A wing surface field inversion method based on distributed pressure measurement

Abstract

The application discloses a kind of wing surface field inversion methods based on distributed pressure measurement, belong to metrological testing technical field.The application is arranged pressure measurement array and verification point array in the test area of wing surface, obtains abundant inversion data by fluid simulation, and constructs wing surface field inversion database, in combination with deep learning and interpolation method, realize the inversion of limited pressure measuring point to spatial position point cluster pressure and spatial field information, can reduce the acquisition time of pressure field, and limit condition reduces, and measurement accuracy is greatly improved.The pressure data of the verification point arranged in the measurement area is used to verify the single-point pressure prediction capability of the field inversion model, and the full-field prediction accuracy of the field inversion model is verified by the measured field measurement information, which more truly represents the application scenario of the model and improves the accuracy of the field inversion model.

Description

Technical Field

[0001] This invention relates to a method for inverting the surface field of an airfoil based on distributed pressure measurement, belonging to the field of metrology and testing technology. Background Technology

[0002] High-precision measurement of the pressure field on the wing surface is crucial for decoupling the structural and aerodynamic loads on the wing surface, and is fundamental to improving the overall performance of the wing. Currently, the main methods for measuring the pressure field on the wing surface are flexible pressure gauges and pressure-sensitive paint. However, due to their inherent technical limitations, these two techniques cannot be widely adopted. For example, pressure-sensitive paint cannot be tested in environments with high light intensity and often only yields satisfactory results in high-speed wind tunnels. Flexible pressure gauges, on the other hand, can alter the structure and flow field characteristics of the test area to some extent.

[0003] While distributed pressure sensors can obtain accurate pressure distribution, they cannot obtain complete pressure field information of the measured area, which limits the further shortening of the development cycle of the wing system and the significant improvement of its performance. Summary of the Invention

[0004] The purpose of this invention is to provide a method for inverting the wing surface field based on distributed pressure measurement. By constructing a wing surface field inversion database, and using a combination of deep learning and interpolation methods, the method can realize the inversion of pressure and spatial field information from a limited number of pressure measurement points to a cluster of spatial locations. Under the premise of satisfying the accuracy and precision of wing surface pressure measurement, the method can improve the dimensionality of spatial pressure measurement and enhance the measurement accuracy of wing surface pressure.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The present invention discloses a method for inverting the surface field of an airfoil based on distributed pressure measurement, comprising the following steps:

[0007] Step 1: Create a fluid simulation model based on wind tunnel tests of the wing. Verify the accuracy of the fluid simulation model using experimental data, and obtain simulation data under different Mach numbers and angles of attack. Analyze the simulation data to identify areas where the pressure gradient changes drastically in the measurement area. Combine this with the measurement points arranged in equal proportions along the chord and span of the test area on the wing surface to determine the placement of pressure sensors on the wing surface, and divide the area into a pressure measurement array (p...). M1 ,p M2 ,…,p Mn ) and pressure verification array (p V1 ,p V2 ,…,p Vm ).

[0008] Step 2: Arrange a pressure measurement array (p) in the test area on the wing surface. M1 ,p M2 ,…,p Mn ) and verification point array (p V1 ,p V2 ,…,p Vm Simultaneously, by considering the sensor wiring method, the impact of wiring on the accuracy of wing surface pressure measurement is reduced. Additionally, pressure-sensitive paint is applied to the test area, and a matching camera measurement device is used to measure the pressure field p on the wing surface under wind tunnel test conditions. f Measurements are performed. The test area is the region where the pressure measurement points and verification points are located.

[0009] Step 3: Based on the fluid simulation model constructed in Step 1, according to the pressure measurement array {(x M1 ,y M1 ,z M1 ),(x M2 ,y M2 ,z M2 ),…,(x Mn ,y Mn ,z Mn )} and stress verification array {(x V1 ,y V1 ,z V1 ),(x V2 ,y V2 ,z V2 ),…,(x Vm ,y Vm ,z Vm The spatial coordinates of the pressure information are extracted to obtain the pressure measurement array on the wing surface under different wind tunnel test conditions, and to verify the evolution of the pressure signal at different measuring points in the array over time, i.e., {(t} i ,p M1 ),(t i ,p M2 ),…,(t i ,p Mn )} and {(t i ,p V1 ),(t i ,p V2 ),…,(t i ,p Vm Simultaneously, pressure field data of the complex wing surface at different times in the test area are extracted, i.e., (t i ,p f ).

[0010] Step 4: Based on the pressure measurement array obtained from simulation {(t i ,p M1 ),(t i,p M2 ),…,(t i ,p Mn Pressure field data (t) at different locations on the wing geometry i ,p f ), and construct a dataset for deep learning training for pressure field inversion.

[0011] Step 5: Based on the deep learning training dataset constructed in Step 4, let t i The pressure field distribution data at different spatial locations on the wing surface at any given time are used as the regression object. The input information of the field inversion model is constructed by combining the R, G, and B multi-channel data of the image. i The input data for training the time-instance model includes initial parameter information such as the pressure measurement array information, the Mach number in matrix form, and the angle of attack information at that time. A field inversion model based on distributed pressure measurement is trained using a fully connected network. This field inversion model is used to predict the pressure information at key points on the wing surface, thereby refining the wing surface pressure information.

[0012] By analyzing t i The pressure information at the location of the pressure field verification array obtained from the time-inversion is extracted and compared with {(t)}. i ,p V1 ),(t i ,p V2 ),…,(t i ,p Vm By comparing the results, the field inversion model is trained with the goal of minimizing the error.

[0013] Step 6: To further verify the effectiveness of the inversion model trained in Step 5 and the accuracy of the field measurements, the field inversion model trained in Step 5 is experimentally verified. Based on the experimental data obtained from the wind tunnel test, and using the pressure data of the measurement array obtained from the test as input information, t i The pressure field information of the measurement area is calculated from the data of the pressure measurement array at any time. Based on the spatial coordinate position of the pressure verification array, the pressure information of each measuring point is extracted and compared with the measured pressure information to verify the accuracy of the single-point prediction of the trained field inversion model.

[0014] In addition, the distribution information of spatial pressure measured by pressure-sensitive paint is introduced into the criterion conditions. The pressure field information obtained by the field inversion model is compared with the distribution information of spatial pressure field measured by pressure-sensitive paint. Through correlation calculation, the consistency between the field information predicted by the model and the field information measured by the experiment is analyzed.

[0015] The accuracy and usability of the trained field inversion model were further verified by using single-point verification and field verification camera methods.

[0016] Step 7: If the relative error between the predicted and measured pressure values ​​at each point on the wing surface pressure verification array is less than φ, then the single-point prediction accuracy of the field inversion model is deemed to meet the requirements. If the correlation between the pressure distribution in the measurement area predicted by the field inversion model and the field distribution measured by the pressure-sensitive paint is greater than σ, then the field prediction capability is deemed to meet the measurement requirements. The field inversion model is deemed to meet the measurement requirements if and only if both conditions are met simultaneously. Otherwise, return to step 6 to refine the pressure field inversion model until the requirements are met.

[0017] Step 8: Predict the pressure information of key points on the wing surface based on the field inversion model trained in Step 7, realize field inversion based on distributed pressure measurement, and improve the dimensionality of spatial pressure measurement and improve the measurement accuracy of wing surface pressure while meeting the accuracy and precision of wing surface pressure measurement.

[0018] Beneficial effects:

[0019] 1. This invention discloses a method for inverting the wing surface field based on distributed pressure measurement. A pressure measurement array and a verification point array are arranged in the test area on the wing surface. Simultaneously, the sensor wiring method is combined to reduce the impact of wiring on the accuracy of wing surface pressure measurement. Pressure-sensitive paint is applied to the test area, and a matching camera measurement device is used to measure the pressure field p on the wing surface under wind tunnel test conditions. f Measurements are taken; based on this, by constructing a wing surface field inversion database, a combination of deep learning and interpolation methods is used to realize the inversion of pressure and spatial field information from a limited number of pressure measurement points to a cluster of spatial locations. Compared with fluid simulation technology, the time for acquiring the pressure field is greatly reduced; compared with test methods such as pressure-sensitive paint and flexible pressure measuring tape, the limiting conditions are reduced and the measurement accuracy is greatly improved.

[0020] 2. The present invention discloses a method for inverting the surface field of an airfoil based on distributed pressure measurement. The input parameters of the fully connected neural network model are used to construct a comprehensive spatial pressure measurement point information and measurement environment parameters, thereby realizing the fusion of spatiotemporal pressure sequence and initial environmental parameter scalar values, which is conducive to the implementation of end-to-end fully connected neural networks.

[0021] 3. The present invention discloses a method for inverting the wing surface field based on distributed pressure measurement. The wing surface pressure field inversion dataset used for model training is mainly obtained through fluid simulation, while the dataset used for model practicality is mainly obtained through experimental means, thereby improving the transfer and generalization ability of the trained field inversion model.

[0022] 4. The present invention discloses a method for inverting the wing surface field based on distributed pressure measurement. The method verifies the single-point pressure prediction capability of the field inversion model by using pressure data from verification points arranged in the measurement area, and verifies the full-field prediction accuracy of the field inversion model by using measured field measurement information. This method more realistically reflects the application scenario of the model and improves the accuracy of the field inversion model. Attached Figure Description

[0023] Figure 1 A method for determining the array of pressure measurement points and verification points on the wing surface;

[0024] Figure 2 This is a principle for inverting the wing surface field based on distributed pressure measurement;

[0025] Figure 3 A method for refining pressure measurement points based on deep convolutional neural networks;

[0026] Figure 4 Pressure field inversion system based on distributed pressure measurement. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] The following is in conjunction with the appendix Figures 1-4 The present invention will be further described in detail with reference to the embodiments. These embodiments are based on the technical solutions of the present invention and provide specific implementation methods, but the scope of protection of the present invention is not limited to the following embodiments.

[0029] like Figure 1 As shown in the figure, the specific implementation steps of the wing surface field inversion method based on distributed pressure measurement disclosed in this embodiment are as follows:

[0030] Step 1: Create a corresponding fluid simulation model based on wind tunnel tests of the airfoil, complete mesh independence verification, and verify the accuracy of the fluid simulation model using historical test data. Simulation data under different Mach numbers, angles of attack, etc., is obtained to analyze the evolution characteristics of the pressure field on the airfoil surface, analyze the pressure gradient changes in the pressure field, and identify areas with significant pressure variations. Simultaneously, combining the conventional method of arranging measurement points proportionally along the chord and span in the airfoil test area, the placement positions of the pressure sensors on the airfoil surface (e.g., ...) are jointly determined. Figure 1 (As shown). Among the determined pressure measuring point locations, a portion of the measuring points are selected as a pressure measuring array (p). M1 ,p M2 ,…,p Mn The other part is the stress verification array (p). V1 ,p V2 ,…,p Vm ).

[0031] Step 2: Arrange a pressure measurement array (p) in the test area on the wing surface. M1 ,p M2 ,…,p Mn ) and verification array (p V1 ,p V2 ,…,p Vm Meanwhile, the wiring positions of the sensors were optimized to reduce the impact of wiring on the measurement accuracy of wing surface pressure. After the pressure sensor passes through the signal conditioning device, the pressure signal is acquired by the data acquisition system and finally stored in text form on the host computer.

[0032] Pressure-sensitive paint was applied to the test area, and the pressure field p on the wing surface under wind tunnel test conditions was measured using a matching camera measurement device. f Measurements were performed. The test area was the region where the pressure measurement points and verification points were located. The pressure field information of the measurement area was obtained by analyzing and extracting the field information measured by the camera.

[0033] Step 3: Based on the fluid simulation model constructed in Step 1, according to the pressure measurement array {(x M1 ,y M1 ,z M1 ),(x M2 ,y M2 ,z M2 ),…,(x Mn ,y Mn ,z Mn )} and stress verification array {(x V1 ,y V1 ,z V1 ),(x V2 ,y V2 ,z V2 ),…,(x Vm ,y Vm ,z Vm The spatial coordinates of the pressure information are extracted to obtain the pressure measurement array on the wing surface under different wind tunnel test conditions, and to verify the evolution of the pressure signal at different measuring points in the array over time, i.e., {(t} i ,p M1 ),(t i ,p M2 ),…,(t i ,p Mn )} and {(t i ,p V1 ),(t i ,p V2 ),…,(t i ,p Vm Simultaneously, pressure field data of the complex wing surface at different times in the test area are extracted, i.e., (ti ,p f Wind tunnel test conditions include different Mach numbers and angles of attack.

[0034] Step 4: Based on the pressure measurement array obtained from simulation {(t i ,p M1 ),(t i ,p M2 ),…,(t i ,p Mn Pressure field data (t) at different locations on the wing geometry i ,p f ), and construct a dataset for deep learning training for pressure field inversion.

[0035] Step 5: Based on the deep learning training dataset constructed in Step 4, let t i The pressure field distribution data at different spatial locations on the wing surface at any given time is used as the regression object. The input information of the field inversion model is constructed by combining multiple channels such as R, G, and B from the image. i The input data for training the time-instance model mainly includes initial parameter information such as the pressure measurement array information, the Mach number in matrix form, and the angle of attack information at that time. A field inversion model based on distributed pressure measurement is trained using a fully connected network. This field inversion model is then used to predict the pressure information at key points on the wing surface, thereby refining the wing surface pressure information.

[0036] By analyzing t i The pressure information at the location of the pressure field verification array obtained from the time-inversion is extracted and compared with {(t)}. i ,p V1 ),(t i ,p V2 ),…,(t i ,p Vm By comparing the results, the field inversion model is trained with the goal of minimizing the error.

[0037] Step 6: To further verify the effectiveness of the model and the accuracy of the field measurements, the field inversion model trained in Step 5 is experimentally verified. Based on the experimental data obtained from the wind tunnel test, and using the pressure data of the measurement array obtained from the test as input information, t i The pressure field information of the measurement area can be calculated from the data of the pressure measurement array at any time. Based on the spatial coordinate position of the pressure verification array, the pressure information of each measuring point is extracted and compared with the measured pressure information to verify the accuracy of the model's single-point prediction.

[0038] In addition, the distribution information of spatial pressure measured by pressure-sensitive paint is introduced into the criterion conditions. The pressure field information obtained by the field inversion model is compared with the distribution information of spatial pressure field measured by pressure-sensitive paint. Through correlation calculation, the consistency between the field information predicted by the model and the field information measured by the experiment is analyzed.

[0039] By employing single-point validation and field validation methods, the accuracy and usability of the model are further clarified.

[0040] Step 7: If the relative error between the predicted and measured pressure values ​​at each point on the wing surface pressure verification array is less than φ, then the single-point prediction accuracy of the model is considered to meet the requirements. If the correlation between the pressure distribution in the measurement area predicted by the field inversion model and the field distribution measured by the pressure-sensitive paint is greater than σ, then the field prediction capability is considered to meet the measurement requirements. The field inversion model is considered to meet the measurement requirements if and only if both conditions are met simultaneously. Otherwise, the pressure field inversion model needs to be further improved until it meets the requirements (e.g., ...). Figure 2 (As shown).

[0041] Step 8: Predict the fine pressure field information of the measurement area on the wing surface based on the field inversion model trained in Step 7 (e.g., Figure 3 In wind tunnel testing, pressure sensors from the wing surface pressure measurement array and pressure verification array are connected to a signal conditioner. Pressure data is acquired via a host computer. After data processing, combined with the real spatial location information of each measuring point, an accurate prediction of the pressure field information for the entire measurement area is achieved based on the pressure field inversion module (e.g., ...). Figure 4 As shown in the figure, this provides technical support for the analysis of the aerodynamic characteristics of the wing surface.

[0042] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for inverting the surface field of an airfoil based on distributed pressure measurement, characterized in that: Includes the following steps, Step 1: Create a fluid simulation model based on wind tunnel tests of the wing. Verify the accuracy of the fluid simulation model using experimental data, and obtain simulation data under different Mach numbers and angles of attack. Analyze the simulation data to identify areas where the pressure gradient changes drastically in the measurement area. Combine this with the measurement points arranged in equal proportions along the chord and span of the test area on the wing surface to determine the placement of pressure sensors on the wing surface, and divide the area into a pressure measurement array (p...). M1 ,p M2 ,…,p Mn ) and pressure verification array (p V1 ,p V2 ,…,p Vm ); Step 2: Arrange a pressure measurement array (p) in the test area on the wing surface. M1 ,p M2 ,…,p Mn ) and verification point array (p V1 ,p V2 ,…,p Vm Simultaneously, by considering the sensor wiring method, the impact of wiring on the accuracy of wing surface pressure measurement is reduced; at the same time, pressure-sensitive paint is applied to the test area, and a matching camera measurement device is used to measure the pressure field p on the wing surface under wind tunnel test conditions. f Measurements are performed; the test area is the region where the pressure measurement points and verification points are located. Step 3: Based on the fluid simulation model constructed in Step 1, according to the pressure measurement array {(x M1 ,y M1 ,z M1 ),(x M2 ,y M2 ,z M2 ),…,(x Mn ,y Mn ,z Mn )} and stress verification array {(x V1 ,y V1 ,z V1 ),(x V2 ,y V2 ,z V2 ),…,(x Vm ,y Vm ,z Vm The spatial coordinates of the pressure information are extracted to obtain the pressure measurement array on the wing surface under different wind tunnel test conditions, and to verify the evolution of the pressure signal at different measuring points in the array over time, i.e., {(t} i ,p M1 ),(t i ,p M2 ),…,(t i ,p Mn )} and {(t i ,p V1 ),(t i ,p V2 ),…,(t i ,p Vm Simultaneously, pressure field data of the complex wing surface at different times in the test area are extracted, i.e., (t i ,p f ); Step 4: Based on the pressure measurement array obtained from simulation {(t i ,p M1 ),(t i ,p M2 ),…,(t i ,p Mn Pressure field data (t) at different locations on the wing geometry i ,p f ), construct a dataset for deep learning training for pressure field inversion; Step 5: Based on the deep learning training dataset constructed in Step 4, let t i The pressure field distribution data at different spatial locations on the wing surface at any given time are used as the regression object. The input information of the field inversion model is constructed by combining the R, G, and B multi-channel data of the image. i The input data for training the time-based model includes the pressure measurement array information, the Ma number in matrix form, and the angle of attack information at that time as initial parameter information. A field inversion model based on distributed pressure measurement is trained using a fully connected network. The field inversion model is used to predict the pressure information of key points on the wing surface, thereby refining the pressure information on the wing surface. By analyzing t i The pressure information at the location of the pressure field verification array obtained from the time-inversion is extracted and compared with {(t)}. i ,p V1 ),(t i ,p V2 ),…,(t i ,p Vm By comparing the results, and aiming to minimize the error, the field inversion model is trained. Step 6: To further verify the effectiveness of the inversion model trained in Step 5 and the accuracy of the field measurements, the field inversion model trained in Step 5 is experimentally verified. Based on the experimental data obtained from the wind tunnel test, and using the pressure data of the measurement array obtained from the test as input information, the model is calculated using t... i The pressure field information of the measurement area is calculated from the data of the pressure measurement array at any time. Based on the spatial coordinate position of the pressure verification array, the pressure information of each measuring point is extracted and compared with the measured pressure information to verify the accuracy of the single-point prediction of the trained field inversion model. In addition, the distribution information of spatial pressure measured by pressure-sensitive paint is introduced into the criterion conditions. The pressure field information obtained by the field inversion model is compared with the distribution information of spatial pressure field measured by pressure-sensitive paint. Through correlation calculation, the consistency between the field information predicted by the model and the field information measured by the experiment is analyzed. Step 7: If the relative error between the predicted and measured pressure values ​​at each point of the wing surface pressure verification array is less than φ, then the single-point prediction accuracy of the field inversion model is deemed to meet the requirements. If the correlation between the pressure distribution of the measurement area predicted by the field inversion model and the field distribution measured by the pressure-sensitive paint is greater than σ, then the field prediction capability is deemed to meet the measurement requirements. The field inversion model is deemed to meet the measurement requirements if and only if both conditions are met simultaneously. Otherwise, return to step 6 to improve the pressure field inversion model until the requirements are met. Step 8: Predict the pressure information of key points on the wing surface based on the field inversion model trained in Step 7, realize field inversion based on distributed pressure measurement, and improve the dimensionality of spatial pressure measurement and improve the measurement accuracy of wing surface pressure while meeting the accuracy and precision of wing surface pressure measurement.

2. The wing surface field inversion method based on distributed pressure measurement as described in claim 1, characterized in that: In step 6, the accuracy of the trained field inversion model is further verified by using single-point verification and field verification camera methods.

Images