simulation operation cycles are repeated.

2. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 1, characterized in that: In the mission scenario design phase: the benchmark scenario includes: scenario deployment, environment configuration, resource configuration, task decomposition, task planning and focused capability verification indicators.

3. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 2, characterized in that: During the mission scenario design phase: Collecting historical data, new data, and simulated data generated in the test definition cycle to analyze system behavior characteristics and identify key system parameters and variables; Based on the key system parameters and variables, extract the features related to building the benchmark scenario, including environmental conditions, operating behaviors, and system status, and define the input and output of the benchmark scenario; The features related to building the benchmark scenario are used for scenario parameterization to form a scenario design model, determine the initial state, and define the boundary conditions.

4. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 3, characterized in that: In the simulation system construction stage: taking the architecture design model as input, combined with the scenario design model, the initialized simulation system is automatically constructed through the simulation system generation technology based on model fusion, and the initialized simulation system is supplemented, refined and configured to form a complete executable simulation system, and the executable simulation system formed is not unique.

5. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 4, characterized in that: During the simulation system construction phase: When building an executable simulation system, data preprocessing is used to process and prepare the data required for simulation to ensure data quality and availability; Based on the collected and pre-processed data, develop data-driven simulation models to simulate the behavior and evolution of real systems; The constructed parallel simulation model is validated using historical data or experimental data to ensure that the model accurately reflects the behavior and characteristics of the system.

6. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 5, characterized in that: In the simulation experiment design stage: By analyzing the task scenarios and architecture design, we identify the factors related to the capability verification issues of concern, including changes in simulation system model parameters and changes in simulation system model structure; Establish the system entity structure corresponding to the simulation model, mark each identified influencing factor in the system entity structure tree, and extract variables; Through the design of the system entity structure and the extraction of variables, the type of each variable and the value space of the variable are determined, and the mapping relationship with the influencing factors is established; The types of variables are divided into discrete variables and continuous variables. Discrete variables correspond to the parameters of the model or the structure of the model. The value spaces of different types of variables are marked using different methods.

7. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 6, characterized in that: In the simulation experiment design stage: Determine the experimental design method based on the variable data, variable type and value space, including orthogonal design, uniform design and Latin hypercube sampling; And execute: Data seeding analysis: When designing simulation experiments, seed key test parameters and variables and analyze the relationship between data; Parameter optimization: Based on the results of data analysis, optimize and adjust the parameters of the simulation test to ensure that the test produces positive results; According to the determined test design method, the test variables are automatically processed to form an test design table. Through mapping, each group of variable values ​​is first mapped to a trimmed system entity structure tree, and then the system entity structure tree is mapped to a simulation model sample, thereby forming an architecture simulation sample set, which contains several executable simulation samples.

8. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 7, characterized in that: In the cloud environment parallel simulation stage: When performing parallel simulation in a cloud environment, large-scale simulation data is analyzed through big data processing, and the simulation process is accelerated using a distributed computing framework; Monitor and analyze the real-time data generated during the parallel simulation process, and dynamically adjust the simulation parameters and configurations based on the analysis results.

9. A parallel simulation experiment construction method based on multi-mode heterogeneous data farming according to claim 8, characterized in that: During the proficiency testing analysis phase: Use data mining techniques to discover hidden information and patterns in simulation data to evaluate the performance and accuracy of simulation systems; Use statistical analysis to evaluate and validate simulation results, including hypothesis testing and variance analysis, to determine whether the simulation system meets the expected performance requirements.

Images