A boundary slip velocity calculation method and perception system based on numerical simulation and data driving
By constructing a paired database and a physical mapping model, the wall slip velocity is calculated using pressure observation data from a combined asymmetric pressure measurement tube group. This solves the problem of accurately obtaining the wall slip velocity in a rarefied gas environment and realizes the engineering-based indirect measurement and high-precision calculation of microscopic quantities.
Patent Information
- Application Number
- CN202610486828.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-14
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2046-04-14
AI Technical Summary
Existing technologies struggle to accurately obtain wall slip velocity in rarefied gas environments. Direct measurement is difficult and easily interferes with the flow field. High-precision numerical simulation results are hard to convert into engineering-usable information, and macroscopic observation data lacks physical fidelity correlation with microscopic flow characteristics.
By constructing a paired database and using machine learning or symbolic regression algorithms to establish a physical mapping model, the wall slip velocity is calculated based on pressure observation data from a combined asymmetric pressure measurement tube group. By combining direct simulation Monte Carlo method and data-driven method, the engineering indirect measurement of micro-quantities is realized.
It achieves accurate calculation and sensing of wall sliding velocity in a rarefied gas environment, avoids sensor interference with micro-flow field, balances physical fidelity and calculation feasibility, adapts to different engineering application scenarios, and ensures the accuracy and physical authenticity of calculation results.
Smart Images

Figure CN122217575B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aircraft flow field sensing technology, and in particular to a boundary slip velocity calculation method and sensing system based on numerical simulation and data-driven methods. Background Technology
[0002] In various engineering fields such as aerospace, microelectromechanical systems (MEMS), and vacuum technology, gas flow often occurs in rarefied or microscale environments. In these conditions, the characteristic size of the gas is on the same order of magnitude as the mean free path of its molecules. The flow enters the slip flow or transition flow region, where the "no-slip" boundary conditions at the wall, as defined by the Navier-Stokes equations based on the continuous medium assumption, completely fail. The gas exhibits macroscopic wall slip velocity at solid walls. As a core physical quantity characterizing the flow properties of rarefied gases, the accurate acquisition of wall slip velocity is crucial for assessing the aerodynamic and aerothermal characteristics of high-speed near-space vehicles, as well as for precise flow control within microchannels. It serves as a key basis for engineering design and performance optimization in related fields.
[0003] However, existing methods for obtaining wall slip velocity face significant technical challenges, failing to meet the demands of practical engineering applications. These challenges manifest in two main ways: First, direct experimental measurement is difficult and prone to interfering with the flow field. Wall slip velocity exists at the bottom of an extremely thin boundary layer, making direct engineering measurement of this physical quantity extremely challenging in real-world scenarios such as flight and wind tunnel testing. Furthermore, the geometry of traditional measurement sensors severely interferes with the microscopic intrinsic flow field, hindering non-invasive slip data acquisition and leading to discrepancies between measurement results and actual flow field characteristics. Second, high-precision numerical simulations are disconnected from conventional engineering observation data. Numerical simulation methods such as Direct Monte Carlo (DSMC) can accurately deconstruct the microscopic or mesoscopic flow characteristics of gases at the molecular dynamics level and accurately capture wall slip velocity. However, the calculation process of these methods depends on the complete analysis of microscopic flow details, resulting in high computational complexity. Furthermore, they cannot establish a direct and physically faithful logical connection with the macroscopic surface pressure data obtained through pressure gauges or pipes in engineering practice, making it difficult to transform high-precision simulation results into effective information that can be used in engineering.
[0004] Currently, the industry lacks a mature technical solution that can effectively map macroscopic observation parameters to microscopic slip characteristics using conventionally measurable macroscopic pressure data combined with standardized numerical simulation processes. It also fails to achieve accurate sensing of wall slip velocity through indirect characterization while maintaining physical fidelity. Therefore, breaking down the technical barriers between macroscopically measurable physical quantities and microscopic aerodynamic characteristics, and constructing a method for obtaining wall slip velocity that combines physical realism and engineering feasibility, has become a critical engineering problem urgently needing to be solved in fields such as rarefied gas dynamics and aircraft flow field sensing. Summary of the Invention
[0005] The purpose of this invention is to provide a boundary slip velocity calculation method and sensing system based on numerical simulation and data-driven methods. This invention solves the technical problems of the difficulty in directly measuring wall slip velocity in rarefied gas dynamic environments and the lack of physical fidelity correlation between microscopic flow characteristics and macroscopic pressure measurement data. It enables accurate calculation and sensing of microscopic wall slip velocity, which cannot be directly measured, based on conventional macroscopic pressure observation data.
[0006] To achieve the above objectives, this invention provides a method for calculating boundary slip velocity based on numerical simulation and data-driven approaches, comprising the following steps: S1. Construct a physically associated paired database: Based on the direct simulation Monte Carlo numerical simulation method, for the same set of environmental parameters, simultaneously acquire the wall slip velocity calibration value under the model without pressure gauge tubes and the characteristic point pressure observation data under the combined asymmetric pressure gauge tube group with specific geometric structure, and form a standard paired dataset by corresponding the environmental parameter set, pressure observation data and slip velocity calibration value one by one. S2. Extract a specific mapping model through data-driven approach: Use machine learning or symbolic regression algorithms to extract features from the standard paired dataset and establish a physical mapping model defined by the physical characteristics of the dataset. The physical mapping model is as follows: ; in, All data are pressure observation data. This represents the feature vector of environmental parameters. S3. Calculation and output of sliding velocity: Determine whether the feature vector formed by the collected environmental parameters and pressure data falls within the parameter space envelope range defined by the standard paired dataset. If it is within the range, input it into the physical mapping model, calculate and output the wall sliding velocity under the corresponding working condition. If it is outside the range, mark the data as non-applicable domain data.
[0007] Preferably, the combined asymmetric pressure gauge assembly adopts an asymmetric orientation layout, including an upstream pressure gauge with its axis perpendicular to the wall and a downstream pressure gauge with its axis inclined relative to the wall. The angle between the axis of the downstream pressure gauge and the wall is 15° to 75°.
[0008] Preferably, the specific method for obtaining the wall slip velocity calibration value is as follows: after the numerical simulation flow field without a piezometer model reaches statistical stability, the average velocity profile in the wall normal direction is extracted, and the wall slip velocity is then... The extrapolated velocity at a given location is defined as the calibrated value of the wall sliding velocity under the corresponding environmental parameters.
[0009] Preferably, the environmental parameter set includes the incoming Mach number and macroscopic parameters characterizing the static state of the flow field; S1 also includes determining the mean free path of gas molecules based on the macroscopic parameters and calculating the Knudsen number in combination with the geometric characteristic length of the pressure measuring tube, and the environmental parameter feature vector includes at least the incoming Mach number and the Knudsen number.
[0010] Preferably, the geometric parameter setting criteria for constructing the combined asymmetric pressure gauge assembly are as follows: The length-to-diameter ratio is set according to the pressure balance requirements inside the pressure measuring tube in order to collect the true local pressure and reduce the geometric interference to the intrinsic flow field. The distance between the axes of the two pipes is set according to the diameter of the pressure measuring tube to balance the spatial local equivalence of the measurement point and the independent interference of the flow field. The upper limit of the distance ensures that the measurement areas of the two pipes are in the same spatial local neighborhood, and the lower limit ensures that the disturbance of the upstream pressure measuring tube does not interfere with the pressure acquisition of the downstream pressure measuring tube.
[0011] Preferably, the machine learning algorithm is an artificial neural network algorithm, and the physical mapping model is a trained artificial neural network model, which includes a network topology structure with neuron connections and weight parameters stored in the topology structure. When using the symbolic regression algorithm, the physical mapping model is an explicit mathematical expression generated by nonlinear analytical optimization in the feature space through the symbolic regression algorithm.
[0012] Preferably, when using artificial neural network algorithms, the local environmental pressure is first utilized. right , The data is then processed to be dimensionless, and the dimensionless pressure data is fused with environmental features as model input. When using the symbolic regression algorithm, the input features are first normalized and preprocessed, and then substituted into the nonlinear polynomial equation to solve for the dimensionless slip velocity.
[0013] Preferably, the environmental parameter set also includes a wall fitness coefficient characterizing the gas-solid interaction properties, which is incorporated into the standard paired dataset for the construction and training of the physical mapping model.
[0014] A boundary slip velocity sensing system based on numerical simulation and data-driven methods includes: The multi-condition flow field correlation simulation and dataset construction module is used to perform synchronous numerical simulations of conditions without and with piezometers based on the direct simulation Monte Carlo method, and to extract and construct a standard paired dataset in which environmental parameters, pressure observation data and slip velocity calibration values correspond one-to-one. The nonlinear mapping relationship extraction module is used to perform machine learning or logistic fitting on standard paired datasets, and generate a physical mapping model after feature extraction; The parameter mapping and velocity calculation module is used to receive the collected flow field pressure data and environmental parameters, perform out-of-distribution detection to determine whether the input feature vector is within the parameter space envelope. If it is within the range, the physical mapping model is called to calculate and output the wall sliding velocity. If it is outside the range, the data is marked as non-applicable domain data.
[0015] Therefore, the present invention employs the above-mentioned boundary slip velocity calculation method and sensing system based on numerical simulation and data-driven approach, and the technical effects are as follows: 1. Achieve indirect engineering measurement of microscopic physical quantities. By using a combined asymmetric pressure measurement tube group, the wall slip velocity, which cannot be directly observed, is transformed into macroscopic pressure data that is easy to collect in engineering. This avoids the interference of sensor geometry on the microscopic intrinsic flow field and solves the engineering problem of direct measurement of slip velocity in rarefied flow environments.
[0016] 2. Balancing physical fidelity and computational feasibility, a mapping model is constructed based on the high-precision numerical simulation of the direct Monte Carlo method, which fully preserves the physical characteristics at the molecular dynamics level. At the same time, the complex micro-flow deconstruction is transformed into a logical mapping of macroscopic parameters, enabling the effective and reliable indirect perception of unmeasurable micro-physical quantities.
[0017] 3. It possesses strong algorithm adaptability and engineering scalability. The established mapping logic does not depend on a specific mathematical form. It can flexibly select the implicit model of artificial neural network or the explicit mathematical analytical expression of symbolic regression according to the processing capabilities of the computing platform, adapting to different engineering application scenarios and having high industrial promotion value.
[0018] 4. Built-in high safety margin solution verification mechanism: By detecting the distribution of the parameter space envelope based on the original standard paired dataset, the validity of the input data is judged, preventing the model from making non-physical predictions in the unknown data domain, and ensuring the physical authenticity and accuracy of the slip velocity solution results from the process. Attached Figure Description
[0019] Figure 1 This is an overall flowchart of the boundary slip velocity calculation method based on numerical simulation and data-driven approach of the present invention. Figure 2This is an isometric schematic diagram of the pressure measuring tube structure used in the embodiments of the present invention; Figure 3 This is a schematic diagram of a calculation model of a smooth wall without a pressure gauge for calibrating the slip velocity of the intrinsic flow field in an embodiment of the present invention; Figure 4 This is a schematic diagram of a flow field calculation model with interference from a pressure measuring tube used to obtain the observed pressure at the pressure measuring hole in an embodiment of the present invention; Figure 5 This is a diagram illustrating the training process and prediction performance evaluation of the neural network mapping model in this embodiment of the invention. Figure 6 This is a graph showing the prediction accuracy and parameter contribution of the explicit analytical expression of symbolic regression in an embodiment of the present invention. Detailed Implementation
[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0021] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0022] Example 1 like Figure 1 As shown, this invention provides a boundary slip velocity calculation method based on numerical simulation and data-driven approach. First, a paired database is constructed through numerical simulations that deconstruct the microscopic or mesoscopic flow characteristics of fluids. Simultaneously, calibration values of wall slip velocity under a piezometer-less model and pressure observation data under a piezometer model with a specific geometry are acquired. Subsequently, machine learning or symbolic regression algorithms are used to extract features from the paired datasets, establishing a physical mapping model defined by the physical properties of the datasets. Finally, the working condition observation data is acquired and input into the model to realize the calculation and sensing of the wall sliding velocity.
[0023] Based on numerical simulation methods capable of deconstructing gas flow characteristics, this study simultaneously performs intrinsic flow field simulations without piezometers and disturbance flow field simulations with combined asymmetric piezometer arrays for the same set of environmental parameters. The velocity profile normal to the wall is extracted from the piezometer-free simulation, and the slip velocity calibration values at the wall are extrapolated. Pressure observation data at specific acquisition points are extracted from the piezometer-equipped simulation. A one-to-one correspondence is established between the environmental parameter set, pressure observation data, and slip velocity calibration values to form a standard structured paired dataset.
[0024] The combined asymmetric pressure gauge assembly employs an asymmetric orientation layout, comprising an upstream pressure gauge with its axis perpendicular to the wall, and a downstream pressure gauge with its axis inclined relative to the wall. Utilizing the different geometric orientations of the two gauges, it captures the orientation-related pressure response differences in the near-wall flow field within its spatial neighborhood caused by molecular non-equilibrium effects. The angle between the downstream pressure gauge's axis and the wall is set between 15° and 75°, preferably 45°. The length-to-diameter ratio of the pressure gauge is set according to internal pressure balance requirements to ensure that the pressure gauge collects accurate local pressure and maintains response fidelity. The pipe diameter is determined accordingly. Set the distance between the two pipe axes The upper limit of the spacing is set to ensure that the measurement areas of the two tubes are within the same local neighborhood in space, and the lower limit of the spacing is set to ensure that the disturbance generated by the upstream tube does not interfere with the pressure acquisition point of the downstream tube.
[0025] Physical features are extracted from the paired dataset using artificial neural network algorithms or symbolic regression algorithms to establish a physical mapping model defined by the physical characteristics of the dataset. When using artificial neural network algorithms, the physical mapping model is represented by a network topology structure containing training weight parameters. When using symbolic regression algorithms, the physical mapping model is represented by an explicit mathematical expression generated by nonlinear analytical optimization in the feature space using symbolic regression algorithms.
[0026] The system acquires environmental parameters and pressure signals from a combined asymmetric pressure measuring tube array. Before calling the physical mapping model, it performs out-of-distribution detection on the input vector to determine whether the feature vector formed by the acquired environmental parameters and pressure data falls within the parameter space envelope range defined by the standard paired dataset. If it is within the range, it is input into the physical mapping model to calculate and output the wall slip velocity under the corresponding working condition. If it is outside the range, the data is marked as non-applicable domain data.
[0027] This embodiment sets the input parameter range for the data-driven model based on the actual operating envelope of the aircraft or the requirements of wind tunnel testing. This embodiment focuses on the near-space high-speed rarefied flow condition, setting the incoming flow Mach number accordingly. The range is 12 to 27, and the altitude range is set at 75 km to 100 km (this altitude range covers typical near-space rarefaction environments).
[0028] In addition, the environmental parameter set also includes wall fitness coefficients characterizing the gas-solid interaction properties. (Values range from 0.5 to 1.0). It should be noted that the wall adaptability coefficient... In numerical simulations, physical boundary conditions are easily given, but in actual flight environments or wind tunnel tests, due to wall contamination, oxidation, and the complexity of material properties, this parameter is difficult to measure directly; this invention constructs a system containing... The influence mapping model enables the perception of slip velocity under the influence of factors that are difficult to observe directly.
[0029] In addition, it is also necessary to consider the mean free path of gas molecules at different altitudes. Compared with the characteristic geometry of the pressure gauge (i.e., the diameter of the pressure gauge) Calculate Knudsen number The calculation formula is: ; The above parameter range ( , , These together constitute the physical space envelope of the mapping model in this embodiment.
[0030] like Figure 2 As shown, the pressure measuring tube structure used in this embodiment includes: a vertical pressure measuring tube located upstream, with its tube axis perpendicular to the wall surface; and an inclined pressure measuring tube located downstream, with its tube axis forming a 45° angle with the wall surface. The relevant geometric parameters are specifically set as follows: Piezometer diameter The setting is 1mm to balance the feasibility of capturing local flow field features and microfabrication processes; The length-to-diameter ratio of the pressure measuring tube ( The values are set to 8 and 12 to ensure that the pressure inside the tube can be fully balanced under different rarefaction levels. The distance between the two pipe axes is set to 5mm. This distance has been verified by simulation to ensure that the two pipes are in the same local neighborhood in space, while effectively avoiding interference from the upstream pressure measuring pipe opening that could affect the downstream acquisition point.
[0031] For each set of environmental parameter tuples within the parameter envelope ( , , ),like Figure 3 As shown, the intrinsic flow field of a smooth wall (without piezometer interference) is simulated using the DSMC method. After the flow field reaches statistical steady state, the average velocity profile is extracted along the wall normal.
[0032] In the simulation results without a pressure gauge, the wall position (i.e., ...) is determined by second-order or linear extrapolation of the near-wall velocity profile. The speed at (the point) is defined as the calibrated value of the wall sliding speed under this working condition. (or its dimensionless form) This serves as the ground truth for subsequent model training.
[0033] To maintain complete consistency of environmental parameters, a combined asymmetric pressure gauge geometry is introduced into the flow field, such as... Figure 4As shown, the DSMC method was used again to calculate the disturbed flow field containing the pressure measurement orifice. After the pressure inside the pipe stabilized, the pressure observation value at the bottom of the upstream vertical pipe was extracted. Pressure observations at the bottom of the downstream inclined pipe .
[0034] Pressure observation data under the same environmental parameters ( , ), environmental characteristics ( , , ) and the corresponding sliding speed calibration value A one-to-one correspondence is established to form standard structured data tuples. By traversing the entire physical envelope, a calibration database is generated for subsequent model training and validation.
[0035] Based on the constructed database, a slip velocity mapping model was established. To eliminate the dimensional influence of absolute pressure values at different altitudes and to ensure the robustness of the model under varying pressure environments, the local environmental pressure was first utilized. Observation pressure , Perform dimensionless processing: ; Subsequently, a generalized mapping function incorporating environmental characteristics and dimensionless pressure characteristics is established: ; in, This represents computational logic acquired through data-driven training, and its specific implementation includes, but is not limited to, explicit algebraic equations or implicit neural computational mappings.
[0036] Regarding Mach number ,altitude For typical high-speed rarefied flow fields, the following two parallel schemes are used to construct the mapping model: Method 1 uses an implicit neural network model based on a multilayer perceptron (MLP). This model constructs a four-layer deep neural network, with the specific architecture as follows: The input layer has four feature neurons, each corresponding to... These are the four physical quantities. The hidden layer consists of two layers: the first hidden layer contains 32 neurons, and the second hidden layer contains 16 neurons. For the activation function, the hidden layer uses the ReLU function, while the output layer uses a linear activation function.
[0037] After training using the backpropagation algorithm, the network maps the physical feature space into a high-dimensional weight matrix, thereby achieving implicit perception of slip velocity. For example... Figure 5As shown, an independent test set error analysis was performed on the neural network model. The results show that the standard deviation between its predicted values and the DSMC calibration values is extremely small, proving that the neural network has extremely high fitting accuracy when dealing with high-dimensional complex nonlinear interferometric flow fields. Figure 5 The error convergence trajectory during the neural network training process and the regression distribution of predicted and true values are shown. The results show that the root mean square error (RMS) is less than 0.04 and the maximum absolute error is about 0.1, which verifies that the model has extremely high representation accuracy for the nonlinear mapping relationship of slip velocity. It also proves that the method has strong robustness in dealing with complex nonlinear pressure signals caused by geometric interference, ensuring the reliability of subsequent solutions.
[0038] Method two is based on explicit mathematical analytical expressions using symbolic regression. It utilizes the symbolic regression algorithm to perform nonlinear analytical optimization within the physical feature space, extracting explicit mapping equations with physical interpretability. In this approach, a normalization preprocessing is first performed, yielding the standardized term as follows: ; Subsequently, the dimensionless slip velocity is calculated by substituting the following nonlinear polynomial equation obtained through analytical optimization. : ; like Figure 6 As shown, an error analysis is performed on the above symbolic regression formula. (Comparison) Figure 5 and Figure 6 It is evident that the explicit analytical expression extracted based on symbolic regression achieves perceptual accuracy comparable to that of the neural network model. Its algebraic analytical form not only establishes a direct physical relationship between input parameters and slip velocity, but also provides a convenient mathematical expression for performing real-time calculations in resource-constrained environments.
[0039] Obtain environmental parameters (including Mach number) under the test condition. Local environmental pressure corresponding to altitude (etc.) and the pressure signal sensed and returned by the pressure measurement unit. and Then, the feature transformation is performed as follows: The dimensionless pressure characteristics were obtained through calculation, namely... and This eliminates the influence of the pressure dimension on subsequent analysis.
[0040] Based on the fixed mean and scaling parameters, the input vector is precisely mapped to the standardized feature space, ultimately obtaining the normalized feature terms. .
[0041] Use the current input vector Compared with the physical space envelope, if all input parameters fall within the envelope (i.e. , If any key parameter exceeds the envelope range, the data set is marked as "non-applicable domain data".
[0042] The generalized mapping function, which incorporates environmental features and dimensionless pressure characteristics, is invoked to create this mapping model. Using the feature quantities processed by the feature transformation operation as input, the wall sliding velocity sensing results under the corresponding working conditions are directly calculated and output. This provides a basis for the aerodynamic / thermal correction of aircraft.
[0043] The sensing system used to implement the above method includes a multi-condition flow field correlation simulation and dataset construction module, which is used to perform synchronous numerical simulation of conditions without and with piezometers based on the direct simulation Monte Carlo method, and extract and construct a standard paired dataset in which environmental parameters, pressure observation data and slip velocity calibration values correspond one-to-one. The nonlinear mapping relationship extraction module is used to perform machine learning or logistic fitting on standard paired datasets, and generate a physical mapping model after feature extraction; The parameter mapping and velocity calculation module is used to receive the collected flow field pressure data and environmental parameters, perform out-of-distribution detection to determine whether the input feature vector is within the parameter space envelope. If it is within the range, the physical mapping model is called to calculate and output the wall sliding velocity. If it is outside the range, the data is marked as non-applicable domain data.
[0044] Therefore, this invention adopts the above-mentioned boundary slip velocity calculation method and sensing system based on numerical simulation and data driving. First, by directly simulating the Monte Carlo method, the wall slip velocity calibration value of the model without pressure gauges and the pressure observation data of the combined asymmetric pressure gauge group are acquired simultaneously for the same set of environmental parameters. A standard paired dataset corresponding to the three is constructed. Then, the features of the dataset are extracted by artificial neural network or symbolic regression algorithm to establish a physical mapping model with pressure observation data and environmental parameter feature vectors as input. Finally, after the parameter space envelope range is determined, the collected data that meets the requirements is input into the model to calculate and output the wall slip velocity. Data that exceeds the range is marked as non-applicable domain data. The pressure gauge group includes a vertical upstream pipe and a downstream pipe inclined at 15°-75°. The environmental parameter feature vector includes at least the incoming Mach number and Knudsen number, and can also include the wall adaptation coefficient. The pressure data is preprocessed by dimensionless and normalized methods. This technology effectively solves the problems of direct engineering measurement of wall slip velocity in rarefied gas dynamic environments and the lack of physical fidelity correlation between microscopic flow characteristics and macroscopic pressure measurement data. It realizes the engineering-based indirect measurement of microscopic slip velocity, avoiding sensor interference to the microscopic flow field. The model is built based on high-precision numerical simulation, taking into account both physical fidelity and solution feasibility. It can flexibly select implicit or explicit mapping models, and the algorithm has strong adaptability and engineering scalability. Furthermore, it establishes a solution verification mechanism through distributed external detection to ensure the physical authenticity and accuracy of the solution results, providing key basis for the evaluation of aerodynamic and aerothermal characteristics of near-space high-speed vehicles.
[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for calculating boundary slip velocity based on numerical simulation and data-driven approach, characterized in that, Includes the following steps: S1. Construct a physically associated paired database: Based on the direct simulation Monte Carlo numerical simulation method, for the same set of environmental parameters, simultaneously acquire the wall slip velocity calibration value under the model without pressure gauge tubes and the characteristic point pressure observation data under the combined asymmetric pressure gauge tube group with specific geometric structure, and form a standard paired dataset by corresponding the environmental parameter set, pressure observation data and slip velocity calibration value one by one. The specific method for obtaining the wall slip velocity calibration value is as follows: After the flow field in the numerical simulation without a piezometer model reaches statistical stability, the average velocity profile in the wall normal direction is extracted, and the wall slip velocity is then... The extrapolated velocity at a given location is defined as the calibrated value of the wall sliding velocity under the corresponding environmental parameters. The combined asymmetric pressure gauge assembly adopts an asymmetric orientation layout, including an upstream pressure gauge with its axis perpendicular to the wall and a downstream pressure gauge with its axis inclined relative to the wall. The angle between the axis of the downstream pressure gauge and the wall is 15° to 75°. The length-to-diameter ratio of the pressure gauge is set according to the internal pressure balance requirements, and the ratio is determined based on the pipe diameter. Set the distance between the two pipe axes The upper limit of the spacing is set to ensure that the measurement areas of the two tubes are within the same local neighborhood in space, and the lower limit of the spacing is set to ensure that the disturbance generated by the upstream tube does not interfere with the pressure acquisition point of the downstream tube. S2. Extract a specific mapping model through data-driven approach: Use machine learning or symbolic regression algorithms to extract features from the standard paired dataset and establish a physical mapping model defined by the physical characteristics of the dataset. The physical mapping model is as follows: ; Where p1 and p2 correspond to the pressure data collected by the upstream vertical pressure gauge and the downstream inclined pressure gauge, respectively. This represents the feature vector of environmental parameters. The environmental parameter feature vector must contain at least the incoming Mach number. With Knudsen , ,in, The mean free path of gas molecules, The diameter of the pressure measuring tube; S3. Calculation and output of sliding velocity: Determine whether the feature vector formed by the collected environmental parameters and pressure data falls within the parameter space envelope range defined by the standard paired dataset. If it is within the range, input it into the physical mapping model, calculate and output the wall sliding velocity under the corresponding working condition. If it is outside the range, mark the data as non-applicable domain data.
2. The boundary slip velocity calculation method based on numerical simulation and data-driven approach according to claim 1, characterized in that, The environmental parameter set includes the incoming Mach number and macroscopic parameters characterizing the static state of the flow field; S1 also includes the mean free path of gas molecules determined based on the macroscopic parameters.
3. The boundary slip velocity calculation method based on numerical simulation and data-driven approach according to claim 1, characterized in that, The machine learning algorithm is an artificial neural network algorithm, and the physical mapping model is a trained artificial neural network model, which includes the network topology of neuron connections and the weight parameters stored in the topology. When using the symbolic regression algorithm, the physical mapping model is an explicit mathematical expression generated by nonlinear analytical optimization in the feature space through the symbolic regression algorithm.
4. The boundary slip velocity calculation method based on numerical simulation and data-driven approach according to claim 3, characterized in that, When using artificial neural network algorithms, the local environmental pressure is first utilized. right , The data is then processed to be dimensionless, and the dimensionless pressure data is fused with environmental features as model input. When using the symbolic regression algorithm, the input features are first normalized and preprocessed, and then substituted into the nonlinear polynomial equation to solve for the dimensionless slip velocity.
5. The boundary slip velocity calculation method based on numerical simulation and data-driven approach according to claim 1, characterized in that, The environmental parameter set also includes the wall fitness coefficient, which characterizes the gas-solid interaction properties. This parameter is incorporated into the standard paired dataset for the construction and training of the physical mapping model.
6. A boundary slip velocity sensing system based on numerical simulation and data-driven method, comprising the boundary slip velocity calculation method based on numerical simulation and data-driven method according to any one of claims 1-5, characterized in that, include: The multi-condition flow field correlation simulation and dataset construction module is used to perform synchronous numerical simulations of conditions without and with piezometers based on the direct simulation Monte Carlo method, and to extract and construct a standard paired dataset in which environmental parameters, pressure observation data and slip velocity calibration values correspond one-to-one. The nonlinear mapping relationship extraction module is used to perform machine learning or logistic fitting on a standard paired dataset, and generates a physical mapping model after feature extraction. The parameter mapping and velocity calculation module is used to receive the collected flow field pressure data and environmental parameters, perform out-of-distribution detection to determine whether the input feature vector is within the parameter space envelope. If it is within the range, the physical mapping model is called to calculate and output the wall sliding velocity. If it is outside the range, the data is marked as non-applicable domain data.
Citation Information
Patent Citations
Aircraft wall surface stress tensor sampling method, device and equipment and storage medium
CN116204985A
Data-driven slip flow near-wall region nonlinear constitutive flow field modeling method
CN116882324A
