Digital-real fusion test total-factor multi-scene acceleration simulation method

Abstract

The invention discloses a total-factor multi-scene acceleration simulation method for a digital-real fusion test, which belongs to the field of digital test verification of high-end equipment, and comprises the following steps: analyzing a test object, proposing a test demand / target and a performance index, and formulating a preliminary test implementation scheme; according to an analysis result, constructing a geometric model and a mechanism model of the test object; performing multi-dimensional model joint simulation operation to realize multi-dimensional simulation data acquisition and acquisition; carrying out physical data acquisition and historical data collection, and combining simulation data to realize fusion and enhancement of data real test data; carrying out dimension reduction on high-dimension data by utilizing a data dimension reduction method so as to extract key features; based on the obtained dimension reduction data and the test object model, constructing a deep learning simulation acceleration network and carrying out training and optimization; and inputting a test object total-factor simulation model and a test full-scene model to realize total-factor multi-scene simulation acceleration and result prediction. According to the invention, total-factor multi-scene accelerated simulation is realized, and the test efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the field of digital testing and verification of high-end equipment, and specifically relates to a full-factor multi-scenario accelerated simulation method for digital-physical fusion testing. Background Art

[0002] With the development of technology, the functions of complex equipment are becoming more complex and the performance indicators are becoming more advanced. In order to achieve the corresponding functions and performance, the demand for complex equipment testing is growing. Although the traditional physical testing method has high accuracy, the testing cost and time are high. Although the pure digital simulation method can shorten the test time to a certain extent, it is usually limited by the computing power of the simulation solver and the processing efficiency of high-dimensional data, resulting in too long simulation time in complex scenarios.

[0003] At present, the research on digital-physical fusion testing has gradually developed, and by combining digital simulation with physical test data, a more accurate test object model can be constructed. However, how to efficiently use simulation data and physical data for fusion, and on this basis, to achieve simulation acceleration through deep learning networks, is still a challenge currently faced. Among the existing methods, the research on dimensionality reduction and accelerated simulation prediction of high-dimensional simulation data is relatively limited, and it is difficult to achieve good scalability in multi-domain model simulation. Summary of the invention

[0004] The present invention discloses a multi-scenario accelerated simulation system and method for digital-real fusion testing, which aims to solve the problems of long simulation time and difficult data fusion in the prior art. Under the premise of ensuring the test accuracy, the accelerated simulation solution capability of the corresponding test object is realized, and the existing test multi-scenario model is integrated to greatly improve the efficiency and credibility of equipment testing.

[0005] The present invention solves the technical problem by adopting the following technical solution: a method for accelerating simulation of all-factors and multiple scenarios in digital-real fusion testing, the method comprising:

[0006] Step 1: Design a test object element model construction module. This construction module implements the model construction of all elements for the test object. The specific implementation is as follows:

[0007] Analyze the test object, propose test requirements or goals, performance indicators, and formulate a preliminary test implementation plan, including detailed decomposition of the test objectives; based on the analysis results, construct a geometric model that reflects the geometric shape characteristics of the test object and a multi-domain mechanism model of the test object's operation process;

[0008] Step 2: Design a digital-real data fusion and enhancement module for the test object element model. The digital-real data fusion and enhancement module uses the constructed test object element model and physical data to achieve data fusion and enhancement. The specific implementation is as follows:

[0009] Conduct joint simulation of geometric models and mechanism models to achieve multi-dimensional simulation data acquisition and collection; conduct physical data acquisition and historical data collection, and combine simulation data to achieve the fusion and enhancement of digital-physical fusion test data; use data dimensionality reduction methods to reduce the dimensionality of high-dimensional data to extract key features and obtain data sets;

[0010] Step 3: Design a test object factor scenario model joint simulation acceleration module. The joint simulation acceleration module combines the full-factor simulation acceleration model and the full-scenario model required for the test to achieve joint simulation acceleration of the test. The specific implementation is as follows:

[0011] For the geometric model of the test object, the signed distance field is used to perform a variant representation of the geometric model; the geometric model, signed distance field, and various operating parameters are input as training input data sets to the deep learning simulation acceleration network for training and optimization; after the training, the deep learning simulation acceleration network accepts the data input after dimensionality reduction and predicts the simulation results; combined with the online learning mechanism, it automatically adapts to the characteristics of different test objects, and dynamically optimizes the simulation process while maintaining the simulation accuracy; finally, the full-factor deep simulation acceleration model and the test full-scene model are input to realize the joint simulation acceleration and result prediction of the full-factor and full-scene model of the digital-real fusion test.

[0012] The present invention has the following beneficial effects: the method disclosed in the present invention can greatly shorten the modeling and simulation time of the test object and improve the test efficiency. At the same time, through the application of deep learning networks, the simulation speed can be significantly accelerated while ensuring the simulation accuracy, thereby providing an efficient and fast solution for the testing of complex equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of a method for accelerating simulation of all-factors and multiple scenarios in digital-real fusion testing of the present invention. DETAILED DESCRIPTION

[0014] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

[0015] The present invention discloses a method for accelerating simulation of all-factors and multiple scenarios in digital-real fusion testing. Figure 1 As shown, including:

[0016] Step 1: Design a test object element model construction module. This module implements the model construction of all elements for the specific application object of the digital-real fusion test. The specific implementation is as follows:

[0017] Analyze the test object, propose test requirements / goals, performance indicators, and formulate a preliminary test implementation plan, including detailed decomposition of the test objectives; based on the analysis results, construct a geometric model that reflects the geometric shape characteristics of the test object and a multi-domain mechanism model of the test object's operation process;

[0018] Step 2: Design the test object factor model digital and physical data fusion and enhancement module. This module uses the constructed full factor model and physical data to achieve data fusion and enhancement. The specific implementation is as follows:

[0019] Conduct joint simulation of multi-dimensional models including geometric models and mechanism models to achieve multi-dimensional simulation data acquisition and collection; conduct physical data acquisition and historical data collection, and combine simulation data to achieve the fusion and enhancement of digital-physical fusion test data; use data dimensionality reduction methods to reduce the dimensionality of high-dimensional data to extract key features and obtain data sets;

[0020] Step 3: Design a test object element scenario model joint simulation acceleration module. This module combines the object element simulation acceleration model and the full scenario model required for the test to achieve joint simulation acceleration of the test. The specific implementation is as follows:

[0021] Based on the acquired data set, the signed distance field is used to convert and represent the data in multiple dimensions. The data is input into the constructed deep learning simulation acceleration network for training and optimization. After training, the deep learning simulation acceleration network can accept the input of data after dimensionality reduction, quickly predict the simulation results, and replace the traditional numerical simulation solver. Combined with the online learning mechanism, it can automatically adapt to the characteristics of different test objects, and dynamically optimize the simulation process while maintaining the simulation accuracy. Finally, the multi-domain mechanism model of the geometric model and the test object operation process and the test full-scene model are input to realize the joint simulation acceleration and result prediction of the full-factor and full-scene model of the digital-real fusion test.

[0022] The multi-domain mechanism model of the test object is jointly constructed by fluid mechanics modeling, structural mechanics modeling, parametric modeling and modeling methods based on two-dimensional modeling language: the fluid mechanics model is based on incompressible fluid model, turbulence model, heat conduction model, etc.; the structural mechanics model is constructed by finite element method (FEM), including nonlinear behavior simulation and dynamic response prediction; parametric modeling includes adaptive adjustment of geometric shape and automatic calibration of simulation parameters; modeling tools based on two-dimensional modeling language are used for coupled simulation of different physical fields to achieve accurate simulation of multiple physical fields.

[0023] The multi-dimensional model joint simulation including geometric models and mechanism models is realized in the following ways: using the general simulation interface (FMI standard) to call different types of models; realizing the interconnection of modules through communication protocols (such as TCP / IP, OPC UA); using the model exchange method to realize data exchange based on simulation models between multiple platforms.

[0024] The fusion of digital-physical test data specifically includes: eliminating noise and error through joint processing of simulation data and physical acquisition data; extracting features of multi-source data to establish unified feature representation of data fusion; and realizing test data result fusion through joint analysis of simulation results and actual physical data.

[0025] The data dimensionality reduction method specifically includes the following steps: extracting main patterns from the simulation data set and decomposing high-dimensional data into a limited number of pattern vectors; retaining the main change patterns and reducing the computational complexity.

[0026] The deep learning simulation acceleration network uses the reduced-dimensional simulation data and model data as data sets to train the constructed deep neural network structure. It adjusts the network parameters and constructs the network's loss function in combination with the physical mechanism. It also ensures the stability of the deep learning network through multiple training and iterations.

[0027] More specifically, the following illustrative examples are given.

[0028] First, the test object is analyzed in detail, including the test requirements, the definition of performance indicators and the determination of test objectives, and a preliminary test plan is formulated. The test object analysis is the basis of the entire process, and its purpose is to refine the test objectives and provide targeted solutions to ensure the rationality of the test plan and the accuracy of execution.

[0029] Next, the geometric model and mechanism model are constructed based on the characteristics of the test object. The geometric model provides a spatial structure basis for physical simulation, while the mechanism model combines knowledge from multiple fields of physics, including fluid mechanics, structural mechanics, etc., to provide a more realistic description of physical behavior for subsequent simulation operations.

[0030] In addition, the present invention also adopts parametric modeling technology, which can achieve model optimization and automatic calibration in various scenarios through adaptive adjustment of geometric shapes and physical parameters, thereby improving the versatility and accuracy of the model.

[0031] In order to improve the efficiency of model simulation, the invention adopts the joint simulation operation of multi-dimensional models, and ensures the interoperability of simulation models in different environments through data exchange and communication protocol connection between multiple simulation tools. This joint simulation method utilizes the universal simulation interface (FMI standard) and communication protocols (such as TCP / IP, OPC UA) to achieve efficient interconnection and calling of simulation modules, greatly enhancing the flexibility and expansion capabilities of simulation. Using cross-platform simulation technology, the compatibility and scalability between different simulation tools are improved.

[0032] In the process of acquiring simulation data, the system also collects physical test data and historical data at the same time, and realizes the joint processing of simulation data, physical data and historical data through multi-source data fusion. Digital-physical fusion not only improves the reliability of data by removing noise and errors, but also further improves the accuracy of the test through data fusion at the feature level and decision level.

[0033] In order to cope with the computational load brought by high-dimensional data, the present invention adopts data dimensionality reduction technology, especially the method based on intrinsic orthogonal decomposition, to decompose high-dimensional data into a limited number of pattern vectors. This dimensionality reduction operation can retain the main change patterns in the data while reducing the computational complexity, thereby effectively improving the training efficiency of the deep learning model.

[0034] In terms of simulation acceleration, the present invention proposes a solution to accelerate the network using deep learning simulation. By using the reduced-dimensional data set, the deep learning network is trained and the loss function is constructed in combination with the physical mechanism to ensure that the network can maintain stability after multiple training and iterations.

[0035] Replacing the traditional numerical simulation solver with this deep learning model can not only accelerate the prediction of simulation results, but also achieve adaptive tuning through online learning mechanisms and dynamically optimize the simulation process to adapt to different test objects and complex scenarios.

[0036] Based on the acquired data set, the signed distance field is used to perform data conversion and multi-dimensional representation.

[0037] Input data into the constructed deep learning simulation acceleration network for training and optimization.

[0038] After training, the deep learning simulation acceleration network can accept reduced-dimensional data input, quickly predict simulation results, and replace traditional numerical simulation solvers.

[0039] Finally, the full-factor deep simulation acceleration model and the test full-scenario model are input to achieve the joint simulation acceleration and result prediction of the full-factor and full-scenario model of digital-real fusion testing.

[0040] In order to better understand the implementation of the present invention, the specific implementation mode of the present invention is described in detail below by taking aircraft aerodynamic simulation as an example in combination with a specific technical process.

[0041] First, analyze the test object and define the simulation requirements:

[0042] In aircraft aerodynamic simulation, it is first necessary to conduct a comprehensive analysis of the test object (aircraft) to clarify the requirements and goals of its aerodynamic characteristics test. This includes but is not limited to the following:

[0043] Fluid mechanics properties, such as lift, drag, pressure distribution, etc.;

[0044] The aerodynamic performance of the aircraft in different flight states (such as take-off, cruising, landing, etc.);

[0045] Structural mechanical properties such as airfoil deformation and fatigue strength.

[0046] By analyzing the above requirements, the corresponding performance indicators (such as lift-to-drag ratio, pressure coefficient, etc.) were determined and a test plan was formulated, which was detailed to the data collection and simulation scenarios under different flight conditions.

[0047] Then, the geometric model and mechanism model of the test object are constructed:

[0048] In this step, a geometric model is constructed based on the geometric structure of the aircraft, such as the wing, fuselage, and tail, and a mechanism model is constructed by combining the basic theories of fluid mechanics and structural mechanics. The specific process includes:

[0049] Fluid mechanics modeling: describing aerodynamic phenomena based on incompressible fluid equations (such as the Navier-Stokes equations) and turbulence models (such as the k-ε model);

[0050] Structural mechanics modeling: The aircraft structure model is constructed by the finite element method (FEM) to simulate its stress and strain distribution under external loads (such as aerodynamic force, gravity, etc.);

[0051] Parametric modeling: Through adaptive adjustment of geometric shape parameters, aerodynamic simulation under different wing structures, angles of attack and other conditions can be achieved.

[0052] These models can be co-simulated with FEM tools through simulation platforms such as CFD (computational fluid dynamics) tools, and model interoperability can be achieved through common interfaces (such as the FMI standard).

[0053] After the model is built, multi-dimensional model joint simulation and data collection are carried out:

[0054] The multi-dimensional simulation models are connected via TCP / IP or OPC UA protocols to achieve data exchange and synchronization between different simulation modules. The aerodynamic performance of the aircraft under different flight conditions is simulated and data collected in the following ways:

[0055] Computational fluid dynamics (CFD) simulation provides data such as airflow distribution and pressure distribution;

[0056] Structural simulation provides stress and deformation data;

[0057] Parametric modeling adjusts the geometry of the aircraft and automatically optimizes simulation parameters.

[0058] These data will serve as the basis for subsequent data fusion and enhancement.

[0059] By integrating physical data, historical data and simulation data, the integrity and consistency of data are ensured. The specific data fusion process includes:

[0060] Data layer fusion: preliminary processing of simulation data and physical acquisition data to eliminate noise and errors;

[0061] Feature layer fusion: extract key features from different data sources to form a unified feature representation;

[0062] Decision-level fusion: Based on the joint analysis of simulation results and physical data, the model is optimized and the simulation results are adjusted.

[0063] By fusing aerodynamic performance data at different angles of attack, a more accurate lift and drag model can be obtained. The fused data can improve the prediction accuracy of aircraft aerodynamic characteristics.

[0064] Integrate reliable physical data and historical data with simulation data.

[0065] After multi-dimensional simulation data collection, high-dimensional data sets are often generated, and the processing efficiency of these data is low.

[0066] The proper orthogonal decomposition (POD) technique is used to reduce the dimensionality of high-dimensional data. The specific process is as follows:

[0067] Data pattern extraction: By performing SVD decomposition on the flow field data, the high-dimensional pressure and velocity field data are decomposed into finite pattern vectors;

[0068] Pattern screening: retain the most representative main patterns, remove the secondary patterns with low contribution, and significantly reduce the data dimension;

[0069] Dimensionality reduction reconstruction: The flow field is reconstructed using the main mode vectors to maintain simulation accuracy at a lower computational cost.

[0070] The specific formula is as follows:

[0071] ,

[0072] in, represents the flow field, is the spatial mode, is the time coefficient, and r is the number of patterns retained.

[0073] Using the reduced-dimensional data, a deep learning simulation acceleration network is constructed. The present invention adopts the Transformer network structure and effectively captures the long-range dependencies of aerodynamic characteristics through the attention mechanism. The specific steps are as follows:

[0074] Data preprocessing: standardize the aerodynamic characteristic data and physical characteristic data after dimension reduction;

[0075] Network design: Based on the encoder part of Transformer, a deep neural network for aerodynamic characteristics prediction is constructed. The input is the reduced-dimensional pattern vector, and the output is the predicted result of the flow field;

[0076] Training and optimization: Through multiple rounds of training and iteration, combined with actual aerodynamic data, the network parameters are optimized; the loss function is combined with the physical mechanism to ensure the physical consistency of the model.

[0077] ,

[0078] in, To predict the results, is the actual simulation result. is the regularization term.

[0079] The trained Transformer network can quickly predict aerodynamic characteristics such as flow field distribution and pressure distribution when new flight conditions (such as flight speed and angle of attack) are input. Compared with traditional simulation tools, this model can significantly speed up calculations while maintaining the same accuracy as traditional simulation methods.

[0080] The full-scene test model includes various high-fidelity scene models required for aircraft operation, including terrain, weather, particles, trees, oceans, people, buildings, etc.

[0081] Through this deep learning acceleration network, in the simulation of aircraft aerodynamic characteristics, the traditional simulation calculation time can be reduced from several hours to a few minutes or even shorter real-time prediction. This greatly improves the simulation efficiency, especially for multi-condition, large-scale aerodynamic simulation tasks.

[0082] The full-scenario test model and the full-element model of the aircraft are integrated in a three-dimensional operating environment. The complete test process is carried out by setting specific simulation goals and tasks, and the test results are used to analyze whether the various performance indicators and functions of the test object meet the test requirements.

[0083] In summary, the present invention can effectively accelerate the simulation process of complex test objects such as aircraft through the combination of deep learning, data fusion and dimensionality reduction technology, and is suitable for testing needs in multiple fields and objects.

[0084] The contents not described in detail in the specification of the present invention belong to the prior art known to the professional and technical personnel in this field.

[0085] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for accelerating simulation of all-factors and multiple scenarios in digital-real fusion testing, characterized in that: The method comprises: Step 1: Design a test object element model construction module. This construction module implements the model construction of all elements for the test object. The specific implementation is as follows: Analyze the test object, propose test requirements or goals, performance indicators, and formulate a preliminary test implementation plan, including detailed decomposition of the test objectives; based on the analysis results, construct a geometric model that reflects the geometric shape characteristics of the test object and a multi-domain mechanism model of the test object's operation process; Step 2: Design a digital-real data fusion and enhancement module for the test object element model. The digital-real data fusion and enhancement module uses the constructed test object element model and physical data to achieve data fusion and enhancement. The specific implementation is as follows: Conduct joint simulation of geometric models and mechanism models to achieve multi-dimensional simulation data acquisition and collection; conduct physical data acquisition and historical data collection, and combine simulation data to achieve the fusion and enhancement of digital-physical fusion test data; use data dimensionality reduction methods to reduce the dimensionality of high-dimensional data to extract key features and obtain data sets; Step 3: Design a test object factor scenario model joint simulation acceleration module. The joint simulation acceleration module combines the full-factor simulation acceleration model and the full-scenario model required for the test to achieve joint simulation acceleration of the test. The specific implementation is as follows: For the geometric model of the test object, the signed distance field is used to perform a variant representation of the geometric model; the geometric model, signed distance field, and various operating parameters are input as training input data sets into the deep learning simulation acceleration network for training and optimization; after the training, the deep learning simulation acceleration network accepts the data input after dimensionality reduction and predicts the simulation results; combined with the online learning mechanism, it automatically adapts to the characteristics of different test objects, and dynamically optimizes the simulation process while maintaining the simulation accuracy; finally, the full-factor deep simulation acceleration model and the test full-scene model are input to realize the joint simulation acceleration and result prediction of the full-factor and full-scene model of the digital-real fusion test.

2. According to claim 1, a method for accelerating simulation of all-factors and multiple scenarios of digital-real fusion testing is characterized in that: In step 1, the multi-domain mechanism model of the test object is jointly constructed by fluid mechanics modeling, structural mechanics modeling, parametric modeling and modeling methods based on two-dimensional modeling language, and the fluid mechanics modeling is based on an incompressible fluid model, a turbulence model, and a heat conduction model; Structural mechanics modeling is constructed using the finite element method, including nonlinear behavior simulation and dynamic response prediction; parametric modeling includes adaptive adjustment of geometry and automatic calibration of simulation parameters; Modeling based on two-dimensional modeling language is used for coupled simulation of different physical fields to achieve accurate simulation of multiple physical fields.

3. According to claim 2, a method for accelerating simulation of all-factors and multiple scenarios of digital-real fusion testing is characterized in that: In step 1, the joint simulation of the geometric model and the mechanism model is achieved in the following ways: using a general simulation interface to call different types of models; interconnecting modules through a communication protocol; and using a model exchange method to achieve model-based data exchange between multiple platforms.

4. According to claim 3, a method for accelerating simulation of all-factors and multiple scenarios of digital-real fusion testing is characterized in that: In step 2, the data-physical fusion specifically includes: eliminating noise and error data fusion by jointly processing simulation data and physical acquisition data; extracting features of multi-source data to establish unified feature representation data fusion; and realizing test result data fusion by jointly analyzing the consistency features of simulation results and actual physical data.

5. A method for accelerating simulation of all-factors and multiple scenarios of digital-real fusion testing according to claim 4, characterized in that: In step 2, the data dimensionality reduction method specifically includes the following steps: extracting primary patterns with weights greater than 95% from the simulation data set, removing secondary patterns with low weights, and thus decomposing high-dimensional data into a limited number of pattern vectors.

6. A method for accelerating simulation of all-factors and multiple scenarios of digital-real fusion testing according to claim 1, characterized in that: The deep learning simulation acceleration network uses the reduced-dimensional simulation data and model data as data sets, adjusts network parameters, and constructs a network loss function to train the constructed deep learning simulation acceleration network structure, and ensures the generalization ability of the deep learning network through multiple training and iterations.

Images