Optimization and prediction method and system of aerodynamic prediction model, equipment and medium

Abstract

The invention provides an aerodynamic force prediction model optimization and prediction method, system, equipment and medium, and the optimization method comprises the steps: inputting current aerodynamic configuration data in a current sample set into N different current aerodynamic force prediction models, and obtaining N pieces of current aerodynamic force data corresponding to the current aerodynamic configuration data; obtaining first prediction data and second prediction data based on the N pieces of current aerodynamic force data; updating the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set, and iteratively updating the N current aerodynamic force prediction models according to the optimized sample set to obtain N optimized aerodynamic force prediction models; the aerodynamic configuration data is optimized, the N optimized aerodynamic prediction models obtained through optimization have extremely high prediction precision, the global coverage capability is high, the operation cost is low, the global optimal solution can be captured more effectively, and therefore the overall optimization effect is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of model technology, and in particular to an optimization and prediction method, system, device and medium for an aerodynamic prediction model. Background Art

[0002] When optimizing the aerodynamic shape of a product (such as a range hood), a single numerical simulation often requires a significant amount of time. To reduce optimization costs and improve effectiveness, optimization methods based on surrogate models are often employed. A surrogate model (aerodynamic force prediction model) is a mathematical mapping model that establishes an approximate correspondence between a product's aerodynamic shape and its aerodynamic forces. When a specific aerodynamic shape is input, the surrogate model quickly outputs a predicted aerodynamic force, thereby supporting the optimization algorithm's search process. This approach significantly reduces optimization costs by reducing the actual computational effort and the number of experiments, while also improving optimization efficiency.

[0003] In surrogate model-based optimization methods, the accuracy of the surrogate model plays a crucial role. The accuracy of the surrogate model typically depends on the number and distribution of training samples. The distribution of these samples directly determines the surrogate model's ability to approximate the true value, thus affecting the reliability and effectiveness of the optimization results. Therefore, when constructing a surrogate model, it is important to ensure that the sample points fully represent the characteristics of the design space to improve the surrogate model's prediction accuracy and generalization capabilities.

[0004] The existing optimization methods are mainly online optimization methods based on the optimal addition point (optimal value) criterion.

[0005] The online optimization method based on the optimal point addition criterion uses a dynamic update mechanism. After determining the initial aerodynamic shape, it is parameterized to identify the optimized location and optimization range of the range hood. The experimental design uses a sampling method to obtain a large number of sample points and conducts numerical simulations to obtain an aerodynamic shape-aerodynamic sample set. A proxy model is then constructed and optimized using an optimization algorithm. After the optimization is complete, the value of the optimal point obtained by the previous generation of optimization is incorporated into the training point set to update the proxy model. This solution does not require the proxy model to have high accuracy throughout the entire design space, but only requires high accuracy near the optimal value. However, this method has weak global coverage and may result in the optimization results being limited to local optimal solutions.

[0006] Online optimization methods based on the optimal addition criterion excel at local performance and have low cost. This approach only maintains high prediction accuracy near the optimal value, reducing computational costs and resource consumption during the optimization process. However, its global performance is relatively weak, which may result in optimization results being limited to local optimal solutions rather than global optimal solutions.

[0007] Therefore, a new optimization method is urgently needed to solve the above problems. Summary of the Invention

[0008] The technical problem to be solved by the present disclosure is to overcome the defect that the aerodynamic prediction model in the prior art has good global characteristics but high calculation cost, and to provide an optimization and prediction method, system, equipment and medium for the aerodynamic prediction model.

[0009] The present disclosure solves the above technical problems through the following technical solutions:

[0010] In a first aspect, a method for optimizing an aerodynamic force prediction model is provided, the method comprising:

[0011] Get the current sample set of the object to be optimized;

[0012] Wherein, the current sample set includes the current aerodynamic shape data of the object to be optimized, and the current aerodynamic force data corresponding to the current aerodynamic shape data;

[0013] Inputting the current aerodynamic shape data in the current sample set into N different current aerodynamic force prediction models respectively to obtain N current aerodynamic force data corresponding to the current aerodynamic shape data;

[0014] Wherein, N ≥ 3 and is an integer, and each current aerodynamic force prediction model corresponds to a different model architecture;

[0015] Obtaining first prediction data and second prediction data based on the N current aerodynamic force data;

[0016] The first prediction data represents the aerodynamic data corresponding to the current prediction optimal value among the N current aerodynamic data, and the second prediction data represents the current prediction error maximum value among the N current aerodynamic data;

[0017] updating the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set;

[0018] Iteratively updating the N current aerodynamic force prediction models according to the optimized sample set until a model convergence condition is met, thereby obtaining N optimized aerodynamic force prediction models;

[0019] The input of the optimized aerodynamic force prediction model is the aerodynamic shape data to be optimized, and the output is the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0020] Optionally, the step of obtaining first prediction data and second prediction data based on the N current aerodynamic force data includes:

[0021] Processing the N current aerodynamic force data using a first preset algorithm to obtain the first prediction data and the second prediction data;

[0022] Wherein, the first preset algorithm includes any one of an evolutionary algorithm, a genetic algorithm, a teaching and learning algorithm, a particle swarm algorithm, and an annealing algorithm;

[0023] Alternatively, the step of obtaining the first prediction data and the second prediction data based on the N current aerodynamic force data includes:

[0024] comparing the numerical values ​​of the N current aerodynamic force data and selecting the current aerodynamic force data with the largest numerical value as the first prediction data;

[0025] The discreteness between the N current aerodynamic force data is calculated, and the discreteness with the largest value is selected as the second prediction data.

[0026] Optionally, the step of updating the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set includes:

[0027] Acquire first current aerodynamic shape data corresponding to the first predicted data, and calculate first optimized aerodynamic force data corresponding to the first current aerodynamic shape data using a second preset algorithm;

[0028] Acquire second current aerodynamic shape data corresponding to the second predicted data, and calculate second optimized aerodynamic force data corresponding to the second current aerodynamic shape data using the second preset algorithm;

[0029] The first optimized aerodynamic force data, the first current aerodynamic shape data, the second optimized aerodynamic force data, and the second current aerodynamic shape data are added to the current sample set to obtain the optimized sample set.

[0030] Optionally, the second preset algorithm includes a numerical simulation method and / or a wind tunnel test method.

[0031] Optionally, the model convergence condition includes that the number of model iterations exceeds a preset number; or the difference between the first prediction data and the second prediction data of multiple consecutive iterations is lower than a preset convergence tolerance;

[0032] And / or, the step of obtaining the current sample set of the object to be optimized includes:

[0033] Obtaining parameters to be optimized of the object to be optimized;

[0034] Obtaining a parameter boundary value of the parameter to be optimized;

[0035] The current sample set is constructed based on the parameter boundary value.

[0036] In a second aspect, an aerodynamic force prediction method is provided, wherein the aerodynamic force prediction method is implemented based on the optimized aerodynamic force prediction model in the above-mentioned aerodynamic force prediction model optimization method;

[0037] The aerodynamic force prediction method comprises:

[0038] Acquire aerodynamic shape data to be optimized for the object to be predicted;

[0039] The aerodynamic shape data to be optimized is input into the optimized aerodynamic force prediction model to predict and obtain the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0040] In a third aspect, an optimization system for an aerodynamic force prediction model is provided, the optimization system comprising:

[0041] A sample set acquisition module is used to obtain the current sample set of the object to be optimized;

[0042] Wherein, the current sample set includes the current aerodynamic shape data of the object to be optimized, and the current aerodynamic force data corresponding to the current aerodynamic shape data;

[0043] an aerodynamic force prediction module, configured to input the current aerodynamic shape data in the current sample set into N different current aerodynamic force prediction models respectively, to obtain N current aerodynamic force data corresponding to the current aerodynamic shape data;

[0044] Wherein, N ≥ 3 and is an integer, and each current aerodynamic force prediction model corresponds to a different model architecture;

[0045] a data prediction module, configured to obtain first prediction data and second prediction data based on the N current aerodynamic force data;

[0046] The first prediction data represents the aerodynamic data corresponding to the current prediction optimal value among the N current aerodynamic data, and the second prediction data represents the current prediction error maximum value among the N current aerodynamic data;

[0047] a sample set optimization module, configured to update the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set;

[0048] a model updating module, configured to iteratively update the N current aerodynamic force prediction models according to the optimized sample set until a model convergence condition is met, thereby obtaining N optimized aerodynamic force prediction models;

[0049] The input of the optimized aerodynamic force prediction model is the aerodynamic shape data to be optimized, and the output is the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0050] In a fourth aspect, an aerodynamic force prediction system is provided, wherein the aerodynamic force prediction system is implemented based on the optimized aerodynamic force prediction model in the aerodynamic force prediction model optimization system described above;

[0051] The aerodynamic force prediction system comprises:

[0052] An aerodynamic shape acquisition module, used to acquire aerodynamic shape data to be optimized for the object to be predicted;

[0053] The aerodynamic force prediction module is used to input the aerodynamic shape data to be optimized into the optimized aerodynamic force prediction model to predict the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0054] In a fifth aspect, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein when the processor executes the computer program, the optimization method of the aerodynamic prediction model described above, or the aerodynamic prediction method described above is implemented.

[0055] In a sixth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the optimization method of the aerodynamic force prediction model or the aerodynamic force prediction method described above is implemented.

[0056] On the basis of conforming to the common sense in this field, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present disclosure.

[0057] The positive progress of this disclosure is:

[0058] The optimization and prediction method, system, device and medium of the aerodynamic prediction model disclosed in the present invention obtain N current aerodynamic data corresponding to the current aerodynamic data by inputting the current aerodynamic shape data in the current sample set into N different current aerodynamic prediction models respectively; then obtain first prediction data and second prediction data, the first prediction data represents the aerodynamic data corresponding to the current prediction optimal value in the N current aerodynamic data, and the second prediction data represents the maximum current prediction error in the N current aerodynamic data; the current sample set is updated according to the first prediction data and the second prediction data to obtain an optimized sample set, thereby realizing the optimization of the aerodynamic shape data; at the same time, the N current aerodynamic prediction models are optimized according to the optimized sample set to obtain N optimized aerodynamic prediction models, the optimized aerodynamic prediction models have extremely high prediction accuracy, strong global coverage capability and low computational cost, can more effectively capture the global optimal solution, thereby improving the overall optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A first flow chart of the method for optimizing the aerodynamic force prediction model provided in Example 1 of the present disclosure;

[0060] Figure 2 A second flow chart of the method for optimizing the aerodynamic force prediction model provided in Example 1 of the present disclosure;

[0061] Figure 3 A third flow chart of the method for optimizing the aerodynamic force prediction model provided in Example 1 of the present disclosure;

[0062] Figure 4 A fourth flow chart of the method for optimizing the aerodynamic force prediction model provided in Example 1 of the present disclosure;

[0063] Figure 5 A fifth flow chart of the method for optimizing the aerodynamic force prediction model provided in Example 1 of the present disclosure;

[0064] Figure 6 A schematic flow chart of the aerodynamic force prediction method provided in Example 2 of the present disclosure;

[0065] Figure 7 A schematic diagram of the structure of an optimization system for an aerodynamic force prediction model provided in Example 3 of the present disclosure;

[0066] Figure 8 A schematic diagram of the structure of the aerodynamic force prediction system provided in Example 4 of the present disclosure;

[0067] Figure 9 This is a structural diagram of an electronic device provided in Example 5 of the present disclosure. DETAILED DESCRIPTION

[0068] The present disclosure is further illustrated below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0069] In the embodiments of the present disclosure, prefixes such as "first" and "second" are used only to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. In the embodiments of the present disclosure, the use of prefixes such as ordinal numbers to distinguish description objects does not constitute a limitation on the described objects. For the statement of the described objects, please refer to the description in the context of the claims or embodiments, and no unnecessary limitations should be constituted due to the use of such prefixes. In addition, in the description of this embodiment, unless otherwise specified, the meaning of "plurality" is two or more.

[0070] Example 1

[0071] This embodiment provides an optimization method for an aerodynamic prediction model, such as Figure 1 As shown, the optimization method includes:

[0072] S1. Obtain the current sample set of the object to be optimized.

[0073] S2. Inputting the current aerodynamic shape data in the current sample set into N different current aerodynamic force prediction models respectively to obtain N current aerodynamic force data corresponding to the current aerodynamic shape data.

[0074] S3. Obtain first prediction data and second prediction data based on N current aerodynamic force data.

[0075] S4. Update the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set.

[0076] S5. Iteratively update N current aerodynamic force prediction models according to the optimized sample set until a model convergence condition is met, thereby obtaining N optimized aerodynamic force prediction models.

[0077] The current sample set includes the current aerodynamic shape data of the object to be optimized and the current aerodynamic force data corresponding to the current aerodynamic shape data.

[0078] The first prediction data represents aerodynamic data corresponding to a current prediction optimal value among the N current aerodynamic data, and the second prediction data represents a current prediction error maximum value among the N current aerodynamic data.

[0079] Specifically, the optimal value is the global optimal aerodynamic value that can be obtained from the N current aerodynamic data predicted by the current aerodynamic prediction model. The maximum error is the maximum estimated error in the current aerodynamic prediction model.

[0080] Each round of optimization can use new first prediction data and new second prediction data. Multiple rounds of updates are performed based on the first and second prediction data of each round until N optimized aerodynamic force prediction models are obtained. The input of the optimized aerodynamic force prediction model is the aerodynamic shape data to be optimized, and the output is the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0081] Each current aerodynamic prediction model corresponds to a different model architecture. These models differ primarily in their mathematical principles and model architecture (structure). Common aerodynamic prediction models include the polynomial response surface model (based on polynomial regression), the Kriging model (based on Gaussian process interpolation), the neural network model (based on artificial neural network architecture), and the radial basis function model (based on radial symmetric function interpolation).

[0082] Specifically, N ≥ 3 and is an integer. For example, when N is 3, the current aerodynamic force prediction model 1 is a polynomial response surface model, the current aerodynamic force prediction model 2 is a Kriging model, and the current aerodynamic force prediction model 3 is a neural network model; alternatively, the current aerodynamic force prediction model 1 is a radial basis function model, the current aerodynamic force prediction model 2 is a Kriging model, and the current aerodynamic force prediction model 3 is a polynomial response surface model; alternatively, the current aerodynamic force prediction model 1 is a radial basis function model, the current aerodynamic force prediction model 2 is a support vector machine model, and the current aerodynamic force prediction model 3 is a polynomial response surface model.

[0083] The optimization method of the aerodynamic prediction model of this embodiment updates the current sample set according to the first prediction data and the second prediction data to obtain the optimized sample set, thereby realizing the optimization of the current aerodynamic shape data; at the same time, N current aerodynamic prediction models are optimized according to the optimized sample set to obtain N optimized aerodynamic prediction models. The optimized aerodynamic prediction model has extremely high prediction accuracy, strong global coverage capability and low computational cost, and can more effectively capture the global optimal solution, thereby improving the overall optimization effect.

[0084] In an optional embodiment, if Figure 2 As shown, the above step S3 includes:

[0085] S31 . Process N current aerodynamic force data using a first preset algorithm to obtain first prediction data and second prediction data.

[0086] The first preset algorithm includes any one of an evolutionary algorithm, a genetic algorithm, a teaching and learning algorithm, a particle swarm algorithm, and an annealing algorithm.

[0087] This embodiment uses a first preset algorithm to find the predicted optimal value and the maximum prediction error (maximum uncertainty), and adds the first prediction data corresponding to the predicted optimal value and the second prediction data corresponding to the maximum prediction error to the current sample set, thereby updating the current sample set, obtaining an optimized sample set, and optimizing the aerodynamic shape data.

[0088] In an optional embodiment, if Figure 3 As shown, the above step S3 includes:

[0089] S32 , comparing the values ​​of N current aerodynamic force data, and selecting the current aerodynamic force data with the largest value as the first prediction data.

[0090] S33. Calculate the discreteness between the N current aerodynamic force data, and select the discreteness with the largest value as the second prediction data.

[0091] The extreme values ​​of the prediction errors of each current aerodynamic prediction model are calculated through an optimization algorithm, and the discreteness of the model prediction results is quantified using statistics such as variance or standard deviation. Subsequently, the maximum value of the prediction error in the global range (the second prediction data) is screened out through cross-comparison. This extreme value represents the maximum deviation area of ​​the prediction data of the N current aerodynamic prediction models. Finally, the aerodynamic shape data corresponding to the maximum prediction error is used as the output result of the current iteration step.

[0092] This embodiment updates the current sample set by adding the first prediction data corresponding to the optimal prediction value and the second prediction data corresponding to the maximum prediction error value to the current sample set, thereby obtaining an optimized sample set.

[0093] In an optional embodiment, if Figure 4 As shown, the above step S4 includes:

[0094] S41. Acquire first current aerodynamic shape data corresponding to the first prediction data, and use a second preset algorithm to calculate first optimized aerodynamic force data corresponding to the first current aerodynamic shape data.

[0095] S42. Obtain second current aerodynamic shape data corresponding to the second predicted data, and use a second preset algorithm to calculate second optimized aerodynamic force data corresponding to the second current aerodynamic shape data.

[0096] S43. Add the first optimized aerodynamic force data, the first current aerodynamic shape data, the second optimized aerodynamic force data, and the second current aerodynamic shape data to the current sample set to obtain an optimized sample set.

[0097] Specifically, the second preset algorithm includes a numerical simulation method and / or a wind tunnel test method.

[0098] Because the first and second predicted data are model predictions and thus contain certain errors, a second preset algorithm is required to calculate the actual aerodynamic force data corresponding to the first current aerodynamic shape data (i.e., the first optimized aerodynamic force data) and the actual aerodynamic force data corresponding to the second current aerodynamic shape data (i.e., the second optimized aerodynamic force data). After each iteration of generating new current aerodynamic shape data, high-precision numerical simulation verification or wind tunnel testing is required to obtain real aerodynamic force data to calibrate the model's prediction errors. This closed-loop mechanism of "model prediction first, real-world verification later" leverages the model's advantage in accelerating global exploration while ensuring the credibility of the final optimization results.

[0099] Since the second prediction data represents the maximum current prediction error among the N current aerodynamic force data, the second prediction data can be used to identify the region with the largest prediction error (region with higher uncertainty). In existing offline optimization methods, the number of sample points depends on the sampling criterion used. To ensure that the sample points accurately reflect the trend of the objective function across the entire design space, a large number of initial sample points is typically required. However, in actual optimization, in regions with low uncertainty, a small number of sample points can effectively fit the relationship between the design variables and the objective function. Excessive sample points have limited impact on accuracy and may even lead to waste of resources. Therefore, to construct a globally high-precision aerodynamic force prediction model, the sample point distribution density must be rationally controlled, with a priority on increasing the sample point density in regions with higher uncertainty. These regions typically contain more critical information. Targeted sampling can not only significantly improve the model's global accuracy but also effectively avoid inefficient resource usage.

[0100] Figure 5 A schematic flow chart of an optimization method for an aerodynamic force prediction model provided in this embodiment is shown as follows: Figure 5 As shown, first, the initial aerodynamic shape of the object to be optimized is determined, the initial aerodynamic shape is parameterized, and the parameters of the object to be optimized and the parameter optimization range are clarified. The experimental design adopts a sampling method to obtain a large number of sample points, and numerical simulation is performed to obtain an aerodynamic shape-aerodynamic force sample set (current sample set), and then N current aerodynamic force prediction models are constructed. According to the output of the model, the aerodynamic shape (first current aerodynamic shape data) corresponding to the predicted optimal value (first predicted data) and the aerodynamic force corresponding to the aerodynamic shape (first optimized aerodynamic force data) are obtained, as well as the aerodynamic shape (second current aerodynamic shape data) corresponding to the maximum prediction error (second predicted data) and the aerodynamic force corresponding to the aerodynamic shape (second optimized aerodynamic force data). These two sets of data are then added to the aerodynamic shape-aerodynamic force sample set (i.e., the current sample set) to optimize and update the current aerodynamic force prediction model. After the optimization is complete, the aerodynamic shape with the optimal model prediction value and its corresponding real aerodynamic force, as well as the aerodynamic shape with the maximum model prediction error and its corresponding real aerodynamic force, are added to the current sample set to obtain an updated sample set. This is used to update the current aerodynamic force prediction model, and when the model converges, the optimized aerodynamic force prediction model is obtained. At the same time, the optimal aerodynamic shape estimated by the last generation model is the optimal aerodynamic shape obtained in this optimization.

[0101] In an optional embodiment, the model convergence condition includes that the number of model iterations exceeds a preset number; or the difference between the first prediction data and the second prediction data of multiple consecutive iterations is lower than a preset convergence tolerance.

[0102] For example, the convergence tolerance can be set to 1e-2 or 5e-3.

[0103] In an optional embodiment, the step of obtaining the current sample set of the object to be optimized includes:

[0104] Obtaining parameters to be optimized of the object to be optimized; obtaining parameter boundary values ​​of the parameters to be optimized; and constructing a current sample set based on the parameter boundary values.

[0105] Specifically, the object to be optimized can be a component in the range hood, and the parameters to be optimized can be blade geometric parameters (such as inlet angle, outlet angle, and bending radius), volute and flow channel structure optimization parameters (such as volute tongue gap, helix angle, etc.), and air inlet channel and smoke hood parameters (such as smoke collection port shape, angle, and negative pressure zone distribution).

[0106] The above-mentioned shape parameters can be used as optimization variables in the optimization process. However, it should be noted that the range of optimization variables usually needs to be set by the designer according to the actual working conditions and optimization constraints.

[0107] Parameterization methods, including but not limited to FFD (Free Form Deformation), CST (Case-like Shape Transformation), and NURBS (Non-Uniform Rational B-Splines), are used to obtain the parameters of the object to be optimized. These methods convert complex geometry into adjustable parameters, enabling efficient modification of the object's aerodynamic shape. They are also used to identify the optimal location and optimization interval (parameter boundary values) of the object to be optimized.

[0108] The working principle of the present disclosure is further explained below with reference to specific examples:

[0109] For example, to optimize a range hood impeller, the optimization objective is the airflow rate (the amount of air extracted per unit time). After parametric design, the three optimization variables (i.e., parameters) are selected for optimization: the blade inlet angle, the blade outlet angle, and the number of blades (i.e., the object to be optimized). The upper and lower bounds (parameter boundary values) of the optimization variables are determined based on engineering constraints and designer experience. The upper and lower bounds are set to [50°, 150°, 20] and [75°, 170°, 40], respectively.

[0110] After selecting the optimization variables, we need to construct an aerodynamic force prediction model. First, we conduct an experimental design. Using Latin hypercube sampling, we select 2n data sets (n is the number of design variables). Here, we have six initial points (the current aerodynamic shape data). These are [72°, 166°, 29], [60°, 160°, 25], [75°, 150°, 33], [66°, 155°, 30], [51°, 169°, 40], and [56°, 163°, 21]. Numerical simulations yielded airflow rates of 10.3 m³ / min, 11 m³ / min, 9.5 m³ / min, 8.2 m³ / min, and 10.1 m³ / min, respectively (the current aerodynamic force data).

[0111] The Kriging model, neural network model, and response surface model were selected as the aerodynamic force prediction models used in this optimization. Based on the five sets of data obtained above, the current aerodynamic force prediction model was constructed.

[0112] The three models were then optimized using the aerodynamic force prediction model optimization method disclosed herein, yielding the optimal prediction value and maximum prediction error. The optimal prediction value was 12 m³ / min, corresponding to the first current aerodynamic shape data set of [62°, 159°, 28]. Uncertainty was assessed using variance. The maximum model prediction error was 14, corresponding to the second current aerodynamic shape data set of [78°, 152°, 40].

[0113] Numerical simulations were performed on the two aerodynamic shape data sets to obtain airflow. The model-estimated optimal shape (first current aerodynamic shape data) [62°, 159°, 28] corresponds to a true airflow (first optimized aerodynamic force data) of 11.7 m³ / min. The model-estimated shape with the maximum error (second current aerodynamic shape data) [78°, 152°, 40] corresponds to a true airflow (second optimized aerodynamic force data) of 10.6 m³ / min. These two aerodynamic shape data sets and their corresponding true values ​​were added to the current dataset to obtain the optimized dataset, which was then used to update the model. The model was then optimized again using the optimization method to obtain the current model's estimated optimal value and maximum error. This process continued until convergence, resulting in the optimized aerodynamic force prediction model. This optimization of the current aerodynamic force prediction model was achieved.

[0114] At the same time, the optimal aerodynamic shape estimated by the last generation model is the optimal aerodynamic shape obtained in this optimization, realizing the optimization of the aerodynamic shape data.

[0115] Example 2

[0116] This embodiment provides an aerodynamic force prediction method, which is implemented based on the optimized aerodynamic force prediction model in the aerodynamic force prediction model optimization method in Example 1.

[0117] like Figure 6 As shown in Figure 2, the aerodynamic force prediction method includes:

[0118] S01. Obtain aerodynamic shape data to be optimized of an object to be predicted.

[0119] S02. Inputting the aerodynamic shape data to be optimized into the optimized aerodynamic force prediction model to predict the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0120] Specifically, the aerodynamic shape data to be optimized can be input into any one of the N optimized aerodynamic force prediction models in Example 1.

[0121] The aerodynamic force prediction method of this embodiment is implemented based on the optimized aerodynamic force prediction model in Example 1. The optimized aerodynamic force prediction model has extremely high prediction accuracy, strong global coverage capability and low computational cost, and can more effectively capture the global optimal solution. With the help of the optimized aerodynamic force prediction model, the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized can be accurately predicted, thereby improving the user experience.

[0122] Example 3

[0123] Corresponding to the optimization method of the aerodynamic force prediction model in the aforementioned embodiment 1, this embodiment provides an optimization system for the aerodynamic force prediction model, such as Figure 7 As shown, the optimization system includes:

[0124] The sample set acquisition module 1 is used to obtain the current sample set of the object to be optimized;

[0125] The current sample set includes the current aerodynamic shape data of the object to be optimized and the current aerodynamic force data corresponding to the current aerodynamic shape data;

[0126] an aerodynamic force prediction module 2, configured to input the current aerodynamic shape data in the current sample set into N different current aerodynamic force prediction models, and obtain N current aerodynamic force data corresponding to the current aerodynamic shape data;

[0127] Wherein, N ≥ 3 and is an integer, and each current aerodynamic force prediction model corresponds to a different model architecture;

[0128] A data prediction module 3, configured to obtain first prediction data and second prediction data based on N current aerodynamic force data;

[0129] The first prediction data represents the aerodynamic data corresponding to the current prediction optimal value among the N current aerodynamic data, and the second prediction data represents the maximum value of the current prediction error among the N current aerodynamic data;

[0130] A sample set optimization module 4 is configured to update the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set;

[0131] A model updating module 5 is configured to iteratively update N current aerodynamic force prediction models according to the optimized sample set until the model convergence condition is met, thereby obtaining N optimized aerodynamic force prediction models;

[0132] The input of the optimized aerodynamic force prediction model is the aerodynamic shape data to be optimized, and the output is the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0133] In an optional embodiment, the data prediction module 3 includes:

[0134] A first prediction unit 31 is configured to process N current aerodynamic force data using a first preset algorithm to obtain first prediction data and second prediction data;

[0135] The first preset algorithm includes any one of an evolutionary algorithm, a genetic algorithm, a teaching and learning algorithm, a particle swarm algorithm, and an annealing algorithm;

[0136] In an optional embodiment, the data prediction module 3 includes:

[0137] The second prediction unit 32 is used to compare the values ​​of N current aerodynamic force data and select the current aerodynamic force data with the largest value as the first prediction data;

[0138] The third prediction unit 33 is used to calculate the discreteness between the N current aerodynamic force data, and select the discreteness with the largest value as the second prediction data.

[0139] In an optional embodiment, the sample set optimization module 4 includes:

[0140] The first optimization unit 41 is configured to obtain first current aerodynamic shape data corresponding to the first prediction data, and calculate first optimized aerodynamic force data corresponding to the first current aerodynamic shape data using a second preset algorithm;

[0141] The second optimization unit 42 is configured to obtain second current aerodynamic shape data corresponding to the second prediction data, and calculate second optimized aerodynamic force data corresponding to the second current aerodynamic shape data using a second preset algorithm;

[0142] The sample updating unit 43 is configured to add the first optimized aerodynamic force data, the first current aerodynamic shape data, the second optimized aerodynamic force data, and the second current aerodynamic shape data to the current sample set to obtain an optimized sample set.

[0143] In an optional embodiment, the second preset algorithm includes a numerical simulation method and / or a wind tunnel test method.

[0144] In an optional embodiment, the model convergence condition includes that the number of model iterations exceeds a preset number; or the difference between the first prediction data and the second prediction data of multiple consecutive iterations is lower than a preset convergence tolerance.

[0145] In an optional implementation manner, the sample set acquisition module 1 is used to acquire parameters to be optimized of the object to be optimized; acquire parameter boundary values ​​of the parameters to be optimized; and construct a current sample set based on the parameter boundary values.

[0146] The optimization system of the aerodynamic prediction model of this embodiment updates the current sample set according to the first prediction data and the second prediction data to obtain the optimized sample set, thereby realizing the optimization of the current aerodynamic shape data; at the same time, N current aerodynamic prediction models are optimized according to the optimized sample set to obtain N optimized aerodynamic prediction models. The optimized aerodynamic prediction model has extremely high prediction accuracy, strong global coverage capability and low computational cost, and can more effectively capture the global optimal solution, thereby improving the overall optimization effect.

[0147] As for the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment, which will not be repeated here.

[0148] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components of the units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the disclosed solution.

[0149] Example 4

[0150] Corresponding to the aerodynamic force prediction method in the aforementioned embodiment 2, this embodiment provides an aerodynamic force prediction system, which is implemented based on the optimized aerodynamic force prediction model in the optimization system of the aerodynamic force prediction model in embodiment 3; Figure 8 As shown in Figure 1, the aerodynamic force prediction system includes:

[0151] The aerodynamic shape acquisition module 6 is used to obtain the aerodynamic shape data to be optimized for the object to be predicted;

[0152] The aerodynamic force prediction module 7 is used to input the aerodynamic shape data to be optimized into the optimized aerodynamic force prediction model, and predict the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

[0153] The aerodynamic force prediction system of this embodiment is implemented based on the optimized aerodynamic force prediction model in Example 3. The optimized aerodynamic force prediction model has extremely high prediction accuracy, strong global coverage capability and low computational cost, and can more effectively capture the global optimal solution. With the help of the optimized aerodynamic force prediction model, the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized can be accurately predicted, thereby improving the user experience.

[0154] As for the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment, which will not be repeated here.

[0155] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components of the units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the disclosed solution.

[0156] Example 5

[0157] Figure 9 This is a structural diagram of an electronic device showing an example embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and for running on the processor. When the processor executes the computer program, the optimization method of the aerodynamic force prediction model or the aerodynamic force prediction method provided in the above embodiment is implemented. Figure 9 The electronic device 90 shown is only an example and should not limit the functionality and scope of use of the embodiments of the present disclosure.

[0158] like Figure 9 As shown, the electronic device 90 may be a general-purpose computing device, such as a server device. Components of the electronic device 90 may include, but are not limited to, the at least one processor 91, the at least one memory 92, and a bus 93 connecting different system components (including the memory 92 and the processor 91).

[0159] The bus 93 includes a data bus, an address bus, and a control bus.

[0160] The memory 92 may include a volatile memory, such as a random access memory (RAM) 921 and / or a cache memory 922 , and may further include a read-only memory (ROM) 923 .

[0161] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) of program modules 924, such program modules 924 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include an implementation of a network environment.

[0162] The processor 91 executes various functional applications and data processing by running computer programs stored in the memory 92 , such as the aerodynamic force prediction model optimization method or the aerodynamic force prediction method provided in the above embodiments.

[0163] The electronic device 90 can also communicate with one or more external devices 94 (e.g., a keyboard, pointing device, etc.). This communication can occur via an input / output (I / O) interface 95. Furthermore, the electronic device 90 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 96. As shown, the network adapter 96 communicates with other modules of the electronic device 90 via a bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 90, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0164] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0165] Example 5

[0166] The embodiments of the present disclosure also provide a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the optimization method of the aerodynamic force prediction model or the aerodynamic force prediction method provided in the above embodiments is implemented.

[0167] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0168] Example 6

[0169] The embodiments of the present disclosure also provide a computer program product, including a computer program. When the computer program is executed by a processor, it implements the optimization method of the aerodynamic force prediction model or the aerodynamic force prediction method provided in the above embodiments.

[0170] The program code for executing the computer program product of the present disclosure may be written in any combination of one or more programming languages, and the program code may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0171] While specific embodiments of the present disclosure have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of protection of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications are intended to fall within the scope of protection of the present disclosure.

Claims

1. A method for optimizing an aerodynamic force prediction model, characterized in that: The optimization method comprises: Get the current sample set of the object to be optimized; Wherein, the current sample set includes the current aerodynamic shape data of the object to be optimized, and the current aerodynamic force data corresponding to the current aerodynamic shape data; Inputting the current aerodynamic shape data in the current sample set into N different current aerodynamic force prediction models respectively to obtain N current aerodynamic force data corresponding to the current aerodynamic shape data; Wherein, N ≥ 3 and is an integer, and each current aerodynamic force prediction model corresponds to a different model architecture; Obtaining first prediction data and second prediction data based on the N current aerodynamic force data; The first prediction data represents the aerodynamic data corresponding to the current prediction optimal value among the N current aerodynamic data, and the second prediction data represents the current prediction error maximum value among the N current aerodynamic data; updating the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set; Iteratively updating the N current aerodynamic force prediction models according to the optimized sample set until a model convergence condition is met, thereby obtaining N optimized aerodynamic force prediction models; The input of the optimized aerodynamic force prediction model is the aerodynamic shape data to be optimized, and the output is the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

2. The method for optimizing the aerodynamic force prediction model according to claim 1, characterized in that: The step of obtaining first prediction data and second prediction data based on the N current aerodynamic force data includes: Processing the N current aerodynamic force data using a first preset algorithm to obtain the first prediction data and the second prediction data; Wherein, the first preset algorithm includes any one of an evolutionary algorithm, a genetic algorithm, a teaching and learning algorithm, a particle swarm algorithm, and an annealing algorithm; Alternatively, the step of obtaining the first prediction data and the second prediction data based on the N current aerodynamic force data includes: comparing the numerical values ​​of the N current aerodynamic force data and selecting the current aerodynamic force data with the largest numerical value as the first prediction data; The discreteness between the N current aerodynamic force data is calculated, and the discreteness with the largest value is selected as the second prediction data.

3. The method for optimizing the aerodynamic force prediction model according to claim 1, characterized in that: The step of updating the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set includes: Acquire first current aerodynamic shape data corresponding to the first predicted data, and calculate first optimized aerodynamic force data corresponding to the first current aerodynamic shape data using a second preset algorithm; Acquire second current aerodynamic shape data corresponding to the second predicted data, and calculate second optimized aerodynamic force data corresponding to the second current aerodynamic shape data using the second preset algorithm; The first optimized aerodynamic force data, the first current aerodynamic shape data, the second optimized aerodynamic force data, and the second current aerodynamic shape data are added to the current sample set to obtain the optimized sample set.

4. The method for optimizing the aerodynamic force prediction model according to claim 3, characterized in that: The second preset algorithm includes a numerical simulation method and / or a wind tunnel test method.

5. The method for optimizing the aerodynamic force prediction model according to claim 1, characterized in that: The model convergence condition includes that the number of model iterations exceeds a preset number; or the difference between the first prediction data and the second prediction data of multiple consecutive iterations is lower than a preset convergence tolerance; And / or, the step of obtaining the current sample set of the object to be optimized includes: Obtaining parameters to be optimized of the object to be optimized; Obtaining a parameter boundary value of the parameter to be optimized; The current sample set is constructed based on the parameter boundary value.

6. A method for predicting aerodynamic forces, characterized in that: The aerodynamic force prediction method is implemented based on the optimized aerodynamic force prediction model in the aerodynamic force prediction model optimization method according to any one of claims 1 to 5; The aerodynamic force prediction method comprises: Acquire aerodynamic shape data to be optimized for the object to be predicted; The aerodynamic shape data to be optimized is input into the optimized aerodynamic force prediction model to predict and obtain the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

7. An optimization system for an aerodynamic prediction model, characterized in that: The optimization system comprises: A sample set acquisition module is used to obtain the current sample set of the object to be optimized; Wherein, the current sample set includes the current aerodynamic shape data of the object to be optimized, and the current aerodynamic force data corresponding to the current aerodynamic shape data; an aerodynamic force prediction module, configured to input the current aerodynamic shape data in the current sample set into N different current aerodynamic force prediction models respectively, to obtain N current aerodynamic force data corresponding to the current aerodynamic shape data; Wherein, N ≥ 3 and is an integer, and each current aerodynamic force prediction model corresponds to a different model architecture; a data prediction module, configured to obtain first prediction data and second prediction data based on the N current aerodynamic force data; The first prediction data represents the aerodynamic data corresponding to the current prediction optimal value among the N current aerodynamic data, and the second prediction data represents the current prediction error maximum value among the N current aerodynamic data; a sample set optimization module, configured to update the current sample set based on the first prediction data and the second prediction data to obtain an optimized sample set; a model updating module, configured to iteratively update the N current aerodynamic force prediction models according to the optimized sample set until a model convergence condition is met, thereby obtaining N optimized aerodynamic force prediction models; The input of the optimized aerodynamic force prediction model is the aerodynamic shape data to be optimized, and the output is the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

8. An aerodynamic force prediction system, characterized in that: The aerodynamic force prediction system is implemented based on the optimized aerodynamic force prediction model in the aerodynamic force prediction model optimization system as claimed in claim 7; The aerodynamic force prediction system comprises: An aerodynamic shape acquisition module, used to acquire aerodynamic shape data to be optimized for the object to be predicted; The aerodynamic force prediction module is used to input the aerodynamic shape data to be optimized into the optimized aerodynamic force prediction model to predict the optimal aerodynamic force data corresponding to the aerodynamic shape data to be optimized.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and configured to run on the processor, wherein: When the processor executes the computer program, the optimization method of the aerodynamic force prediction model according to any one of claims 1 to 5 or the aerodynamic force prediction method according to claim 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for optimizing the aerodynamic force prediction model according to any one of claims 1 to 5 or the method for predicting aerodynamic force according to claim 6 is implemented.

Images