A wind tunnel test balance structure optimization design method and system
By using parametric modeling, finite element simulation, and intelligent optimization algorithms, combined with machine learning, the structural parameters of the balance are optimized, solving the problem of time-consuming and labor-intensive traditional design and achieving efficient and accurate balance design.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- AVIC SHENYANG AERODYNAMICS RES INST
- Filing Date
- 2026-03-02
- Publication Date
- 2026-05-29
AI Technical Summary
Traditional balance design methods are time-consuming and labor-intensive, making it difficult to achieve optimal results, and they also have low design accuracy and efficiency.
By employing parametric modeling, finite element simulation, intelligent optimization algorithms, and machine learning algorithms, combined with ANSYS and DOE methods, an intelligent model is established to optimize the structural parameters of the balance, thereby achieving automated simulation and design.
It improves the accuracy and efficiency of balance design, optimizes structural parameters, reduces repetitive work, and provides convenience.
Smart Images
Figure CN122113638A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of wind tunnel test balance design technology, specifically relating to a method and system for optimizing the design of a wind tunnel test balance structure. Background Technology
[0002] Traditional balance design methods rely on the balance designer's experience to initially select a suitable balance structure and structural parameters for finite element simulation. Based on the simulation results, the structure and parameters are then modified in a targeted manner until the basic design requirements are met. This approach is not only labor-intensive but also makes it difficult to obtain the optimal solution for the balance structure. To address these issues, parametric modeling, finite element simulation, and intelligent optimization algorithms are used to quickly and accurately find the optimal solution for the structural parameters. With the continuous acquisition and accumulation of large amounts of simulation data, an intelligent model based on machine learning algorithms is established to predict the optimization objective from the model parameters, further improving the accuracy and efficiency of balance design. Based on these methods, a wind tunnel test balance optimization design system has been developed, replacing a large amount of repetitive and tedious work in traditional wind tunnel balance structural design, providing significant convenience for balance designers. Summary of the Invention
[0003] The problem this invention aims to solve is to achieve intelligent optimization design of wind tunnel balance structures, while improving the performance and design efficiency of the balances. It proposes a method and system for optimizing the design of wind tunnel test balance structures.
[0004] To achieve the above objectives, the present invention provides the following technical solution:
[0005] A method for optimizing the design of a wind tunnel test balance structure includes the following steps:
[0006] S1. Determine the input parameters of the balance according to its design shape and requirements;
[0007] S2. Based on the input parameters of the balance obtained in step S1, a three-dimensional model of the balance is generated using an assembly modeling method;
[0008] S3. Taking into account the strength of the balance structure, the maximum stress at the strain gauge bonding location, and the mutual interference between components, an optimization model is established;
[0009] S4. Based on the DOE method, conduct experimental design and parameter analysis on the optimized model obtained in step S3. Combined with correlation analysis, obtain the degree of influence of the input parameters of the balance on the measurement performance of the balance, and extract the key parameters of the balance.
[0010] S5. Based on the three-dimensional model of the balance under different parameters obtained in step S2 and the key parameters of the balance obtained in step S4, the parameterized design language command flow of ANSYS is used to perform parameterized automatic simulation on the optimized model in step S3, generating sample data of the key parameters of the balance and the optimization target values corresponding to the simulation analysis results, forming a key parameter-optimization target sample pair of the balance.
[0011] S6. Based on the intelligent optimization algorithm, optimize the key parameters of the balance obtained in step S5 - the key parameters of the target sample pair of the balance, and extract the optimal solution of the target value.
[0012] S7. Repeat steps S5 and S6 to obtain the key parameters of the balance through multiple simulations - optimize the target sample pair dataset;
[0013] S8. Based on the dataset obtained in step S7, train a machine learning regression model to obtain an intelligent mathematical proxy model for the structural optimization design of the wind tunnel balance.
[0014] Furthermore, in step S1, based on the application scenario of the balance, the requirements of the measurement range, and the structural strength, stiffness, and sensitivity requirements of the balance, the input parameters of the balance are determined, including the diameter, the length, width, height and position dimensions of the measuring element, the thickness, gap dimension, number of support plates, position dimensions and slot angle of the support plates; the input parameters of the balance with the axial force measuring unit also include the form of the axial force element support plate, and the specific parameters and dimensions of the axial force element support plate.
[0015] Furthermore, during the modeling process in step S2, it is checked whether the maximum size of the beam cross section exceeds the diameter of the balance, whether the components are in good contact, and whether there is any overlap or offset interference between the models, so as to ensure that the size and position of the balance components are in a coordinated state.
[0016] Furthermore, in step S3, an optimization model is established with the optimization objectives of ensuring that the design strain output of each component of the balance is reasonable and that the strength at the key position of the balance meets the material yield strength requirements.
[0017] Furthermore, the DOE method in step S4 is implemented based on the orthogonal array method, the Latin hypercube method, and the optimized Latin hypercube method.
[0018] Furthermore, the parameterized automatic simulation process in step S5 is as follows: Based on the batch processing method, the APDL script program is run to start ANSYS and the Mechanical APDL module. By batch setting the material properties, load positions and sizes, mesh size and position, and fixed end constraints in the script language, the automatic simulation process of the model driven by different parameters is realized, and sample data of the optimized target values corresponding to the parameter values and simulation analysis results are generated.
[0019] Furthermore, the intelligent optimization algorithm in step S6 is an improved genetic algorithm NSGA-II, and the key parameters of the optimized balance include the thickness of the measuring element. Support beam length Width of the support beam Width of the slot in the axial force element of the balance Distance from the center of the balance to the support beam .
[0020] Furthermore, the machine learning regression model in step S8 includes one of support vector machines, neural networks, and Bayesian regression models.
[0021] A wind tunnel test balance structure optimization design system includes a processor, a memory, and a computer program stored in the memory and run on the processor. When the computer program runs, it implements the steps of the wind tunnel test balance structure optimization design method.
[0022] The beneficial effects of this invention are:
[0023] This invention discloses a wind tunnel test balance structural optimization design method. It comprehensively considers the balance structure's strength, stiffness, sensitivity, maximum stress at strain gauge attachment points, inter-component interference, and structural thermal deformation. The method establishes an optimization model with the goals of reasonable design strain output for each component, minimal interference between components, and ensuring the strength at key locations meets material yield strength requirements. This process involves batch modeling, simulation, dimensional parameter optimization, and automatic design, simulation, and modeling based on optimal dimensions. Through batch modeling and simulation, a data sample library is constructed between balance dimensional parameters and optimization objectives. Based on this library, a machine learning intelligent regression model is trained. For newly designed balances, the trained model is used, and based on the optimization algorithm, model dimensional parameters that satisfy the optimization objectives are searched. This forms an intelligent optimization design process for wind tunnel balances based on optimized model dimensional parameters and automatic parameterized modeling, improving the accuracy and efficiency of balance design. Attached Figure Description
[0024] Figure 1 This is a flowchart of a wind tunnel test balance structure optimization design method according to the present invention. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.
[0026] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.
[0027] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 Detailed explanation is as follows:
[0028] Example 1:
[0029] A method for optimizing the design of a wind tunnel test balance structure includes the following steps:
[0030] S1. Determine the input parameters of the balance according to its design shape and requirements;
[0031] Furthermore, in step S1, based on the application scenario of the balance, the requirements of the measurement range, and the structural strength, stiffness, and sensitivity requirements of the balance, the input parameters of the balance are determined, including the diameter, the length, width, height and position dimensions of the measuring element, the thickness, gap dimension, number of support plates, position dimensions and slot angle of the support plates; the input parameters of the balance with the axial force measuring unit also include the form of the axial force element support plate, and the specific parameters and dimensions of the axial force element support plate.
[0032] S2. Based on the input parameters of the balance obtained in step S1, a three-dimensional model of the balance is generated using an assembly modeling method;
[0033] Furthermore, during the modeling process in step S2, it is checked whether the maximum size of the beam cross section exceeds the diameter of the balance, whether the components are in good contact, and whether there is any overlap or offset interference between the models, so as to ensure that the size and position of the balance components are in a coordinated state.
[0034] S3. Taking into account the strength of the balance structure, the maximum stress at the strain gauge bonding location, and the mutual interference between components, an optimization model is established;
[0035] Furthermore, in step S3, an optimization model is established with the optimization objectives of ensuring that the design strain output of each component of the balance is reasonable and that the strength at the key position of the balance meets the material yield strength requirements.
[0036] Furthermore, considering the strength of the balance structure, the maximum stress at the strain gauge mounting location, and the mutual interference between components, an objective function is established. The objective function is a constrained multi-objective optimization form, as shown in the following formula:
[0037]
[0038] in, The objective function to be optimized is abstracted from the following: reasonable strain output of each component of the balance, small interference between each component of the balance, and the strength of the key position of the balance meets the material yield strength requirements. These are variables that need to be optimized, such as the dimensions and structural parameters of the balance. and Variables Lower and upper bound constraints; For variables Linear equality constraints; For variables Linear inequality constraints.
[0039] S4. Based on the DOE method, conduct experimental design and parameter analysis on the optimized model obtained in step S3. Combined with correlation analysis, obtain the degree of influence of the input parameters of the balance on the measurement performance of the balance, and extract the key parameters of the balance.
[0040] Furthermore, the DOE method in step S4 is implemented based on the orthogonal array method, the Latin hypercube method, and the optimized Latin hypercube method.
[0041] Furthermore, in experimental design, all the input parameters under study are called factors, the numerical value defined for each factor is called a level, and the output parameter obtained from the analysis is called the response. The experimental design method is based on an integrated implementation of the following three methods, specifically including:
[0042] (1) Orthogonal array method: The orthogonal array method, also known as the partial factorial method, ensures the orthogonality between factors, and the number of trials is approximately equal to the number of factors selected. The advantage is that it covers the entire space under full factors and can obtain greater accuracy with fewer trials. The disadvantage is that the interaction effect between factors is relatively small.
[0043] (2) Latin hypercube method: The Latin method uniformly samples within a spatial distribution and randomly combines a specified number of points at a horizontal level to obtain the interaction effects between factors. If the number of factors is m, then the minimum number of sampling points N to ensure accuracy is:
[0044]
[0045] (3) Optimized Latin hypercube method: Since the Latin method randomly samples within the domain, it has poor control over space and corner points. The optimized Latin method optimizes the sampling based on the Latin method, so that the sampling points are distributed as evenly as possible in space.
[0046] S5. Based on the three-dimensional model of the balance under different parameters obtained in step S2 and the key parameters of the balance obtained in step S4, the parameterized design language command flow of ANSYS is used to perform parameterized automatic simulation on the optimized model in step S3, generating sample data of the key parameters of the balance and the optimization target values corresponding to the simulation analysis results, forming a key parameter-optimization target sample pair of the balance.
[0047] Furthermore, the parameterized automatic simulation process in step S5 is as follows: Based on the batch processing method, the APDL script program is run to start ANSYS and the Mechanical APDL module. By batch setting the material properties, load positions and sizes, mesh size and position, and fixed end constraints in the script language, the automatic simulation process of the model driven by different parameters is realized, and sample data of the optimized target values corresponding to the parameter values and simulation analysis results are generated.
[0048] S6. Based on the intelligent optimization algorithm, optimize the key parameters of the balance obtained in step S5 - the key parameters of the target sample pair of the balance, and extract the optimal solution of the target value.
[0049] Furthermore, the intelligent optimization algorithm in step S6 is an improved genetic algorithm NSGA-II, and the key parameters of the optimized balance include the thickness of the measuring element. Support beam length Width of the support beam Width of the slot in the axial force element of the balance Distance from the center of the balance to the support beam .
[0050] Furthermore, the key size parameter values are optimized based on the improved genetic algorithm (NSGA-II), and the specific implementation process is as follows:
[0051] S6.1. Initializing the Population: First, randomly generate an initial population containing multiple individuals. Each individual represents a potential solution;
[0052] S6.2. Calculate Fitness: For each individual, calculate its fitness in the objective function space. The calculation of fitness depends on the specific problem; generally, evaluation indicators (such as distance, superiority, etc.) are used to represent the degree of superiority or inferiority of the individual.
[0053] S6.3. Non-dominated ordination: NSGA-II uses non-dominated ordination techniques to divide individuals in the population into multiple fronts. This ordination process determines the non-dominated level of each individual to identify which individuals occupy different positions on the Pareto front. Individuals at higher levels are not dominated by individuals at lower levels;
[0054] S6.4. Crowding Distance Calculation: To maintain the diversity of the Pareto front, NSGA-II calculates the crowding distance to measure the density of individuals in the target space. A larger crowding distance indicates greater distance between individuals, thus helping to maintain a uniform distribution on the Pareto front.
[0055] S6.5. Selection Operation: NSGA-II uses a binary tournament as the selection operation. In each generation, two individuals are first randomly selected from the current population using a binary tournament selection operation, and then the individual with the higher non-dominance level is selected. This selection process is performed by comparing the individuals' non-dominance level and crowding distance;
[0056] S6.6. Crossover / Mutation Operations: Selected individuals are crossovered and mutated to generate a parent population. The parent and offspring generations are then merged into a larger candidate population.
[0057] S6.7. Iteration: Re-sort the candidate populations for non-dominated order, calculate crowding distance, select, cross over, and mutate to generate new offspring populations. Iterate until the stopping condition is met (such as reaching the maximum number of iterations or converging to a satisfactory Pareto front solution).
[0058] S7. Repeat steps S5 and S6 to obtain the key parameters of the balance through multiple simulations - optimize the target sample pair dataset;
[0059] Furthermore, through multiple simulations, sample pairs of key dimensional parameters and performance indicators of the balance were obtained. These sample pairs mainly included: optimizing the thickness of the measuring elements. Support beam length Width of the support beam Width of the slot in the axial force element of the balance Distance from the center of the balance to the support beam Key parameter values and corresponding simulation output results: strength of the balance structure, maximum stress at the strain gauge bonding location, and mutual interference between components;
[0060] S8. Based on the dataset obtained in step S7, train a machine learning regression model to obtain an intelligent mathematical proxy model for the structural optimization design of the wind tunnel balance.
[0061] Furthermore, the machine learning regression model in step S8 includes one of support vector machines, neural networks, and Bayesian regression models.
[0062] Furthermore, the three model forms are as follows:
[0063] 1. The method for optimizing the target value prediction of a balance scale based on support vector machines is as follows:
[0064] The support vector regression model is represented as:
[0065]
[0066] in, For Gaussian kernel function, As the first weight value, For deviation, The hyperplane is defined; the loss function is defined according to backpropagation theory. ;
[0067] loss function Represented as:
[0068]
[0069] in, This is the actual output;
[0070] Introducing the first slack variable Second relaxation variable The support vector regression model is solved using the Lagrange function method.
[0071] The support vector regression model is represented as:
[0072]
[0073]
[0074] in, As a penalty factor, The optimal parameters are to be determined. As the second weight value, This refers to the deviation range;
[0075] 2. The method for predicting the target value of balance design optimization based on Bayesian regression is as follows:
[0076] Bayesian regression learning machines can automatically and recursively derive output weights without requiring cross-validation to determine regularization parameters, while also avoiding the ill-conditioned solution problem of pseudo-inverse methods. This method assumes that the learning errors are independent and follow a zero-mean Gaussian distribution, i.e., the likelihood function of the training data follows the following Gaussian distribution:
[0077]
[0078] The prior distribution of the network output weights is set as follows:
[0079]
[0080] The corresponding posterior distribution is also a Gaussian distribution, with its mean and variance matrices being respectively... and :
[0081]
[0082] Determined by the method of approximate evidence and The value of is calculated using the following formula:
[0083]
[0084]
[0085]
[0086] in, for The eigenvalues. First, initialize the parameters. and , and then use the initialized and calculate and , and then use the estimated and Recalculate and The value is calculated repeatedly until the algorithm converges, and the model parameters are finally obtained. When a new balance design is used to optimize the prediction of the target value, it can be directly input into the model and the trained parameters can be used for calculation.
[0087] 3. The method for predicting the target value of a balance scale based on convolutional neural networks is as follows:
[0088] Traditional regression methods (such as linear regression and decision trees) typically require extensive feature engineering to extract effective features from the data. Convolutional neural networks (CNNs), on the other hand, can automatically learn and extract features from raw data, simplifying the modeling process. The convolutional layer, the core structure of DCNNs, scans the input data along the time axis using convolutional kernels and performs convolution operations on the input, as shown in the following equation.
[0089]
[0090] Among them, L n m and Z n m X represents the weights and biases of the nth convolutional kernel in the mth layer; m (p) represents the p-th local region in the m-th layer; y n m+1 (p) is the input of the p-th neuron in the result of the n-th convolution kernel operation in the (m+1)-th layer.
[0091] After predicting the target value of balance design optimization using the above method, the performance of each balance optimization design prediction model is quantitatively evaluated in the performance evaluation process using absolute error (root mean square error, RMSE), normalized root mean square error (NRMSE), mean absolute percentage error (MAPE), and mean absolute deviation (MAD).
[0092]
[0093]
[0094]
[0095]
[0096] In the above formula, For the first The predicted output for each test sample For the first The target output of each test sample The average value of the target output vector. This represents the total number of test samples.
[0097] Example 2:
[0098] A wind tunnel test balance structure optimization design system includes a processor, a memory, and a computer program stored in the memory and run on the processor. When the computer program runs, it implements the steps of the wind tunnel test balance structure optimization design method as described in Example 1.
[0099] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0100] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
Claims
1. A method for optimizing the design of a wind tunnel test balance structure, characterized in that, Includes the following steps: S1. Determine the input parameters of the balance according to its design shape and requirements; S2. Based on the input parameters of the balance obtained in step S1, a three-dimensional model of the balance is generated using an assembly modeling method; S3. Taking into account the strength of the balance structure, the maximum stress at the strain gauge bonding location, and the mutual interference between components, an optimization model is established; S4. Based on the DOE method, conduct experimental design and parameter analysis on the optimized model obtained in step S3. Combined with correlation analysis, obtain the degree of influence of the input parameters of the balance on the measurement performance of the balance, and extract the key parameters of the balance. S5. Based on the three-dimensional model of the balance under different parameters obtained in step S2 and the key parameters of the balance obtained in step S4, the parameterized design language command flow of ANSYS is used to perform parameterized automatic simulation on the optimized model in step S3, generating sample data of the key parameters of the balance and the optimization target values corresponding to the simulation analysis results, forming a key parameter-optimization target sample pair of the balance. S6. Based on the intelligent optimization algorithm, optimize the key parameters of the balance obtained in step S5 - the key parameters of the target sample pair of the balance, and extract the optimal solution of the target value. S7. Repeat steps S5 and S6 to obtain the key parameters of the balance through multiple simulations - optimize the target sample pair dataset; S8. Based on the dataset obtained in step S7, train a machine learning regression model to obtain an intelligent mathematical proxy model for the structural optimization design of the wind tunnel balance.
2. The method for optimizing the design of a wind tunnel test balance structure according to claim 1, characterized in that, In step S1, based on the application scenario, measurement range requirements, and structural strength, stiffness, and sensitivity requirements of the balance, the input parameters of the balance are determined, including the diameter, length, width, height, and position dimensions of the measuring element, the thickness, gap dimensions, number of support plates, position dimensions, and slot angle. The input parameters of the balance with the axial force measuring unit also include the form of the axial force element support plate, and the specific parameters and dimensions of the axial force element support plate.
3. The method for optimizing the design of a wind tunnel test balance structure according to claim 2, characterized in that, In step S2, during the modeling process, it is necessary to check whether the maximum size of the beam cross section exceeds the diameter of the balance, whether the components are in good contact, and whether there is any overlap or offset interference between the models, so as to ensure that the size and position of the balance components are in a coordinated state.
4. The method for optimizing the design of a wind tunnel test balance structure according to claim 3, characterized in that, In step S3, an optimization model is established with the optimization objectives of ensuring that the strain output of each component of the balance is reasonable and that the strength at the key position of the balance meets the material yield strength requirements.
5. The method for optimizing the design of a wind tunnel test balance structure according to claim 4, characterized in that, The DOE method in step S4 is implemented based on the orthogonal array method, the Latin hypercube method, and the optimized Latin hypercube method.
6. The method for optimizing the design of a wind tunnel test balance structure according to claim 5, characterized in that, The parameterized automatic simulation process in step S5 is as follows: The APDL script program is run based on the batch processing method to start ANSYS and the Mechanical APDL module. By batch setting the material properties, load positions and sizes, mesh size and position, and fixed end constraints in the script language, the automatic simulation process of the model driven by different parameters is realized, and sample data of the optimized target values corresponding to the parameter values and simulation analysis results are generated.
7. The method for optimizing the design of a wind tunnel test balance structure according to claim 6, characterized in that, The intelligent optimization algorithm in step S6 is an improved genetic algorithm NSGA-II. The key parameters of the optimized balance include the thickness of the measuring element. Support beam length Width of the support beam Width of the slot in the axial force element of the balance Distance from the center of the balance to the support beam .
8. The method for optimizing the design of a wind tunnel test balance structure according to claim 7, characterized in that, The machine learning regression model in step S8 includes one of the following: support vector machine, neural network, and Bayesian regression model.
9. A wind tunnel test balance structural optimization design system, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed, implements the steps of the wind tunnel test balance structure optimization design method as described in any one of claims 1-8.