Aerodynamic force prediction method and device for large transportation equipment

Abstract

The invention discloses an aerodynamic force prediction method and device for large-scale transportation equipment, and relates to the field of aerodynamic force prediction.The method comprises the steps that firstly, structural parameters and historical flight data of the large-scale transportation equipment are obtained, an aerodynamic force prediction knowledge graph is constructed based on a cloud server, and a self-adaptive CFD numerical simulation model is constructed; and carrying out adaptive optimization on the adaptive CFD numerical simulation model by adopting an adaptive local alternating genetic algorithm, and carrying out iterative updating on the aerodynamic prediction knowledge graph. Then, aerodynamic force prediction is carried out by adopting a deep reinforcement learning model, and knowledge reasoning is carried out in combination with an aerodynamic force prediction knowledge graph; the method solves the problem that the accuracy of a CFD (computational fluid dynamics) method is limited by a physical model and a computational grid in the aerodynamic force prediction of the large transportation equipment, so that the calculation failure under the conditions of complex aerodynamic force phenomena and flow characteristics is easily caused, and can comprehensively consider various physical phenomena and flow characteristics of the large transportation equipment during high-speed operation. And the prediction accuracy and reliability are improved.

Description

Technical Field

[0001] The invention relates to the field of aerodynamic force prediction, and in particular to an aerodynamic force prediction method and device for large-scale transportation equipment. Background Art

[0002] With the rapid development of industrialization and urbanization, the demand for large-scale logistics and freight is strong, and the importance of large-scale transportation equipment has become increasingly prominent. How to ensure the reliability and safety of large-scale transportation equipment at high speed is of great significance to the development of modern economy. For large-scale transportation equipment, aerodynamic performance directly affects the flight efficiency, stability and safety of the equipment. Therefore, it is very important to accurately predict the aerodynamic force on large-scale transportation equipment.

[0003] Computational fluid dynamics (CFD) is a most economical, efficient and accurate numerical calculation method. It has been widely used in the field of aerodynamic force prediction in recent years. For example, the Chinese patent document with publication number CN111859817A discloses an aircraft aerodynamic selection method based on the shock wave simulation capability of CFD software; the Chinese patent document with publication number CN114611437A discloses a method and device for establishing an aircraft aerodynamic model database based on CFD technology. However, the accuracy of the CFD method depends on the physical model and the computational grid. Large-scale transportation equipment involves a variety of physical phenomena and flow characteristics when running at high speed. The high complexity of aerodynamic force prediction will lead to CFD calculation failure.

[0004] Therefore, a method and device for predicting aerodynamic force of large-scale transportation equipment are needed to solve the above problems. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention discloses a method and device for predicting aerodynamic forces of large-scale transport equipment, which combines multiple technical means such as CFD numerical simulation, aerodynamic force prediction knowledge graph, adaptive optimization algorithm and deep reinforcement learning model to form a comprehensive aerodynamic force prediction system. This system can more comprehensively consider various physical phenomena and flow characteristics of large-scale transport equipment when running at high speed, and improve the accuracy and reliability of prediction.

[0006] The present invention adopts the following technical solutions:

[0007] A method for predicting aerodynamic force of large-scale transportation equipment comprises the following steps:

[0008] Step 1: Obtain the structural parameters of the target large-scale transport equipment, as well as historical flight data of large-scale transport equipment with different structural parameters under different flight condition parameters and meteorological condition parameters;

[0009] Step 2: Use a cloud server to build an aerodynamic prediction knowledge graph. The cloud server extracts structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of various large-scale transportation equipment from the data obtained in step 1 as nodes of the aerodynamic prediction knowledge graph, and integrates aerodynamic prediction subject knowledge and aerodynamic formulas to form a network structure of the aerodynamic prediction knowledge graph;

[0010] Step 3: construct an adaptive CFD numerical simulation model based on the aerodynamic prediction knowledge graph and the structural parameters of the target large-scale transportation equipment, and use an adaptive optimization algorithm to optimize the adaptive CFD numerical simulation model;

[0011] Step 4: Use the optimized adaptive CFD numerical simulation model to simulate the aerodynamic performance of the target large-scale transport equipment when it is running at high speed under different flight conditions and meteorological conditions, and iteratively update the aerodynamic prediction knowledge graph based on the simulation process data;

[0012] Step 5: Use the deep reinforcement learning model to predict the aerodynamic performance of large transportation equipment, and use the updated aerodynamic prediction knowledge graph to train the deep reinforcement learning model.

[0013] Furthermore, in step one, the structural parameters include at least fuselage length, fuselage width, fuselage height, fuselage shape, wing shape, wing area, vertical tail area, center of gravity position, engine thrust, Mach number and distribution of angle of attack along the curved surface; the meteorological condition parameters include at least wind speed, wind direction, temperature, humidity and air pressure; the flight condition parameters include at least flight altitude, flight speed, flight attitude and maneuvering movements; and the flight data include at least lift, drag, lateral force and aerodynamic moment.

[0014] Furthermore, in the step 2, the cloud server includes a data preprocessing unit, a domain classification unit, an entity recognition and linking unit, a relationship extraction unit, a knowledge organization representation unit and an incremental update unit. The data preprocessing unit uses a natural language processor to extract structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of large-scale transportation equipment, and cleans and organizes the extracted parameters. The domain classification unit classifies the extracted parameters according to the subject domain distribution through the text domain mining model LDA. The entity recognition and linking unit automatically recognizes the aerodynamic entities in the extracted parameters through the named entity recognition task NER, and links the aerodynamic entities through the entity linking task EL. The dynamic entity is linked to the corresponding entity in the aerodynamic prediction knowledge graph. The relationship extraction unit automatically extracts the relationship between entities from the aerodynamic prediction subject knowledge and aerodynamic formulas through a relationship extraction algorithm, and stores them in the aerodynamic prediction knowledge graph to form a network structure. The relationship between entities at least includes the relationship between structural parameters and aerodynamic performance, the relationship between flight condition parameters and aerodynamic changes, and the relationship between meteorological condition parameters and aerodynamic changes. The knowledge organization representation unit uses a graph database or a triple storage format to store the extracted parameters, aerodynamic prediction subject knowledge and aerodynamic formulas. The incremental update unit updates the aerodynamic prediction knowledge graph in real time through a time window mechanism.

[0015] Furthermore, in step three, constructing an adaptive CFD numerical simulation model includes the following steps:

[0016] S301, receiving structural parameters of the target large-scale transport equipment, and extracting structural parameters, flight condition parameters and meteorological condition parameters matching the target large-scale transport equipment from the aerodynamic force prediction knowledge graph, and initializing the CFD numerical simulation model based on the extracted parameters, wherein the initialization setting at least includes geometric model setting, mesh division and boundary condition setting;

[0017] S302, setting an adaptive optimization algorithm according to the structural parameters of the target large-scale transport equipment and the simulation target, and using the adaptive optimization algorithm to optimize the initialization settings of the CFD numerical simulation model, wherein the adaptive optimization algorithm sets the aerodynamic performance simulation performance as the optimization target, and sets constraint conditions based on aerodynamic prediction subject knowledge and aerodynamic formula;

[0018] S303, selecting a physical model according to the characteristics of the target large-scale transport equipment and the expected flight environment to reflect the aerodynamic performance of the target large-scale transport equipment when running at high speed;

[0019] S304, integrating the adaptive optimization algorithm with the CFD numerical simulation model to form an adaptive CFD numerical simulation model, and numerically simulating the aerodynamic performance of the target large-scale transport equipment under different flight conditions and meteorological conditions according to the selected physical model and the adaptive CFD numerical simulation model;

[0020] S305. Compare the historical flight data with the numerical simulation, and calibrate the adaptive CFD numerical simulation model according to the verification and comparison results.

[0021] Furthermore, the adaptive optimization algorithm optimizes the initialization settings of the CFD numerical simulation model based on the multi-agent proximal strategy model and the local alternating genetic model, and the working method of the adaptive optimization algorithm includes the following steps:

[0022] S3021. Consider the initialization setting variables in the CFD numerical simulation model as an intelligent agent, the operating environment of the intelligent agent is the CFD numerical simulation model, and the strategy function and value function of the intelligent agent are defined by using the aerodynamic prediction subject knowledge and the aerodynamic formula. The strategy function adjusts the value of the current initialization setting variables according to the current state of the CFD numerical simulation model and the historical experience of the initialization setting variables, and the value function evaluates the value of the current initialization setting variables according to the current state of the CFD numerical simulation model and the strategy function;

[0023] S3022, the agent performs adjustment actions in the operating environment according to the policy function, and updates the policy function parameters and value function parameters of the agent based on a proximal policy optimization method, wherein the proximal policy optimization method constrains the update step size of the policy and value functions through a policy loss function and a value loss function;

[0024] S3023. In the process of updating and iterating the parameters of the strategy function and the value function, the agents are divided into multiple groups, and a local alternating genetic model is used to perform local collaboration and overall competition among the multiple agents.

[0025] S3024, inputting the updated and iterated initialization setting variable values ​​into the CFD numerical simulation model for numerical calculation, and feeding back the calculation results to the intelligent agent as the current state of the CFD numerical simulation model;

[0026] S3025. Repeat step S3022 and step S3023 until a predetermined aerodynamic performance simulation is achieved.

[0027] Furthermore, the expression of the strategy loss function is:

[0028]

[0029] In formula (1), L P(θ) represents the updated loss function of the policy function parameter θ, which is used to measure the difference between the new policy and the old policy, E t represents the expected value of all states and actions at time step t, r t (θ) represents the policy function parameters in state s at time step t t Take action a t The probability ratio of the strategy, Represents the advantage estimate at time step t, which is used to measure the advantage of taking actions relative to the average action selection in the state, ∈ represents a hyperparameter used to limit the probability ratio r t The range of variation of (θ), clip(r t (θ),1-∈,1+∈) represents the probability ratio r t (θ) is limited to between 1-∈ and 1+∈, and the expression of the value loss function is:

[0030]

[0031] In formula (2), Represents the value function parameter The updated loss function is Indicates that in state s t The value function of the agent is: represents the reward value obtained at time step t, E t represents the expected value of all states and actions at time step t.

[0032] Furthermore, the deep reinforcement learning model extracts aerodynamic prediction features from the adaptive CFD numerical simulation results as aerodynamic prediction input features, and uses a graph neural network to perform knowledge reasoning on the aerodynamic prediction knowledge graph to mine the aerodynamic characteristics inside large transportation equipment.

[0033] Furthermore, an aerodynamic force prediction device for large-scale transportation equipment comprises:

[0034] The data acquisition module acquires the structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of the target large-scale transport equipment, and transmits them to the cloud server for processing and storage;

[0035] The knowledge graph construction module uses a cloud server to construct an aerodynamic prediction knowledge graph. The cloud server extracts information from structural parameters, flight conditions, meteorological conditions, and historical flight data as nodes of the aerodynamic prediction knowledge graph, and integrates aerodynamic knowledge, aerodynamic formulas, and physical models to form a multi-dimensional prediction framework.

[0036] Numerical simulation module, which builds an adaptive CFD numerical simulation model based on the aerodynamic prediction knowledge graph and the structural parameters of the target large-scale transportation equipment to simulate aerodynamic performance, and iteratively updates the aerodynamic prediction knowledge graph based on the simulation process data;

[0037] The aerodynamic prediction module uses a deep reinforcement learning model to predict the aerodynamic performance of large transport equipment and uses the updated aerodynamic prediction knowledge graph for training;

[0038] The user interaction module uses a graphical interface to view the real-time status of the adaptive CFD numerical simulation model and the aerodynamic performance prediction results, and adjust the flight conditions and meteorological conditions. Then, the aerodynamic performance report of large transport equipment is automatically generated based on the simulation and prediction results.

[0039] The beneficial effects of the present invention are:

[0040] 1. The present invention can more accurately reflect the aerodynamic performance of the equipment when running at high speed by constructing an adaptive CFD numerical simulation model and selecting a physical model according to the characteristics of the target large-scale transportation equipment and the expected flight environment. In addition, the adaptive optimization algorithm is used to optimize the CFD numerical simulation model, which can further improve the accuracy and efficiency of the calculation.

[0041] 2. The present invention forms an aerodynamic prediction knowledge graph by integrating aerodynamic prediction subject knowledge and aerodynamic formulas. This method can extract structural parameters, flight condition parameters, and meteorological condition parameters that match the target large-scale transportation equipment, and provide more accurate and comprehensive initialization settings for CFD numerical simulation. At the same time, the iterative update of the knowledge graph can also continuously optimize the accuracy of CFD numerical simulation.

[0042] 3. The present invention uses a deep reinforcement learning model to predict the aerodynamic performance of large-scale transport equipment, and uses the updated aerodynamic prediction knowledge graph to train the model. This method combining deep learning and reinforcement learning can mine the aerodynamic characteristics inside large-scale transport equipment and improve the accuracy and generalization ability of prediction.

[0043] 4. The present invention adopts a multi-agent proximal strategy model and a local alternating genetic model to carry out local collaboration and overall competition among agents, which can more effectively explore and optimize the initialization settings of the CFD numerical simulation model, thereby improving the accuracy and efficiency of the calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0045] Figure 2 A schematic diagram of the workflow for constructing an adaptive CFD numerical simulation model in the present invention;

[0046] Figure 3 Schematic diagram of the workflow of the adaptive optimization algorithm in the present invention;

[0047] Figure 4 It is a schematic diagram of the overall equipment architecture of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the attached embodiment of the present invention Figure 1 To Attachment Figure 4 , the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The embodiment of the present invention discloses a method for predicting aerodynamic force of large-scale transportation equipment, as shown in the attached Figure 1 As shown, the following steps are included:

[0050] Step 1: Get data;

[0051] The structural parameters of the target large-scale transport equipment are obtained by consulting technical documents or design drawings. The structural parameters include at least the length of the fuselage, the width of the fuselage, the height of the fuselage, the shape of the fuselage, the shape of the wing, the wing area, the area of ​​the vertical tail, the position of the center of gravity, the thrust of the engine, the Mach number and the distribution of the angle of attack along the curved surface. The historical flight data of large-scale transport equipment with various different structural parameters under different flight condition parameters and meteorological condition parameters are recorded by flight recorders, data recorders, flight simulators and other equipment. The collected structural parameters and historical flight data are stored on a cloud server, and a database management system is used to manage and maintain the data. The data interface is connected to the database, cloud storage and other data sources to automatically or manually import relevant data. The meteorological condition parameters include at least wind speed, wind direction, temperature, humidity and air pressure. The flight condition parameters include at least flight altitude, flight speed, flight attitude and maneuvering movements. The flight data includes at least lift, drag, lateral force and aerodynamic moment.

[0052] During the data collection process, ensure the accuracy, completeness and consistency of the data. Perform preliminary screening and cleaning of the data to remove outliers, duplicate values ​​or invalid data. Standardize the data to ensure that data from different sources are consistent in format, unit and precision. Store the collected data in a secure and reliable database for subsequent analysis and processing. Establish data indexes and metadata for quick retrieval and access to data. Back up data regularly to prevent data loss or damage.

[0053] Step 2: Construct aerodynamic prediction knowledge graph;

[0054] The cloud server is used to build the aerodynamic prediction knowledge graph. The cloud server integrates the various data types obtained in step 1 to build nodes of the aerodynamic prediction knowledge graph. Each node represents a structural parameter, a flight condition parameter or a meteorological condition parameter, and the relationship between each parameter (for example, the relationship between flight speed and lift and drag) will be connected as an edge. On the cloud server, the network structure of the knowledge graph is constructed based on the existing aerodynamic theory, aerodynamic formulas (such as calculation formulas for lift, drag, aerodynamic load, etc.), and empirical formulas extracted from historical data. The edges in the graph can be physical models, empirical formulas, or even mathematical relationships obtained through historical data regression. Nodes are structural parameters, flight conditions, meteorological conditions, etc. Edges are aerodynamic relationships, mechanical formulas, historical data regression results, etc. The network structure of the knowledge graph is used to realize data transmission and learning between different nodes, and establish an aerodynamic prediction knowledge graph. All data and calculations will be performed on the cloud server, which can meet large-scale data processing and computing needs and facilitate later iterative updates.

[0055] In the step 2, the cloud server includes a data preprocessing unit, a domain classification unit, an entity recognition and linking unit, a relationship extraction unit, a knowledge organization representation unit, and an incremental update unit;

[0056] Data preprocessing unit: Use natural language processor (NLP) technology to extract structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of large transport equipment from original data sources (such as databases, documents, log files, etc.). NLP tools include word segmenters, part-of-speech taggers, named entity recognizers, etc. to accurately identify and extract relevant parameters. The extracted data is processed to remove duplicates, fill in missing values, correct erroneous values, etc. to ensure the accuracy and completeness of the data. The data is formatted into a unified format for subsequent processing and storage.

[0057] Domain classification unit: LDA (Latent Dirichlet Allocation) text domain mining model is used to perform topic modeling on the extracted parameter text. By analyzing the topic distribution of the text, the parameters are classified according to subject areas (such as aerodynamics, mechanical engineering, meteorology, etc.).

[0058] Entity Recognition and Linking Unit: Use Named Entity Recognition (NER) technology to automatically identify aerodynamic related entities in the extracted parameters, such as equipment names, component names, parameter names, etc. NER tools are based on deep learning models such as LSTM-CRF, BERT, etc. to improve the accuracy of recognition. Through the Entity Linking (EL) technology, the identified aerodynamic entities are linked to the corresponding entities that already exist in the aerodynamic prediction knowledge graph. If the corresponding entity does not exist in the knowledge graph, a new entity is created and added to the graph.

[0059] Relationship extraction unit: Apply relationship extraction algorithms (such as rule-based methods, machine learning algorithms, deep learning models, etc.) to extract the relationship between entities from the aerodynamic prediction subject knowledge and aerodynamic formulas. The relationship extraction algorithm involves natural language understanding, semantic analysis and other technologies. The extracted relationships are stored in the aerodynamic prediction knowledge graph to form a network structure. The relationships include at least the relationship between structural parameters and aerodynamic performance, the relationship between flight condition parameters and aerodynamic changes, and the relationship between meteorological condition parameters and aerodynamic changes.

[0060] Knowledge organization and representation unit: Choose a graph database (such as Neo4j, JanusGraph, etc.) or a triple storage format (such as RDF, OWL, etc.) to store the extracted parameters, aerodynamic prediction subject knowledge, and aerodynamic formulas. Graph databases are suitable for storing complex network structures, while triple storage formats facilitate semantic query and reasoning. Store the extracted parameters and their attribute values. Store the structured representation of aerodynamic prediction subject knowledge and aerodynamic formulas. Store the relationships between entities and their attributes.

[0061] Incremental update unit: Set up a time window mechanism to regularly obtain new data from the data source. The time window can be set to a fixed time interval (such as daily, weekly, monthly, etc.), or dynamically adjusted according to the frequency of data updates. Perform preprocessing, entity recognition and linking, and relationship extraction on new data. Integrate new knowledge into the aerodynamic prediction knowledge graph and update the nodes and relationships of the graph. Verify the updated knowledge graph to ensure the accuracy and consistency of the data. If errors or inconsistent data are found, make corrections and adjustments in a timely manner.

[0062] Through the above units, the cloud server can build an accurate and complete aerodynamic prediction knowledge graph, providing strong support for subsequent aerodynamic prediction. At the same time, through the continuous updating and improvement of the incremental update unit, the knowledge graph can maintain consistency with the actual situation and improve the accuracy and reliability of the prediction.

[0063] Step 3: Build an adaptive CFD numerical simulation model based on knowledge graph

[0064] Input the structural parameters and historical flight data into the aerodynamic prediction knowledge graph to obtain the learning parameters in the network structure. Use the learning parameters to perform adaptive meshing and fixed computational domain management. Based on the aerodynamic prediction knowledge graph and the structural parameters of the target large-scale transportation equipment, construct its aerodynamic simulation model through computational fluid dynamics (CFD) method. CFD simulation will consider the changes in the air flow field during flight and calculate the aerodynamic response of the equipment under different conditions. In order to improve the accuracy and computational efficiency of the CFD numerical model, an adaptive optimization algorithm is used to optimize the CFD model. The algorithm can automatically adjust parameters such as grid density and flow field simulation accuracy, so that CFD simulation can reduce the calculation time while ensuring the calculation accuracy.

[0065] In step 3, an adaptive CFD numerical simulation model is constructed, as shown in the attached Figure 2 As shown, the following steps are included:

[0066] S301. Initialize the CFD numerical simulation model

[0067] The structural parameters of the target large-scale transport equipment are received. These parameters usually include the size, shape, mass distribution, etc. of the equipment, which are the basis for establishing the geometric model. Parameters are extracted from the aerodynamic prediction knowledge graph, which stores various parameters related to the target large-scale transport equipment, including structural parameters, flight condition parameters (such as speed, altitude, angle of attack, etc.) and meteorological condition parameters (such as wind speed, wind direction, temperature, etc.). These parameters are used to perform more accurate initialization settings for the CFD numerical simulation model.

[0068] The CFD numerical simulation model is initialized based on the extracted parameters. The initialization settings include at least geometric model settings, mesh division and boundary condition settings. The geometric model setting is to build a three-dimensional model of the equipment according to the structural parameters; the mesh division is to divide the geometric model into multiple small units for numerical calculation; the boundary condition setting is to set the boundary conditions of the simulation model according to the flight conditions and meteorological conditions, such as inlet velocity, outlet pressure, etc.

[0069] S302. Optimizing the initialization settings of the CFD numerical simulation model

[0070] According to the structural parameters of the target large-scale transport equipment and the simulation objectives, the relevant parameters of the adaptive optimization algorithm are set, such as the maximum number of grid lines and the number of multi-grid iterations, etc., to optimize the initialization settings of the CFD numerical simulation model, and the aerodynamic performance simulation performance is used as the optimization target, such as lift coefficient, drag coefficient, lift-to-drag ratio, etc. In the simulation process, aerodynamic performance is the key indicator that the model needs to achieve. According to the aerodynamic force prediction knowledge and physical formulas, the constraints in the optimization process are set. Common constraints include: flight safety constraints (such as avoiding excessive aerodynamic loads), structural strength constraints (such as the carrying capacity of the wings), and economic constraints such as range and fuel consumption. The adaptive optimization algorithm will adjust the settings of the CFD model (such as grid density, physical model parameters, etc.) through repeated iterations until the optimization target is improved.

[0071] S303, select physical model

[0072] Select a suitable physical model based on the characteristics of the target large-scale transport equipment and the expected flight environment. Common physical models include: Turbulence model: Select a suitable turbulence model (such as k-ε model, k-ω model, large eddy simulation (LES), etc.) according to the characteristics of the flow field. Heat transfer model: If you need to simulate temperature effects, select a suitable heat transfer model, such as the convection-conduction model. Shock wave and high-speed flow model: If the equipment involves high-speed flight, it is necessary to use an aerodynamic model suitable for supersonic or subsonic speeds, considering shock waves, shock waves and other phenomena. Structural-aerodynamic coupling model: For equipment involving aeroelastic effects, it is necessary to select a model that couples structure and fluid.

[0073] S304. Numerical simulation of the aerodynamic performance of target large-scale transport equipment

[0074] The adaptive optimization algorithm is integrated with the CFD numerical simulation model to form a closed-loop system. The optimization algorithm adjusts the simulation parameters through iterations and evaluates the optimization target after each simulation result. The selected physical model and the optimized CFD model are used to simulate the aerodynamic performance under different flight conditions and meteorological conditions. This process includes: changes in flight speed, changes in altitude, changes in meteorological conditions (such as different wind speeds, temperatures, etc.). The optimized aerodynamic performance data, such as lift, drag, stability coefficient, etc., are output for subsequent design and verification.

[0075] S305. Calibrate the adaptive CFD numerical simulation model

[0076] Use historical flight data (such as aerodynamic performance data measured during actual flight) to compare with numerical simulation results. The purpose of this step is to verify the accuracy of the CFD numerical simulation model. According to the results of the verification and comparison, the accuracy of the CFD model is evaluated. If there is a large deviation between the numerical simulation results and the actual flight data, the CFD numerical simulation model needs to be calibrated. This can include modifying the physical model parameters, the weights of the optimization algorithm, or readjusting the boundary conditions. The calibrated model can be re-input into the optimization process to form a more accurate simulation model to improve the simulation accuracy and optimization results.

[0077] The adaptive optimization algorithm optimizes the initialization settings of the CFD numerical simulation model based on the multi-agent proximal strategy model and the local alternating genetic model. The working method of the adaptive optimization algorithm is shown in the attached figure. Figure 3 As shown, the following steps are included:

[0078] S3021, Agent and Strategy Function Initialization

[0079] The initialization setting variables in the CFD numerical simulation model are regarded as multiple agents, each of which corresponds to a variable in the CFD simulation (such as initial velocity, initial pressure, boundary conditions, etc.). The task of the agent is to optimize these initialization variables by interacting with the environment to improve the simulation results. Each agent has a policy function and a value function. The policy function is used to determine the action that the agent should take in a specific state (i.e., adjust the value of the initialization variable), while the value function is used to evaluate the "goodness" of the current state. According to the current state of the CFD numerical simulation model (such as the temperature, pressure, and velocity distribution of the current flow field) and the historical experience of the initialization setting variables, adjust the value of the current initialization variable. The role of the value function is to evaluate whether the strategy adopted by the current agent is helpful to improve the quality of the simulation results, which is usually measured by the reward value.

[0080] S3022. Use proximal strategy optimization methods to optimize the strategy function and value function of the intelligent agent

[0081] The agent performs adjustment actions in the operating environment according to the policy function and observes the resulting results (i.e., new states or rewards). Proximal policy optimization methods constrain the update step size of the policy and value functions by maximizing the policy loss function and minimizing the value loss function, thereby ensuring stability during the learning process.

[0082] The policy loss function is used to limit the step size of the policy update to avoid excessive changes that lead to unstable learning. The expression of the policy loss function is:

[0083]

[0084] In formula (1), LP (θ) represents the updated loss function of the policy function parameter θ, which is used to measure the difference between the new policy and the old policy, E t represents the expected value of all states and actions at time step t, r t (θ) represents the policy function parameters in state s at time step t t Take action a t The probability ratio of the strategy is the ratio of the new strategy to the old strategy (i.e. the probability ratio of the new strategy to the old strategy), which is used to calculate the difference between the new strategy and the old strategy. Represents the advantage estimate at time step t, which is used to measure the advantage of taking actions relative to the average action selection in the state, ∈ represents a hyperparameter used to limit the probability ratio r t (θ) changes, ensuring that the change between the new strategy and the old strategy does not exceed a certain threshold, clip(r t (θ),1-v,1+∈) represents the probability ratio r t (θ) is restricted to the range 1-∈ to 1+∈.

[0085] The value loss function is used to optimize the value function so that it can better predict the expected return of each state. The expression of the value loss function is:

[0086]

[0087] In formula (2), Represents the value function parameter The updated loss function is Indicates that in state s t The value function of the agent is: represents the reward value obtained at time step t, E t It represents the expected value of all states and actions at time step t, and is used to estimate the average return of all actions in a certain state.

[0088] S3023, using the local alternating genetic model to update the strategy function parameters and value function parameters

[0089] According to the scale and complexity of the problem, as well as the number and capabilities of the agents, the agents are divided into multiple groups. Ensure that the number of agents in each group is appropriate to allow for effective collaboration and competition. Set initial parameter values ​​for each agent, including policy function parameters and value function parameters. Initialize the parameters of the genetic algorithm, such as population size, crossover probability, mutation probability, etc.

[0090] In each group, agents solve problems together through information sharing and collaboration. Communication protocols are used to ensure the reliability and real-time nature of information transmission, as well as task allocation and coordination mechanisms to optimize cooperation between agents. Excellent agents are selected as parents based on the fitness of the agents (i.e., their problem-solving effectiveness). In the group, crossover operations are performed on the selected parent agents to generate new child agents. Crossover operations can include single-point crossover, multi-point crossover, etc. The genes of the child agents are mutated with a certain probability to increase the diversity of solutions.

[0091] Competition is used between different groups to drive the entire optimization process in a better direction. The performance of each group is evaluated through competition or tournaments, and the groups are ranked according to their performance. The policy function parameters and value function parameters of the agents are updated according to their performance in the competition. For agents with excellent performance, their parameter values ​​can be retained as the starting point for the next round of iterations. For agents with poor performance, they can be adjusted or replaced according to the rules of the genetic algorithm.

[0092] The process of local collaboration, overall competition, and parameter updating is repeated until the predetermined number of iterations is reached or a certain convergence condition is met. In each iteration, the performance of the agent is evaluated, and the parameters and strategies of the genetic algorithm are adjusted based on the evaluation results. Through continuous iteration and optimization, a set of excellent agent parameter values ​​can be obtained, which can be used to solve similar problems or as a starting point for further research.

[0093] S3024, Calculation and feedback of CFD numerical simulation model

[0094] The updated and iterated initialization variable values ​​are input into the CFD numerical simulation model for numerical calculation. The calculation results (such as aerodynamic performance parameters) are fed back to the intelligent agent as the current state of the CFD numerical simulation model, which is used by the intelligent agent to update its strategy and value function according to the new state and continue the optimization process. The feedback simulation results usually include key indicators such as aerodynamic performance, fluid flow characteristics, temperature field, etc. The intelligent agent adjusts its strategy based on these feedbacks to achieve more accurate simulation results.

[0095] S3025, repeat the iteration until the predetermined goal is achieved

[0096] Repeat steps S3022 and S3023 until the simulation results meet the predetermined aerodynamic performance requirements. This process is a dynamic adjustment process, which gradually optimizes the initialization settings through multiple iterations, so that the calculation accuracy and efficiency of the CFD model are continuously improved. When the model reaches the predetermined aerodynamic performance simulation target (for example, the simulation error is lower than a certain threshold or the performance index meets the requirements), the algorithm terminates and outputs the optimized initialization settings.

[0097] Step 4: Aerodynamic performance simulation and knowledge graph iterative update

[0098] The historical flight data of the target large transport equipment under different flight conditions and meteorological conditions are input into the optimized adaptive CFD numerical simulation model to simulate the aerodynamic performance. During the simulation process, the adaptive CFD numerical simulation model automatically adjusts the grid division and computational domain management according to the input parameters to ensure the accuracy and stability of the simulation results. These simulations will generate a large amount of simulation data, including aerodynamic parameters such as lift, drag, and pitch moment. According to the comparison between the simulation results and historical flight data, the aerodynamic prediction knowledge graph is updated in real time. After each simulation, the simulation data is analyzed to extract key indicators such as aerodynamic coefficient, lift-to-drag ratio, side force coefficient, pitch moment, etc. The extracted key indicators are added as new nodes to the aerodynamic prediction knowledge graph, and new aerodynamic laws are extracted from the simulation data through machine learning algorithms (such as regression analysis, Kriging interpolation, etc.), and the network structure of the graph is updated according to the relationship between the nodes and the principle of action. According to the relationship between the newly added nodes and the original nodes, the knowledge graph is iteratively learned and optimized to improve the prediction precision and accuracy.

[0099] By continuously iterating and updating the aerodynamic prediction knowledge graph, the prediction model can have better adaptability and accuracy, providing more scientific and effective aerodynamic prediction methods and technical support for the design, production and operation of large-scale transportation equipment.

[0100] Step 5: Deep reinforcement learning model training and aerodynamic performance prediction

[0101] A deep reinforcement learning (DRL) model is constructed and trained to predict the aerodynamic performance of large transport equipment under various flight conditions. Deep reinforcement learning optimizes decision-making strategies by training from historical data and simulation results to make predictions more accurate. State space: flight conditions, meteorological conditions, structural parameters, etc. Action space: adjust flight parameters (such as speed, angle of attack, etc.). Reward function: evaluate the predicted aerodynamic performance through actual aerodynamic data and simulation results, and give rewards or penalties. The updated aerodynamic prediction knowledge graph is used as an important input of the reinforcement learning model, so that the model can refer to known aerodynamic laws during the training process and quickly converge to accurate aerodynamic prediction results. The trained deep reinforcement learning model can predict the aerodynamic performance of the target large transport equipment under different flight conditions in real time, and provide it to practical applications such as flight control systems, design optimization, or pilot decision support.

[0102] A deep reinforcement learning (DRL) method is used to further predict aerodynamic performance. Deep reinforcement learning continuously optimizes the decision-making strategy in the prediction process by learning the dynamic laws in the simulation data. Aerodynamic features are extracted from the CFD simulation results as the input features of the model, including aerodynamic lift, drag, side force, etc. The reward function is designed according to the target aerodynamic performance (such as the ratio of lift to drag, fuel efficiency, flight stability, etc.), and the reinforcement learning algorithm optimizes the control strategy through continuous iteration. In the deep reinforcement learning model, a graph neural network (GNN) is used to perform knowledge reasoning on the aerodynamic prediction knowledge graph. The graph neural network can handle complex dependencies between nodes and can mine and model the complex relationship between the internal and external aerodynamic characteristics of large transportation equipment. It can be used as a graph embedding module in the reinforcement learning model to reason about the structural and flight condition parameters and improve the prediction ability of the model. Through GNN, deeper laws can be extracted from historical flight data and the aerodynamic prediction knowledge graph, and the aerodynamic behavior of the equipment under different conditions can be mined, thereby better improving the accuracy and reliability of the prediction model.

[0103] A device for predicting the aerodynamic force of large transport equipment, such as the attached Figure 4 As shown, including:

[0104] Data acquisition module, which is responsible for collecting structural parameters (such as size, shape, weight distribution, etc.), flight condition parameters (such as speed, altitude, Mach number, angle of attack, sideslip angle, etc.), meteorological condition parameters (such as wind speed, wind direction, temperature, humidity, etc.) and historical flight data of the target large-scale transport equipment. Through sensors, data recording devices or data interfaces, data is acquired from the target large-scale transport equipment and its related environment in real time or regularly, and transmitted to the cloud server for processing and storage.

[0105] The knowledge graph construction module uses cloud servers to process data and construct an aerodynamic prediction knowledge graph. Key information is extracted from structural parameters, flight conditions, meteorological conditions, and historical flight data as nodes of the aerodynamic prediction knowledge graph. Aerodynamics knowledge, aerodynamic formulas, and physical models are integrated to form a multi-dimensional prediction framework. This may include fluid mechanics principles, aerodynamic equations, wind tunnel test data, etc. By associating nodes and integrating knowledge, a complete aerodynamic prediction knowledge graph is constructed.

[0106] The numerical simulation module builds an adaptive CFD numerical simulation model based on the aerodynamic prediction knowledge graph and the structural parameters of the target large-scale transport equipment to simulate aerodynamic performance. An adaptive CFD numerical simulation model is built based on the structural parameters of the target large-scale transport equipment and the aerodynamic prediction knowledge graph. The aerodynamic performance of the target large-scale transport equipment under different flight conditions and meteorological conditions is simulated using the CFD numerical simulation method. Based on the simulation process data, the aerodynamic prediction knowledge graph is iteratively updated to improve its accuracy and reliability.

[0107] The aerodynamic prediction module uses a deep reinforcement learning model to predict the aerodynamic performance of large transport equipment and uses the updated aerodynamic prediction knowledge graph for training. The updated aerodynamic prediction knowledge graph is used as training data to train the deep reinforcement learning model. The trained deep reinforcement learning model is used to predict the aerodynamic performance of the target large transport equipment under different conditions.

[0108] The user interaction module uses a graphical interface to view the real-time status and aerodynamic performance prediction results of the adaptive CFD numerical simulation model, and adjust the flight conditions and meteorological conditions. Then, the aerodynamic performance report of large transport equipment is automatically generated based on the simulation and prediction results. A user-friendly graphical interface is designed to display the real-time status and aerodynamic performance prediction results of the adaptive CFD numerical simulation model. Allow users to adjust flight conditions and meteorological conditions through the graphical interface to observe their impact on aerodynamic performance. Based on the simulation and prediction results, an aerodynamic performance report of large transport equipment is automatically generated, including key indicators such as aerodynamic drag, lift, and stability analysis.

[0109] In practical applications, these modules work together to form a complete aerodynamic prediction system. Users can input the structural parameters and flight conditions of the target large-scale transport equipment through the user interaction module, and the system will automatically perform data processing, knowledge graph construction, numerical simulation, aerodynamic prediction and report generation. This not only helps to reduce the cost and time of aerodynamic prediction, but also improves the accuracy and reliability of the prediction, providing strong support for the design and optimization of large-scale transport equipment.

[0110] The output end of the data acquisition module is connected to the input end of the knowledge graph construction module, the knowledge graph construction module is bidirectionally connected to the numerical simulation module, the output end of the numerical simulation module is connected to the input end of the aerodynamic prediction module, the output end of the knowledge graph construction module is connected to the input end of the aerodynamic prediction module, the output ends of the data acquisition module and the aerodynamic prediction module are connected to the input end of the user interaction module, and the numerical simulation module is bidirectionally connected to the user interaction module.

[0111] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A method for predicting aerodynamic forces of large transport equipment, characterized in that: The following steps are involved: Step 1: Obtain the structural parameters of the target large-scale transport equipment, as well as historical flight data of large-scale transport equipment with different structural parameters under different flight condition parameters and meteorological condition parameters; Step 2: Use a cloud server to build an aerodynamic prediction knowledge graph. The cloud server extracts structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of various large-scale transportation equipment from the data obtained in step 1 as nodes of the aerodynamic prediction knowledge graph, and integrates aerodynamic prediction subject knowledge and aerodynamic formulas to form a network structure of the aerodynamic prediction knowledge graph; Step 3: construct an adaptive CFD numerical simulation model based on the aerodynamic prediction knowledge graph and the structural parameters of the target large-scale transportation equipment, and use an adaptive optimization algorithm to optimize the adaptive CFD numerical simulation model; Step 4: Use the optimized adaptive CFD numerical simulation model to simulate the aerodynamic performance of the target large-scale transport equipment when it is running at high speed under different flight conditions and meteorological conditions, and iteratively update the aerodynamic prediction knowledge graph based on the simulation process data; Step 5: Use the deep reinforcement learning model to predict the aerodynamic performance of large transportation equipment, and use the updated aerodynamic prediction knowledge graph to train the deep reinforcement learning model.

2. The aerodynamic prediction method of large-scale transportation equipment according to claim 1, characterized in that: In step one, the structural parameters include at least the fuselage length, fuselage width, fuselage height, fuselage shape, wing shape, wing area, vertical tail area, center of gravity position, engine thrust, Mach number and distribution of angle of attack along the curved surface; the meteorological condition parameters include at least wind speed, wind direction, temperature, humidity and air pressure; the flight condition parameters include at least flight altitude, flight speed, flight attitude and maneuvering movements; and the flight data include at least lift, drag, lateral force and aerodynamic moment.

3. The aerodynamic prediction method of large-scale transportation equipment according to claim 1, characterized in that: In the step 2, the cloud server includes a data preprocessing unit, a domain classification unit, an entity recognition and linking unit, a relationship extraction unit, a knowledge organization representation unit and an incremental update unit. The data preprocessing unit uses a natural language processor to extract structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of large-scale transportation equipment, and cleans and organizes the extracted parameters. The domain classification unit classifies the extracted parameters according to the subject domain distribution through the text domain mining model LDA. The entity recognition and linking unit automatically recognizes the aerodynamic entities in the extracted parameters through the named entity recognition task NER, and connects the aerodynamic entities to the aerodynamic entities through the entity linking task EL. The entity is linked to the corresponding entity in the aerodynamic prediction knowledge graph, the relationship extraction unit automatically extracts the relationship between entities from the aerodynamic prediction subject knowledge and aerodynamic formulas through a relationship extraction algorithm, and stores them in the aerodynamic prediction knowledge graph to form a network structure, the relationship between entities at least includes the relationship between structural parameters and aerodynamic performance, the relationship between flight condition parameters and aerodynamic changes, and the relationship between meteorological condition parameters and aerodynamic changes, the knowledge organization representation unit uses a graph database or a triple storage format to store the extracted parameters, aerodynamic prediction subject knowledge and aerodynamic formulas, and the incremental update unit updates the aerodynamic prediction knowledge graph in real time through a time window mechanism.

4. The aerodynamic prediction method of large-scale transportation equipment according to claim 1, characterized in that: In the step 3, constructing an adaptive CFD numerical simulation model includes the following steps: S301, receiving structural parameters of the target large-scale transport equipment, and extracting structural parameters, flight condition parameters and meteorological condition parameters matching the target large-scale transport equipment from the aerodynamic force prediction knowledge graph, and initializing the CFD numerical simulation model based on the extracted parameters, wherein the initialization setting at least includes geometric model setting, mesh division and boundary condition setting; S302, setting an adaptive optimization algorithm according to the structural parameters of the target large-scale transport equipment and the simulation target, and using the adaptive optimization algorithm to optimize the initialization settings of the CFD numerical simulation model, wherein the adaptive optimization algorithm sets the aerodynamic performance simulation performance as the optimization target, and sets constraint conditions based on aerodynamic prediction subject knowledge and aerodynamic formula; S303, selecting a physical model according to the characteristics of the target large-scale transport equipment and the expected flight environment to reflect the aerodynamic performance of the target large-scale transport equipment when running at high speed; S304, integrating the adaptive optimization algorithm with the CFD numerical simulation model to form an adaptive CFD numerical simulation model, and numerically simulating the aerodynamic performance of the target large-scale transport equipment under different flight conditions and meteorological conditions according to the selected physical model and the adaptive CFD numerical simulation model; S305. Compare the historical flight data with the numerical simulation, and calibrate the adaptive CFD numerical simulation model according to the verification and comparison results.

5. The aerodynamic force prediction method of large-scale transportation equipment according to claim 4, characterized in that: The adaptive optimization algorithm optimizes the initialization settings of the CFD numerical simulation model based on the multi-agent proximal strategy model and the local alternating genetic model. The working method of the adaptive optimization algorithm includes the following steps: S3021. Consider the initialization setting variables in the CFD numerical simulation model as an intelligent agent, the operating environment of the intelligent agent is the CFD numerical simulation model, and the strategy function and value function of the intelligent agent are defined by using the aerodynamic prediction subject knowledge and the aerodynamic formula. The strategy function adjusts the value of the current initialization setting variables according to the current state of the CFD numerical simulation model and the historical experience of the initialization setting variables, and the value function evaluates the value of the current initialization setting variables according to the current state of the CFD numerical simulation model and the strategy function; S3022, the agent performs adjustment actions in the operating environment according to the policy function, and updates the policy function parameters and value function parameters of the agent based on a proximal policy optimization method, wherein the proximal policy optimization method constrains the update step size of the policy and value functions through a policy loss function and a value loss function; S3023. In the process of updating and iterating the parameters of the strategy function and the value function, the agents are divided into multiple groups, and a local alternating genetic model is used to perform local collaboration and overall competition among the multiple agents. S3024, inputting the updated and iterated initialization setting variable values ​​into the CFD numerical simulation model for numerical calculation, and feeding back the calculation results to the intelligent agent as the current state of the CFD numerical simulation model; S3025. Repeat step S3022 and step S3023 until a predetermined aerodynamic performance simulation is achieved.

6. The aerodynamic prediction method of large-scale transportation equipment according to claim 5, characterized in that: The expression of the policy loss function is: In formula (1), L P (θ) represents the updated loss function of the policy function parameter θ, which is used to measure the difference between the new policy and the old policy, E t represents the expected value of all states and actions at time step t, r t (θ) represents the policy function parameters in state s at time step t t Take action a t The probability ratio of the strategy, represents the advantage estimate at time step t, which is used to measure the advantage of taking actions relative to the average action selection in the state, ∈ represents a hyperparameter used to limit the probability ratio r t The range of variation of (θ), clip(r t (θ),1-∈,1+∈) represents the probability ratio r t (θ) is limited to between 1-∈ and 1+∈, and the expression of the value loss function is: In formula (2), Represents the value function parameter The updated loss function is Indicates that in state s t The value function of the agent is: represents the reward value obtained at time step t, E t represents the expected value of all states and actions at time step t.

7. The aerodynamic prediction method of large-scale transportation equipment according to claim 1, characterized in that: The deep reinforcement learning model extracts aerodynamic prediction features from the adaptive CFD numerical simulation results as aerodynamic prediction input features, and uses a graph neural network to perform knowledge reasoning on the aerodynamic prediction knowledge graph to mine the aerodynamic characteristics inside large transportation equipment.

8. An aerodynamic force prediction device for large-scale transport equipment, applied to an aerodynamic force prediction method for large-scale transport equipment as claimed in any one of claims 1 to 7, characterized in that: include: The data acquisition module acquires the structural parameters, flight condition parameters, meteorological condition parameters and historical flight data of the target large-scale transport equipment, and transmits them to the cloud server for processing and storage; The knowledge graph construction module uses a cloud server to construct an aerodynamic prediction knowledge graph. The cloud server extracts information from structural parameters, flight conditions, meteorological conditions, and historical flight data as nodes of the aerodynamic prediction knowledge graph, and integrates aerodynamic knowledge, aerodynamic formulas, and physical models to form a multi-dimensional prediction framework. Numerical simulation module, which builds an adaptive CFD numerical simulation model based on the aerodynamic prediction knowledge graph and the structural parameters of the target large-scale transportation equipment to simulate aerodynamic performance, and iteratively updates the aerodynamic prediction knowledge graph based on the simulation process data; The aerodynamic prediction module uses a deep reinforcement learning model to predict the aerodynamic performance of large transport equipment and uses the updated aerodynamic prediction knowledge graph for training; The user interaction module uses a graphical interface to view the real-time status of the adaptive CFD numerical simulation model and the aerodynamic performance prediction results, and adjust the flight conditions and meteorological conditions. Then, the aerodynamic performance report of large transport equipment is automatically generated based on the simulation and prediction results.

Images