A transition model debugging method for accurate prediction of compressor flow

Abstract

The application provides a transition model debugging method for accurate prediction of compressor flow, and belongs to the technical field of aero-engines, and specifically comprises the following steps: obtaining key flow physical quantities of a typical compressor cascade under different inflow conditions; determining a parameter set to be debugged and a variation range of each parameter to be debugged in the parameter set, and generating an initial sample set of the parameter set to be debugged; taking the obtained key flow physical quantities as a calibration target, and iteratively updating the parameter set to be debugged based on the initial sample set and an iterative calibration framework based on a DIOM; when a cycle reaches a maximum number of iterations or a prediction accuracy reaches a convergence condition, an optimal estimated parameter set is obtained, and an average value and a standard deviation of the optimal estimated parameter set are output as a parameter debugging result. The application provides a systematic parameter debugging method for improving the numerical simulation accuracy of a compressor, which is simple and effective.

Description

Technical Field

[0001] This application relates to the field of aero-engine technology, and in particular to a method for debugging a transition model for accurate prediction of compressor flow. Background Technology

[0002] The development of advanced aero-engines in the future places higher demands on the aerodynamic performance of compressors. With increasing flight altitude, the transition process within the boundary layer on the blade surface under real flight conditions has a decisive impact on the compressor's loss characteristics, load capacity, and stability margin. Accurately predicting the transition location and the length of the transition region is crucial for the refined aerodynamic design of compressors. Currently, the design quality of high-performance aero-engine compressors highly depends on the modeling accuracy of the Reynolds-averaged Navier-Stokes (RANS) transition model. However, existing transition models are mostly calibrated based on flat plate flow experimental data, resulting in poor prediction accuracy for flow transition phenomena in real compressors. Studies have shown that when using the standard γ transition model to numerically simulate compressor blades, there are problems such as overestimating the size of the laminar separation bubble and inaccurate prediction of blade performance. Therefore, there is an urgent need to develop a RANS transition model calibration method for accurate compressor flow prediction, providing a reliable tool for the refined design of advanced aero-engine compressors. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method for debugging a transition model for accurate prediction of compressor flow, which at least partially solves the defects of the γ transition model in compressor flow prediction in the prior art.

[0004] This application provides a method for debugging a transition model for accurate prediction of compressor flow, including:

[0005] Step 1: Obtain key flow physical quantities of a typical compressor blade under different incoming flow conditions; Step 2: Determine the set of parameters to be debugged and the range of variation for each parameter in the set, and generate an initial sample set for the set of parameters to be debugged; Step 3: Using the key flow physical quantities obtained as calibration targets, iteratively update the parameter set to be debugged based on the initial sample set and the DIOM-based iterative calibration framework; Step 4: When the loop reaches the maximum number of iterations or the prediction accuracy reaches the convergence condition, obtain the optimal estimated parameter set, and output the mean and standard deviation of the optimal estimated parameter set as the parameter adjustment result.

[0006] According to a specific implementation of an embodiment of this application, obtaining key flow physical quantities of a typical compressor blade under different incoming flow conditions includes: Select a typical compressor blade cascade and obtain the key flow physical quantities of the typical compressor blade cascade under different incoming flow conditions through experiments or high-fidelity simulation methods.

[0007] According to a specific implementation of an embodiment of this application, the expression for the key flow physical quantity is: , in, As a key flow physical quantity, The surface static pressure rise coefficient, The coefficient of surface friction, For different flow directions and positions, This is the total pressure loss coefficient at the export end. This is to account for errors in experimental or high-fidelity simulation data.

[0008] According to a specific implementation of an embodiment of this application, the expression of the parameter set to be debugged is: , in, For the parameter set to be debugged, C TU1 To control the parameters of the natural transition critical Reynolds number, C TU2 To control the parameters of the bypass transition critical Reynolds number, C TU3 To control the parameters of the bypass transition critical Reynolds number, the intensity of free-flow turbulence, and the rate of change of the pressure gradient, C PG1 The parameter C is used to control the value of the critical Reynolds number at transition under the compressive gradient. PG2 The parameter C is used to control the value of the critical Reynolds number at transition under the inverse pressure gradient. PG3 The parameter L is used to control the critical Reynolds number of the transition zone. T The length scale is the inlet turbulence length scale.

[0009] According to a specific implementation of an embodiment of this application, the initial sample set for generating the parameter set to be debugged includes: The Latin hypercube sampling algorithm is used to generate an initial sample set of parameters to be debugged.

[0010] According to a specific implementation of an embodiment of this application, the iterative update of the parameter set to be debugged based on the initial sample set and the DIOM-based iterative calibration framework includes: For the sample set of the parameter set to be debugged in the i-th iteration, each set of parameter sets to be debugged in the sample set is input into the RANS solver to simulate different incoming flow conditions of the selected typical compressor blade cascade and calculate the predicted values ​​of key flow physical quantities for each set of parameter sets to be debugged. Based on the predicted values ​​of key flow physical quantities in the i-th iteration and the key flow physical quantities obtained in step 1, the prediction error and Kalman gain matrix are calculated. The sample set is updated based on the prediction error and the Kalman gain matrix.

[0011] According to a specific implementation of an embodiment of this application, the expression for the predicted value of the key flow physical quantity is: , in, Let R(·) be the predicted value of the key flow physical quantity for the j-th sample in the i-th iteration, and R(·) be the RANS solver. Let ε be the j-th sample in the sample set during the i-th iteration. (j) This represents the uncertainty error in RANS simulation.

[0012] According to a specific implementation of an embodiment of this application, the expression for the prediction error is: , in, This represents the prediction error; The expression for the Kalman gain matrix is: , in, Here is the Kalman gain matrix. for and The covariance matrix between them for and The covariance matrix, This represents the average value of the high-precision data error. This is the average value of the RANS solver predictions for the current iteration step. This represents the average value of the samples in the current iteration step. Let N be the identity matrix and N be the total number of samples.

[0013] According to a specific implementation of an embodiment of this application, the expression for updating the sample set is: , in, It represents the j-th sample in the sample set during the (i+1)-th iteration.

[0014] According to a specific implementation of an embodiment of this application, the expression for the convergence condition is: , in, Let be the average absolute error of the prediction samples in the i-th iteration. This represents the average RANS simulation uncertainty for each sample. It is the Manhattan norm.

[0015] Beneficial effects: The transition model debugging method for accurate compressor flow prediction in this application embodiment systematically debugs the γ transition model, provides an optimization strategy for the transition model under compressor transition conditions, solves the problem of excessively large separation bubble size in the prediction of compressor blade transition flow in existing models, and reveals the key sensitive parameters that determine the prediction accuracy. It provides a systematic parameter debugging method for improving the accuracy of compressor numerical simulation, which is simple and effective. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart of a method for debugging a transition model for accurate prediction of compressor flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the fluid domain, boundary conditions, and mesh topology of a compressor blade cascade according to an embodiment of the present invention; Figure 3 This is a schematic diagram showing the prediction results of the momentum thickness of the suction surface boundary layer before and after model parameter adjustment according to an embodiment of the present invention. Figure 4 The following is a schematic diagram of the prediction results of the static pressure rise coefficient and the friction coefficient of the suction surface before and after the model parameter adjustment according to an embodiment of the present invention. (a) is the prediction result of the static pressure rise coefficient of the blade surface before and after the model parameter adjustment under operating condition 1. (b) is the prediction result of the static pressure rise coefficient of the blade surface before and after the model parameter adjustment under operating condition 2. (c) is the prediction result of the friction coefficient of the suction surface before and after the model parameter adjustment under operating condition 1. (d) is the prediction result of the friction coefficient of the suction surface before and after the model parameter adjustment under operating condition 2. Detailed Implementation

[0018] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0019] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this application, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0021] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The illustrations only show the components related to this application and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0022] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0023] This application provides a method for debugging a transition model for accurate prediction of compressor flow, referring to... Figure 1 This includes the following steps: Step 1: Obtain key flow physical quantities of a typical compressor blade under different incoming flow conditions; Step 2: Determine the set of parameters to be debugged and the range of variation for each parameter in the set, and generate an initial sample set for the set of parameters to be debugged; Step 3: Using the obtained key flow physical quantities as calibration targets, iteratively update the set of parameters to be debugged based on the initial sample set and the iterative calibration framework based on DIOM (data-driven iterative optimization method). Step 4: When the loop reaches the maximum number of iterations or the prediction accuracy reaches the convergence condition, obtain the optimal estimated parameter set, and output the mean and standard deviation of the optimal estimated parameter set as the parameter adjustment result.

[0024] In this embodiment, a data-driven iterative optimization method (DIOM) is proposed to perform transition model tuning. This method first determines the parameter set to be tuned, including transition model parameters and the inlet turbulence length scale; then, it constructs the DIOM iterative calibration framework, using high-fidelity compressor flow data to iteratively drive parameter optimization and updates; when the maximum number of iterations is reached or the prediction accuracy meets the convergence condition, the final parameter tuning result is obtained.

[0025] In one embodiment, obtaining key flow physical quantities of a typical compressor cascade under different incoming flow conditions includes: Select a typical compressor blade cascade and obtain the key flow physical quantities of the typical compressor blade cascade under different incoming flow conditions through experiments or high-fidelity simulation methods.

[0026] Furthermore, the expression for the key flow physical quantity is: , in, As a key flow physical quantity, The surface static pressure rise coefficient, The coefficient of surface friction, For different flow directions and positions, This is the total pressure loss coefficient at the export end. This is to account for errors in experimental or high-fidelity simulation data.

[0027] Furthermore, the expression for the set of parameters to be debugged is: , in, For the parameter set to be debugged, C TU1 To control the parameters of the natural transition critical Reynolds number, C TU2 To control the parameters of the bypass transition critical Reynolds number, C TU3 To control the parameters of the bypass transition critical Reynolds number, the intensity of free-flow turbulence, and the rate of change of the pressure gradient, C PG1The parameter C is used to control the value of the critical Reynolds number at transition under the compressive gradient. PG2 The parameter C is used to control the value of the critical Reynolds number at transition under the inverse pressure gradient. PG3 The parameter L is used to control the critical Reynolds number of the transition zone. T The length scale is the inlet turbulence length scale.

[0028] In one embodiment, the initial sample set for generating the parameter set to be debugged includes: The Latin hypercube sampling algorithm is used to generate the initial sample set of the parameter set to be debugged. Where j represents the sample number and N represents the total number of samples. This is the j-th initial sample, i.e., the j-th initial parameter sample. Each initial sample represents a set of parameters to be debugged.

[0029] In one embodiment, the iterative update of the parameter set to be debugged based on the initial sample set and the DIOM-based iterative calibration framework includes: Step 3.1: For the sample set of the parameter set to be debugged in the i-th iteration... Each set of parameters to be debugged in the sample set is input into the RANS solver to simulate different inflow conditions of the selected typical compressor blade cascade and calculate the predicted values ​​of key flow physical quantities for each set of parameters to be debugged. Step 3.2: Based on the predicted values ​​of the key flow physical quantities in the i-th iteration and the key flow physical quantities obtained in Step 1, calculate the prediction error and the Kalman gain matrix; Step 3.3: Update the sample set based on the prediction error and the Kalman gain matrix.

[0030] Furthermore, the expression for the predicted value of the key flow physical quantity is as follows: , in, Let R(·) be the predicted value of the key flow physical quantity for the j-th sample in the i-th iteration, and R(·) be the RANS solver. Let ε be the j-th sample in the sample set during the i-th iteration. (j) ε represents the uncertainty error of the RANS simulation. (j) Estimated by grid discretization error and boundary condition error.

[0031] Furthermore, the expression for the prediction error is: , in, This represents the prediction error; The expression for the Kalman gain matrix is: , in, Here is the Kalman gain matrix. for and The covariance matrix between them for and The covariance matrix, This represents the average value of the high-precision data error. This is the average value of the RANS solver predictions for the current iteration step. This represents the average value of the samples in the current iteration step. Let N be the identity matrix and N be the total number of samples.

[0032] Furthermore, the expression for updating the sample set is: , in, It represents the j-th sample in the sample set during the (i+1)-th iteration.

[0033] Furthermore, if the prediction accuracy does not meet the convergence condition, the process returns to step 3.1 for the next iteration. The expression for the convergence condition is: , in, Let be the average absolute error of the prediction samples in the i-th iteration. This represents the average RANS simulation uncertainty for each sample. It is the Manhattan norm.

[0034] Assume that the results of the experiment or high-fidelity simulation follow a standard deviation of . The Gaussian distribution and the RANS simulation results follow a standard deviation of ε. (j) The two are Gaussian distributed and uncorrelated. Based on the above assumptions, the prediction error of RANS simulation... The standard deviation is ε. (j) + The model follows a Gaussian distribution. When the prediction error is below the above threshold, it indicates that the accuracy of the debugged model has reached a level comparable to that of experiments or high-fidelity simulations, and the debugging process can be considered complete.

[0035] The debugging method of this application will be described in detail below with a specific embodiment, which includes the following contents.

[0036] S1. Transition model parameters were calibrated based on high-precision data from three sets of different inflow conditions for a compressor blade cascade. The Mach number ranged from 0.10 to 0.67, and the Reynolds number ranged from 8 × 10⁻⁶. 4 -1.5×10 5The surface static pressure rise coefficient of the blade cascade under three different inflow conditions was obtained using high-fidelity simulation methods. With surface friction coefficient As a key flow physical quantity for model debugging In this example, the error of the high-fidelity simulation results is estimated at 3% of the measured physical quantity.

[0037] S2. Determine the parameter set to be debugged. The parameter tuning range is limited to the original parameter values ​​of 0%-300%. An initial sample set of the parameter set to be tuned is generated using the Latin hypercube sampling algorithm. , where j represents the sample number and N represents the total number of samples. In this example, N=80.

[0038] S3. Using key flow physical quantities obtained from high-fidelity simulation as calibration targets, an iterative calibration framework based on DIOM is used for the parameter set to be debugged. The iterative update process includes the following three steps: S3.1, For the parameter set of the i-th iteration Each set of parameter samples in the dataset is input into the RANS solver to simulate the three different inflow conditions of the compressor blade cascade selected in step S1, and the key flow physical quantities predicted for each sample are calculated. The specific algorithm is as follows: .

[0039] like Figure 2 The figure shows the computational domain, boundary conditions, and mesh topology of the compressor blade cascade in a RANS simulation. Figure 2 p in t T t α, TI, and L are the inlet conditions, representing the total inlet pressure, total temperature, airflow angle, turbulence intensity, and turbulence length scale, respectively. Figure 2 c x This is the axial chord length of the blade cascade. The mesh adopts an O-4H type topology, and the height of the first layer of mesh near the wall ensures... , This represents the height of the first layer of mesh near the wall. This represents the uncertainty error of the RANS simulation, which is estimated at 3% of the measured physical quantity in this example.

[0040] S3.2 Key flow physical quantities based on the i-th iteration prediction Key flow physical quantities obtained from experiments or high-fidelity simulations in step S1 Calculate the prediction error With Kalman gain matrix :

[0041] , S3.3, Based on prediction error With Kalman gain matrix Update the parameter set: .

[0042] S4. When the loop reaches the maximum number of iterations or the prediction accuracy reaches the convergence condition, the optimal estimated parameter set is considered to have been found, and the mean and standard deviation of the optimal estimated parameter set are output as the parameter tuning result; otherwise, return to step S3.1 for the next iteration. The prediction accuracy convergence condition is defined as follows: , in, Let be the average absolute error of the prediction samples in the i-th iteration. This represents the average RANS simulation uncertainty for each sample. This is the Manhattan norm. In this example, the maximum number of iterations is 30.

[0043] Comparative analysis of calculation results from different aspects: such as Figure 3 As shown, θ is the momentum thickness of the flow boundary layer, and x c Here, C represents the flow-length coordinate, and C is the length scale parameter used for dimensionless optimization; in this example, it is the chord length. SST-γ is the turbulence / transition model used for DIOM optimization in this example. Figure 3 The comparison is with the original model ( Figure 3 SST-γ (before debugging) and the model after debugging ( Figure 3 The suction surface boundary layer momentum thickness distribution along the flow direction is predicted using SST-γ (after debugging) and high-precision data from Large Eddy Simulation (LES). LES data shows that the boundary layer develops rapidly after 60% chord length. The original model underestimated the momentum thickness growth because it failed to capture flow reattachment. The model optimized using the debugging method proposed in this application significantly improves prediction accuracy after 75% chord length, accurately capturing flow reattachment in this region. Figure 4 The surface static pressure coefficient C of the blade is given. p and surface friction coefficient C fThe distribution of the pressure rise coefficient on the blade surface is shown in Figure 1. (a) Figure 2 shows the predicted results of the static pressure rise coefficient on the blade surface before and after model parameter adjustment for operating condition 1. (b) Figure 3 shows the predicted results of the static pressure rise coefficient on the blade surface before and after model parameter adjustment for operating condition 2. (c) Figure 4 shows the predicted results of the friction coefficient on the suction surface before and after model parameter adjustment for operating condition 1. (d) Figure 5 shows the predicted results of the friction coefficient on the suction surface before and after model parameter adjustment for operating condition 2. The data from direct numerical simulation show that there is a closed separation bubble and separation-induced transition near 60% chord length of the suction surface. The original model predicted open separation near 40% chord length, resulting in a significant deviation in the blade load prediction. In contrast, the model calibrated by the adjustment method proposed in this application predicted the separation flow topology better, and the prediction results were in better agreement with the data from direct numerical simulation. The blade cascade results for another operating condition show that the physical process is similar to the former. The original model also predicted open separation, resulting in a lower blade load prediction. The calibrated model predicted a closed separation bubble, thus significantly improving the prediction accuracy of the blade load.

[0044] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for calibrating a transition model for accurate prediction of compressor flow, characterized in that, include: Step 1: Obtain key flow physical quantities of a typical compressor blade under different incoming flow conditions; Step 2: Determine the set of parameters to be debugged and the range of variation for each parameter in the set, and generate an initial sample set for the set of parameters to be debugged; Step 3: Using the key flow physical quantities obtained as calibration targets, iteratively update the parameter set to be debugged based on the initial sample set and the DIOM-based iterative calibration framework; Step 4: When the loop reaches the maximum number of iterations or the prediction accuracy reaches the convergence condition, obtain the optimal estimated parameter set, and output the mean and standard deviation of the optimal estimated parameter set as the parameter tuning result. The key flow physical quantities for obtaining typical compressor blades under different incoming flow conditions include: Select a typical compressor blade cascade and obtain the key flow physical quantities of the typical compressor blade cascade under different incoming flow conditions through experiments or high-fidelity simulation methods; The expression for the key flow physical quantity is: ， wherein, is the key flow physical quantity, is the surface static pressure rise coefficient, is the surface friction coefficient, is the velocity profile at different streamwise locations, is the exit total pressure loss coefficient, is the test or high-fidelity simulation data error; The expression for the set of parameters to be debugged is: ， in, For the parameter set to be debugged, C TU1 To control the parameters of the natural transition critical Reynolds number, C TU2 To control the parameters of the bypass transition critical Reynolds number, C TU3 To control the parameters of the bypass transition critical Reynolds number, the intensity of free-flow turbulence, and the rate of change of the pressure gradient, C PG1 The parameter C is used to control the value of the critical Reynolds number at transition under the compressive gradient. PG2 The parameter C is used to control the value of the critical Reynolds number at transition under the inverse pressure gradient. PG3 The parameter L is used to control the critical Reynolds number of the transition zone within the separation region. T The length scale is the inlet turbulence length scale.

2. The method for debugging a transition model for accurate prediction of compressor flow according to claim 1, characterized in that, The initial sample set for generating the parameter set to be debugged includes: The Latin hypercube sampling algorithm is used to generate an initial sample set of parameters to be debugged.

3. The method for debugging a transition model for accurate prediction of compressor flow according to claim 1, characterized in that, The iterative calibration framework based on the initial sample set and DIOM iteratively updates the parameter set to be debugged, including: For the sample set of the parameter set to be debugged in the i-th iteration, each set of parameter sets to be debugged in the sample set is input into the RANS solver to simulate different incoming flow conditions of the selected typical compressor blade cascade and calculate the predicted values ​​of key flow physical quantities for each set of parameter sets to be debugged. Based on the predicted values ​​of key flow physical quantities in the i-th iteration and the key flow physical quantities obtained in step 1, the prediction error and Kalman gain matrix are calculated. The sample set is updated based on the prediction error and the Kalman gain matrix.

4. The method for debugging a transition model for accurate prediction of compressor flow according to claim 3, characterized in that, The expression for the predicted value of the key flow physical quantity is: ， in, Let R(·) be the predicted value of the key flow physical quantity for the j-th sample in the i-th iteration, and R(·) be the RANS solver. Let ε be the j-th sample in the sample set during the i-th iteration. (j) This represents the uncertainty error in RANS simulation.

5. The method for debugging a transition model for accurate prediction of compressor flow according to claim 4, characterized in that, The expression for the prediction error is: ， in, This represents the prediction error; The expression for the Kalman gain matrix is: ， in, Here is the Kalman gain matrix. for and The covariance matrix between them for and The covariance matrix, This represents the average error of the high-precision data. This is the average value of the RANS solver predictions for the current iteration step. This is the average value of the samples in the current iteration step. Let N be the identity matrix and N be the total number of samples.

6. The method for debugging a transition model for accurate prediction of compressor flow according to claim 5, characterized in that, The expression for updating the sample set is: ， in, It represents the j-th sample in the sample set during the (i+1)-th iteration.

7. The method for debugging a transition model for accurate prediction of compressor flow according to claim 1, characterized in that, The expression for the convergence condition is: ， in, Let be the average absolute error of the prediction samples in the i-th iteration. This represents the average RANS simulation uncertainty for each sample. It is the Manhattan norm.

Images