Method, device and equipment for reconstructing physical field based on limited measuring points

Abstract

The application discloses a physical field reconstruction method and device based on limited measuring points and equipment, and relates to the technical field of liquid rocket engine data processing. The method comprises the following steps: in the process of physical test of a liquid rocket engine, multi-source heterogeneous sensor data collected by a plurality of multi-source heterogeneous sensors is acquired; the sensor data corresponding to a limited number of measuring points is preprocessed to obtain target sensor data; a physical correction reconstruction model based on the limited measuring points is constructed by using a regression machine learning algorithm and combining simulation boundary conditions of actual working conditions of a test bench; simulation analysis is performed on the actual working conditions of the test bench to generate simulation physical field data; a mapping relationship among the target sensor data, high-fidelity data corresponding to the reconstructed physical field, and the simulation physical field data is established based on a data fusion technology, and a multi-fidelity fusion model is constructed by optimizing fusion weights and fusion strategies. The application solves the problems of insufficient space of the test bench and limited sensor layout, and improves the physical field reconstruction precision.

Description

Technical Field

[0001] This invention relates to the field of liquid rocket launcher data processing technology, and in particular to a method, apparatus and equipment for physical field reconstruction based on a finite number of measurement points. Background Technology

[0002] The inverse finite element method (IFM) is used for deformation monitoring of plate and shell structures based on the Reissner-Mindlin medium-thick plate theory and the least squares principle. The IFM has also been applied in the field of structural damage and deformation. Colombo et al. defined a percentage error between the equivalent strain of the structural field and the reconstructed strain using the IFM, demonstrating the effectiveness of this method in predicting cantilever beam damage. For deformation reconstruction of ship and marine structures, a four-node inverse shell element has been disclosed in existing technologies, and its robustness and accuracy have been verified through deformation monitoring of container ships and bulk carriers. Since the location and number of measuring points have a significant impact on the accuracy of deformation prediction results using the IFM, significant progress has been made in optimizing sensor layout. Currently, research on deformation monitoring using the IFM for plate and shell structures is relatively mature, and related algorithms are quite sophisticated. However, research on deformation reconstruction of complex structures is limited. Most studies simplify complex structures to simple plates, beams, etc., and accurate reconstruction methods are still lacking, especially for airfoil beams with irregular cross-sections. Taking the strain field as an example, the test bench blades are subject to complex operating conditions during operation, are affected by load uncertainties, and involve the coupling between multiple physical fields and multiple components, making it difficult to achieve rapid calculation and evaluation of their structural performance.

[0003] In a digital twin test system for liquid rocket engines, the digital twin model needs to be synchronized with the engine test using sensors at measurement points to reflect the engine's physical field information in real time. Traditionally, this involves deploying densely packed measurement points and constructing a point cloud physical field from the data collected by these sensors. The accuracy of this type of physical field reconstruction depends on the data collected from the measurement points, and its effectiveness hinges on the number and location of these points. However, in actual rocket engines with complex structures (such as those inside capillary tubes) and harsh test environments (such as the high-temperature environment of the thrust chamber), it is difficult to deploy too many sensor measurement points. Furthermore, the location of these measurement points usually does not affect engine operation, impacting their effectiveness and resulting in low accuracy in physical field reconstruction.

[0004] Therefore, there is an urgent need to provide a more reliable physical field reconstruction scheme based on finite measurement points. Summary of the Invention

[0005] The purpose of this invention is to provide a physical field reconstruction method, apparatus, and equipment based on a finite number of measurement points, in order to solve the problems of poor effectiveness of measurement points and low reconstruction accuracy of physical fields in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a physical field reconstruction method based on finite measurement points, the method comprising:

[0008] During the physical test of a liquid rocket engine, multi-source heterogeneous sensing data were acquired from multi-source heterogeneous sensors.

[0009] The target sensing data is obtained by preprocessing the sensing data corresponding to a limited number of measurement points selected from the multi-source heterogeneous sensing data; the preprocessing includes at least data cleaning, noise reduction and format unification.

[0010] By using regression machine learning algorithms and combining simulation boundary conditions of actual test bench conditions, a physical correction and reconstruction model based on finite measurement points is constructed.

[0011] The actual working conditions of the test bench are simulated and analyzed to generate simulated physical field data; the simulated physical field data is low-fidelity data.

[0012] Based on data fusion technology, a mapping relationship is established between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data. By optimizing the fusion weights and fusion strategies, a multi-fidelity fusion model is constructed.

[0013] Optionally, a physical correction and reconstruction model based on a finite number of measurement points can be constructed using regression machine learning algorithms and simulation boundary conditions of the actual working conditions of the test bench. Specifically, this includes:

[0014] By utilizing Kriging, radial basis function neural networks, or polynomial response surfaces, and combining the simulated boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on finite measurement points is constructed.

[0015] Based on the physical correction and reconstruction model, reconstructed physical field data at a limited number of measurement points are obtained.

[0016] Optionally, using the Kriging method and combining the simulation boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on a finite number of measurement points is constructed, specifically including:

[0017] Based on the actual working conditions of the test bench, determine the boundary conditions of the simulation model; the boundary conditions include at least temperature, pressure, and flow rate.

[0018] The Kriging model construction method is used to calculate predicted values ​​based on known measurement point data and determine the spatial correlation between data.

[0019] By using an iterative optimization method, the parameters of the Kriging model are continuously adjusted until the preset conditions are met, thus obtaining the target Kriging model.

[0020] Based on the differences between the simulation model and the actual working conditions, the physical parameters that need to be corrected are determined; the physical parameters include at least material properties, geometric structure, and boundary conditions.

[0021] Using the trained target Kriging model, correction values ​​for the parameters to be corrected are generated based on limited measurement point data;

[0022] The correction values ​​are applied to the simulation model to reconstruct the physical correction and reconstruction model.

[0023] Optionally, the step of establishing a mapping relationship between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data based on data fusion technology, and constructing a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies, specifically includes:

[0024] The target sensing data, the reconstructed physical field data, and the simulated physical field data are aligned using a unified grid data alignment.

[0025] The aligned target sensing data, the reconstructed physical field data, and the simulated physical field data are fused with multi-fidelity. Combined with data fusion technology, a mapping relationship is established between the aligned target sensing data, the reconstructed physical field data, and the simulated physical field data to form a multi-fidelity fusion model.

[0026] Optionally, after constructing a multi-fidelity fusion model by optimizing the fusion weights and fusion strategy, the model may also include:

[0027] Establish an objective function that reflects the displacement error of the reconfiguration system under multiple operating conditions;

[0028] Based on the objective function, an optimization algorithm is applied to optimize the multi-fidelity model.

[0029] Optionally, a multi-fidelity fusion model can be constructed, specifically including:

[0030] Construct a three-dimensional geometric model of the liquid rocket engine, perform finite element method simulation, and build a physical field model;

[0031] The measurement data collected from the physical test of the liquid rocket engine were used as the boundary conditions for the finite element simulation to calculate the simulation results under different operating conditions.

[0032] The simulation results under different working conditions are used to train and construct a multi-fidelity physical field model.

[0033] Compared with existing technologies, the physical field reconstruction method based on finite measurement points provided by this invention acquires multi-source heterogeneous sensor data collected by multi-source heterogeneous sensors during the physical test of a liquid rocket engine; preprocesses the sensor data corresponding to a finite number of measurement points from the multi-source heterogeneous sensor data to obtain target sensor data; utilizes a regression machine learning algorithm, combined with the simulation boundary conditions of the actual test bench, to construct a physical correction and reconstruction model based on finite measurement points; performs simulation analysis on the actual test bench conditions to generate simulated physical field data; and establishes a mapping relationship between the target sensor data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data based on data fusion technology, and constructs a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies. This invention reconstructs the physical field using finite measurement point data, combines it with physical field simulation data, and utilizes a multi-fidelity fusion method to achieve multi-fidelity physical field reconstruction, solving problems such as insufficient test bench space and limited sensor layout, improving the accuracy of physical field reconstruction, and laying the foundation for real-time perception of the physical field by sensor models.

[0034] Secondly, the present invention provides a physical field reconstruction device based on a finite number of measurement points, the device comprising:

[0035] The multi-source heterogeneous sensor data acquisition module is used to acquire multi-source heterogeneous sensor data collected by multi-source heterogeneous sensors during the physical test of a liquid rocket engine.

[0036] The target sensing data determination module is used to select a limited number of sensing data corresponding to measurement points from the multi-source heterogeneous sensing data and perform preprocessing to obtain target sensing data; the preprocessing includes at least data cleaning, noise reduction and format unification processing;

[0037] The physical correction and reconstruction model building module is used to construct a physical correction and reconstruction model based on a finite number of measurement points by using regression machine learning algorithms and combining the simulation boundary conditions of the actual working conditions of the test bench.

[0038] The simulated physical field data generation module is used to perform simulation analysis on the actual working conditions of the test bench and generate simulated physical field data; the simulated physical field data is low-fidelity data.

[0039] The multi-fidelity fusion model construction module is used to establish the mapping relationship between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data based on data fusion technology, and to construct a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies.

[0040] Thirdly, the present invention provides a physical field reconstruction device based on a finite number of measurement points, the device comprising:

[0041] The system includes a memory, a processor, and a communication interface coupled to the processor; the memory stores a computer program that can be run by the processor; when the processor runs the computer program, it executes the aforementioned physical field reconstruction method based on finite measurement points.

[0042] Fourthly, the present invention provides a computer storage medium storing instructions that, when executed, implement the above-described physical field reconstruction method based on finite measurement points.

[0043] The technical effects achieved by the device-type solution provided in the second aspect, the equipment-type solution provided in the third aspect, and the computer storage medium solution provided in the fourth aspect are the same as those achieved by the method-type solution provided in the first aspect, and will not be repeated here. Attached Figure Description

[0044] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0045] Figure 1 This is a schematic diagram of the physical field reconstruction method based on finite measurement points provided by the present invention;

[0046] Figure 2 A schematic diagram of the multi-fidelity physical field reconstruction technology based on finite measurement points provided by the present invention;

[0047] Figure 3 A schematic diagram of the physical field reconstruction device based on finite measurement points provided by the present invention;

[0048] Figure 4 A schematic diagram of the physical field reconstruction device based on finite measurement points provided by the present invention. Detailed Implementation

[0049] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0050] It should be noted that in this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0051] In this invention, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, a combination of a and b, a combination of a and c, a combination of b and c, or a, b, and c, where a, b, and c can be single or multiple.

[0052] In the process of constructing a monitoring system, reconstructing the system's physical field from a small number of discrete measuring points is an important research topic. In the research of structural field reconstruction methods, the mainstream methods for structural deformation reconstruction include modal analysis, Ko displacement theory, and the inverse finite element method. Experiments have shown that all three methods have high reconstruction accuracy, and the inverse finite element method is more effective and accurate, making it one of the most promising methods currently available.

[0053] In a digital twin test system for liquid rocket engines, the digital twin model needs to be synchronized with the engine test using sensors at measurement points to reflect the engine's physical field information in real time. Traditionally, this involves deploying densely packed measurement points and constructing a point cloud physical field from the data from these sensors. The accuracy of this type of physical field reconstruction depends on the data collected from the measurement points, and its effectiveness hinges on the number and location of these points. However, in actual rocket engines with complex structures (such as those inside capillary tubes) and harsh test environments (such as the high-temperature environment of the thrust chamber), it is difficult to deploy too many sensor measurement points. Furthermore, the location of these measurement points usually does not affect engine operation, thus impacting their effectiveness.

[0054] To address the aforementioned shortcomings in existing technologies, this invention researches multi-fidelity physical field reconstruction technology based on finite measurement points. It reconstructs the physical field from finite measurement point data, combines it with physical field simulation data, and utilizes a multi-fidelity fusion method to achieve multi-fidelity physical field reconstruction. This solves problems such as insufficient test bench space and limited sensor layout, laying the foundation for real-time physical field sensing by sensor models. The following description, in conjunction with the accompanying drawings, illustrates the solutions provided in the embodiments of this specification:

[0055] like Figure 1 As shown, the process may include the following steps:

[0056] Step 110: During the physical test of the liquid rocket engine, acquire multi-source heterogeneous sensing data collected by multi-source heterogeneous sensors.

[0057] Multi-source heterogeneous sensors can include at least a test bench strain acquisition system, inertial navigation sensors, pressure sensors, fiber Bragg grating sensors, displacement sensors, and other sensors. The data acquired by these sensors is multi-source heterogeneous sensing data.

[0058] Step 120: Select a limited number of sensor data corresponding to measurement points from the multi-source heterogeneous sensor data and preprocess them to obtain target sensor data; the preprocessing includes at least data cleaning, noise reduction and format unification processing.

[0059] Data cleaning, noise reduction, and format standardization are performed to filter and remove data that does not meet preset conditions in order to obtain valid data. Data that does not meet preset conditions includes at least incomplete data, erroneous data, and duplicate data.

[0060] The finite number of measurement points can be those among the measurement points of multi-source heterogeneous sensors whose importance exceeds a preset threshold. The importance can be determined based on the sensor's installation location, the type of object being tested, and user-preset conditions (e.g., operating boundary conditions).

[0061] Step 130: Using a regression machine learning algorithm and combining the simulation boundary conditions of the actual working conditions of the test bench, construct a physical correction and reconstruction model based on a finite number of measurement points.

[0062] Regression machine learning algorithms can include Kriging algorithms, radial basis function neural networks, or polynomial response surface methods, etc.

[0063] Simulation boundary conditions can include inlet boundary conditions (e.g., setting parameters such as propellant flow rate, pressure, and temperature), wall boundary conditions (e.g., heat conduction and radiation of the wall, including material properties, thermal conductivity, and emissivity), outlet boundary conditions (e.g., the flow state of the combustion gas, such as velocity, pressure, and temperature), combustion chamber internal conditions (e.g., temperature, pressure, and component distribution within the combustion chamber; chemical reaction kinetics within the combustion chamber, including reaction rate, reactant and product concentrations; flow state within the combustion chamber, such as turbulence intensity and vortex structure), and rotor and turbopump boundary conditions (e.g., rotor speed and torque; turbopump operating conditions, such as pump efficiency and head).

[0064] Step 140: Perform simulation analysis on the actual working conditions of the test bench to generate simulated physical field data; the simulated physical field data is low-fidelity data.

[0065] In the field of liquid rocket engines, low-fidelity data generally refers to data that is less precise, less detailed, or simplified, in contrast to high-precision, highly detailed data. This data may originate from various sources, including but not limited to simplified physical models, empirical formulas, rough estimates, preliminary processing of experimental data, or low-resolution measurements.

[0066] Step 150: Based on data fusion technology, establish the mapping relationship between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data, and construct a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies.

[0067] Figure 1 The method described herein involves acquiring multi-source heterogeneous sensor data from multiple heterogeneous sensors during the physical test of a liquid rocket engine; preprocessing the sensor data corresponding to a limited number of measurement points from the multi-source heterogeneous sensor data to obtain target sensor data; constructing a physical correction and reconstruction model based on a limited number of measurement points using a regression machine learning algorithm combined with the simulation boundary conditions of the actual test bench; performing simulation analysis on the actual test bench conditions to generate simulated physical field data; and establishing a mapping relationship between the target sensor data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data based on data fusion technology, and constructing a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies. This invention reconstructs the physical field from limited measurement point data, combines it with physical field simulation data, and utilizes a multi-fidelity fusion method to achieve multi-fidelity physical field reconstruction, solving problems such as insufficient test bench space and limited sensor layout, improving the accuracy of physical field reconstruction, and laying the foundation for real-time perception of the physical field by sensor models.

[0068] based on Figure 1In addition to the method described herein, this specification also provides some specific implementation methods of this method, which will be described below.

[0069] Step 130, in its specific implementation, utilizes a regression machine learning algorithm, combined with the simulation boundary conditions of the actual working conditions of the test bench, to construct a physical correction and reconstruction model based on a finite number of measurement points. Specifically, this may include:

[0070] By utilizing Kriging, radial basis function neural networks, or polynomial response surfaces, and combining the simulated boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on finite measurement points is constructed.

[0071] Based on the physical correction and reconstruction model, reconstructed physical field data at a finite number of measurement points are obtained;

[0072] A physical correction and reconstruction model based on a finite number of measurement points is constructed. This model combines Kriging, radial basis function neural network (RBF-NN) or polynomial response surface (PRS) methods, and simulates the boundary conditions of the actual test bench. The specific implementation methods are as follows:

[0073] 1) Data Collection: Obtain measurement data under actual working conditions from the test bench, including physical quantity values ​​at a limited number of measuring points and corresponding simulation boundary conditions. Ensure the accuracy and completeness of the data.

[0074] 2) Data preprocessing: Clean the data to remove outliers and noise. Normalize or standardize the data to improve the model's generalization ability.

[0075] Of the three methods—Kriging, Radial Basis Function Neural Networks (RBF-NNs), and Polynomial Response Surfaces (PRS)—Kriging is suitable for interpolation and prediction of spatially distributed data. It requires calculating a semi-variogram (or covariance function) to describe the spatial correlation between data points. Specifically, a Kriging equation can be constructed, and the predicted values ​​for unknown points can be obtained by solving the equation. RBF-NNs are suitable for nonlinear mappings and function approximation. Appropriate radial basis functions (such as Gaussian functions, multiple quadratic functions, etc.) can be selected, the neural network trained, and the prediction error minimized by adjusting the weights and biases. PRS is suitable for modeling linear or low-order nonlinear relationships. Specifically, an appropriate polynomial order can be selected, and the polynomial coefficients can be fitted using least squares or other optimization methods.

[0076] When constructing the corresponding model, physical quantities and simulation boundary conditions at a finite number of measurement points can be input. For the Kriging method, the semi-variogram function can be calculated and the Kriging equation can be constructed. For the RBF-NN method, an appropriate radial basis function can be selected and the neural network can be trained. For the PRS method, polynomial coefficients can be fitted.

[0077] After the model is built, it can be validated using a validation dataset. The prediction error (such as mean square error, root mean square error, etc.) can be calculated to evaluate the model's performance. Based on the validation results, the model parameters can be adjusted or a more suitable model can be selected.

[0078] As for the physical correction and reconstruction model, the physical quantities of a limited number of measurement points and the simulation boundary conditions can be input into the trained model to obtain the predicted physical quantities of unknown points. Then, the predicted values ​​are corrected according to actual physical laws and the characteristics of the test bench. The reconstruction results are compared with the actual measured values ​​on the test bench to evaluate the accuracy and reliability of the reconstruction results, and the model is optimized based on the evaluation results.

[0079] By following the steps above, a physical correction and reconstruction model based on a finite number of measurement points can be constructed and applied in conjunction with the simulation boundary conditions of the actual working conditions of the test bench. In practice, it is necessary to select appropriate models and parameters according to specific circumstances and conduct thorough verification and optimization.

[0080] Next, we will use the implementation of the Kriging model as an example to illustrate:

[0081] The Kriging model is an interpolation model based on statistical principles that can generate predicted values ​​from known data points. In finite element model (FEM) correction, the Kriging model is frequently used to address the problem of insufficient data. When the error of the FEM model is correlated with certain parameters, the Kriging model can model these parameters and generate more accurate predicted values ​​at unknown points.

[0082] Using the Kriging method and combining the simulation boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on a finite number of measurement points is constructed, which may specifically include:

[0083] Based on the actual working conditions of the test bench, determine the boundary conditions of the simulation model; the boundary conditions include at least temperature, pressure, and flow rate.

[0084] The Kriging model construction method is used to calculate predicted values ​​based on known measurement point data and determine the spatial correlation between data.

[0085] By using an iterative optimization method, the parameters of the Kriging model are continuously adjusted until the preset conditions are met, thus obtaining the target Kriging model.

[0086] Based on the differences between the simulation model and the actual working conditions, the physical parameters that need to be corrected are determined; the physical parameters include at least material properties, geometric structure, and boundary conditions.

[0087] Using the trained target Kriging model, correction values ​​for the parameters to be corrected are generated based on limited measurement point data;

[0088] The correction values ​​are applied to the simulation model to reconstruct the physical correction and reconstruction model.

[0089] More specifically, the boundary conditions of the simulation model are first determined based on the actual operating conditions of the test bench. These boundary conditions should reflect the actual operating state of the test bench as accurately as possible, including but not limited to physical parameters such as temperature, pressure, and flow rate. Then, limited measurement point data is acquired, which will serve as input to the Kriging model for model construction and training. When constructing the Kriging model, the Kriging model construction method can be used to generate predicted values ​​based on known measurement point data. This typically involves calculating a weighted average of known values ​​near the predicted points, considering spatial correlations between data points, and then continuously adjusting the parameters of the Kriging model through iterative optimization methods to improve the model's prediction accuracy. Next, the parameters to be corrected are determined: based on the differences between the simulation model and the actual operating conditions, the physical parameters that need correction are determined. These parameters may include material properties, geometry, boundary conditions, etc. Correction values ​​are generated: using the trained Kriging model, correction values ​​for the parameters to be corrected are generated based on the limited measurement point data. These correction values ​​will be used to adjust the corresponding parameters of the simulation model. Finally, the correction values ​​are applied to the simulation model to reconstruct the physically corrected and reconstructed model. This typically involves iteratively optimizing the simulation model to minimize the differences from actual operating conditions.

[0090] It's important to note that when choosing a Kriging model, the model's complexity should be determined based on the complexity of the actual problem and the amount of data. This ensures the model accurately reflects the inherent relationships between the data. Furthermore, when constructing a Kriging model, it's crucial to select appropriate model parameters, such as variance model and weighting functions. The choice of these parameters directly impacts the model's predictive accuracy and stability.

[0091] Step 150, in specific implementation, can be optionally implemented as follows: Construct a multi-fidelity fusion model based on data fusion technology, aiming to improve the prediction and reconstruction capabilities of target sensing data and its corresponding physical field by integrating data from different sources and with varying fidelities. Specifically, this may include:

[0092] The target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data are preprocessed (data cleaning: removing noise, outliers, and duplicate data; data alignment: ensuring consistency of data from different sources in time and space; data normalization or standardization: improving data consistency and comparability). Then, a mapping relationship is established, wherein the specific implementation of establishing the mapping relationship may include:

[0093] 1) Feature Extraction: Extract key features from target sensing data, medium-fidelity data, and simulated physical field data. These features can be direct measurements of physical parameters or data that has undergone transformation or dimensionality reduction.

[0094] 2) Establish a mapping model: Use machine learning or deep learning algorithms (such as regression models, neural networks, etc.) to establish the mapping relationship. The input includes target sensing data and mid-fidelity data, and the output is the simulated physical field data or the predicted value of the target sensing data. The mapping model can be a direct point-to-point mapping or a complex nonlinear relationship.

[0095] To optimize fusion weights and fusion strategies, we can first determine the fusion weights: assign different weights to data with different fidelity levels based on the accuracy and reliability of the data. Data-driven methods (such as Bayesian optimization, genetic algorithms, etc.) can be used to automatically determine the optimal weights. The fusion strategy can be a simple weighted average, or a more complex one such as weighted summation, minimizing the squared error of the weighted average, etc. Considering the spatiotemporal correlations of the data, we can design fusion strategies that capture these correlations.

[0096] When constructing a multi-fidelity fusion model, a mapping model and a fusion strategy can be integrated to form a multi-fidelity fusion model. This model can receive target sensing data and mid-fidelity data as input and output fused physical field data or predicted values ​​of target sensing data.

[0097] Through the above steps, a multi-fidelity fusion model based on data fusion technology can be established to achieve accurate mapping and prediction of the complex relationship between target sensing data, reconstructed physical field and simulated physical field data.

[0098] The technical solution of the present invention, such as Figure 2As shown, after reading data from a limited number of measurement points using multi-source heterogeneous sensor data processing and fusion technology, research was conducted on a method for reconstructing data from these limited measurement points. Kriging, radial basis function neural networks, and multinomial response surface regression machine learning methods were used, combined with the simulation boundary of actual working conditions, to develop a physical correction reconstruction model based on limited measurement points, achieving the acquisition of reconstructed physical field data under limited measurement points. Subsequently, simulation analysis was conducted on the actual working conditions of the test bench to obtain simulation data. A unified grid alignment was performed on the high-fidelity data from the limited measurement points, the medium-fidelity data from the reconstructed physical field, and the low-fidelity data from the simulated physical field, laying the foundation for multi-fidelity data fusion. Then, multi-fidelity fusion was performed using the three types of fidelity data, and advanced data fusion techniques such as bridge functions were combined to establish a mapping relationship between the three types of data, forming a multi-fidelity fusion model. Finally, to ensure that the reconstruction model can well reflect the test bench state under all working conditions, an objective function reflecting the displacement error of the reconstruction system under multiple working conditions was established. Optimization algorithms were applied to optimize the multi-fidelity model for key performance aspects such as accuracy and stability, maximizing reconstruction accuracy, minimizing information redundancy, and maximizing information density.

[0099] A digital twin model of a liquid rocket engine is constructed. During liquid rocket engine testing, the digital twin model is driven by a limited number of sensor data points. This model reflects the changes in the physical field of the liquid rocket engine in real time, providing more comprehensive, intuitive, in-depth, and accurate monitoring and early warning of engine testing. It also reduces sensor costs and the impact of sensors on engine testing. The physical field reconstruction technology using limited measurement points is entirely feasible for handling complex multi-condition problems in the test rig structure field. Previous research has applied this technology to the placement of mechanical sensors for the load-bearing structure during recovery. This enabled the selection of measurement points from complex structural information that can accurately and reliably perceive the mechanical performance of the load-bearing structure, meeting the requirement for precise reconstruction of the structural field with limited measurement point information. This provides data support for the calibration of simulation models and the training of real-time prediction models.

[0100] Following the same line of thought, this invention also provides a physical field reconstruction device based on a finite number of measurement points, such as... Figure 3 As shown, the device may include:

[0101] The multi-source heterogeneous sensor data acquisition module 310 is used to acquire multi-source heterogeneous sensor data collected by multi-source heterogeneous sensors during the physical test of a liquid rocket engine.

[0102] The target sensing data determination module 320 is used to select a limited number of sensing data corresponding to measurement points from the multi-source heterogeneous sensing data and perform preprocessing to obtain target sensing data; the preprocessing includes at least data cleaning, noise reduction and format unification processing;

[0103] The physical correction and reconstruction model building module 330 is used to construct a physical correction and reconstruction model based on a finite number of measurement points by using a regression machine learning algorithm and combining the simulation boundary conditions of the actual working conditions of the test bench.

[0104] The simulation physics field data generation module 340 is used to perform simulation analysis on the actual working conditions of the test bench and generate simulation physics field data; the simulation physics field data is low-fidelity data.

[0105] The multi-fidelity fusion model construction module 350 is used to establish a mapping relationship between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data based on data fusion technology, and to construct a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies.

[0106] based on Figure 3 The device may also include specific implementation units:

[0107] Optionally, the physical correction and reconstruction model building module 330 may specifically include:

[0108] The physical correction and reconstruction model building unit is used to construct a physical correction and reconstruction model based on finite measurement points by using Kriging, radial basis function neural networks or polynomial response surfaces, combined with the simulation boundary conditions of the actual working conditions of the test bench.

[0109] The reconstructed physical field data acquisition unit is used to acquire reconstructed physical field data at a limited number of measurement points based on the physical correction and reconstruction model.

[0110] Optionally, the physical correction and reconstruction model building blocks can be specifically used for:

[0111] Based on the actual working conditions of the test bench, determine the boundary conditions of the simulation model; the boundary conditions include at least temperature, pressure, and flow rate.

[0112] The Kriging model construction method is used to calculate predicted values ​​based on known measurement point data and determine the spatial correlation between data.

[0113] By using an iterative optimization method, the parameters of the Kriging model are continuously adjusted until the preset conditions are met, thus obtaining the target Kriging model.

[0114] Based on the differences between the simulation model and the actual working conditions, the physical parameters that need to be corrected are determined; the physical parameters include at least material properties, geometric structure, and boundary conditions.

[0115] Using the trained target Kriging model, correction values ​​for the parameters to be corrected are generated based on limited measurement point data;

[0116] The correction values ​​are applied to the simulation model to reconstruct the physical correction and reconstruction model.

[0117] Optionally, the multi-fidelity fusion model building module 350 may specifically include:

[0118] A grid data alignment unit is used to perform unified grid data alignment on the target sensing data, the reconstructed physical field data, and the simulated physical field data.

[0119] The multi-fidelity fusion model construction unit is used to perform multi-fidelity fusion of the aligned target sensing data, the reconstructed physical field data, and the simulated physical field data. Combined with data fusion technology, it establishes a mapping relationship between the aligned target sensing data, the reconstructed physical field data, and the simulated physical field data to form a multi-fidelity fusion model.

[0120] Optionally, the device also includes:

[0121] The objective function establishment module is used to establish an objective function that reflects the displacement error of the reconstructed system under multiple operating conditions;

[0122] The multi-fidelity model optimization module is used to perform multi-fidelity model optimization by applying optimization algorithms based on the objective function.

[0123] Optional, the multi-fidelity fusion model building module 350 may specifically include:

[0124] The physics model building unit is used to construct a three-dimensional geometric model of a liquid rocket engine, perform finite element simulation, and build a physics model.

[0125] The boundary condition determination element is used to take the measurement data collected from the physical test of the liquid rocket engine as the boundary conditions for finite element simulation and calculate the simulation results under different working conditions.

[0126] The multi-fidelity physics model building unit is used to train simulation results under different working conditions and construct multi-fidelity physics models.

[0127] Optionally, the device also includes:

[0128] A digital twin system building module for constructing digital twin systems for liquid rocket engines;

[0129] The engine test real-time monitoring module is used to monitor the engine test in real time by driving a digital twin model through a limited number of measurement points during the physical test of a liquid rocket engine.

[0130] Based on the same idea, embodiments of this specification also provide a physical field reconstruction device based on a finite number of measurement points. For example... Figure 4 As shown, the device includes:

[0131] The system includes a memory, a processor, and a communication interface coupled to the processor; the memory stores a computer program that can be run by the processor; when the processor runs the computer program, it executes the aforementioned physical field reconstruction method based on finite measurement points.

[0132] like Figure 4 As shown, the processor described above can be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present invention. The communication interface described above can be one or more. The communication interface can use any transceiver-like device for communicating with other devices or communication networks.

[0133] like Figure 4 As shown, the terminal device described above may also include a communication line. The communication line may include a path for transmitting information between the components described above.

[0134] Optional, such as Figure 4 As shown, the terminal device may further include a memory. The memory stores a computer program that can be executed by the processor; when the processor executes the computer program, it implements the method provided in the embodiments of the present invention.

[0135] like Figure 4 As shown, the memory can be read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed discs, laser discs, optical discs, digital universal discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited to these. The memory can exist independently and be connected to the processor via communication lines. The memory can also be integrated with the processor.

[0136] Optionally, the computer execution instructions in the embodiments of the present invention may also be referred to as application code, and the embodiments of the present invention do not specifically limit this.

[0137] In a specific implementation, as one example, such as Figure 4 As shown, a processor may include one or more CPUs, such as Figure 4 CPU0 and CPU1 in the CPU.

[0138] In a specific implementation, as one example, such as Figure 4 As shown, the terminal device may include multiple processors, such as Figure 4 The processors in the system. Each of these processors can be a single-core processor or a multi-core processor.

[0139] Based on the same idea, this specification also provides a computer storage medium corresponding to the above embodiments. The computer storage medium stores instructions that, when executed, implement the methods in the above embodiments.

[0140] The foregoing mainly describes the solutions provided by the embodiments of the present invention from the perspective of the interaction between various modules. It is understood that each module, in order to achieve the above functions, includes corresponding hardware structures and / or software units for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0141] The embodiments of the present invention can divide functional modules according to the above method examples. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in the embodiments of the present invention is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0142] The processor described in this specification may also function as a memory. The memory stores computer execution instructions for carrying out the present invention, and its execution is controlled by the processor. The processor executes the computer execution instructions stored in the memory, thereby implementing the method provided in the embodiments of the present invention.

[0143] The memory can be read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed discs, laser discs, optical discs, universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited to these. The memory can exist independently and be connected to the processor via communication lines. The memory can also be integrated with the processor.

[0144] Optionally, the computer execution instructions in the embodiments of the present invention may also be referred to as application code, and the embodiments of the present invention do not specifically limit this.

[0145] The methods disclosed in the above embodiments of the present invention can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above methods can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0146] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, the disclosure, and the appended claims in carrying out the claimed invention. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0147] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely exemplary descriptions of the invention as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A physical field reconstruction method based on finite measurement points, characterized in that, The methods include: During the physical test of a liquid rocket engine, multi-source heterogeneous sensing data were acquired from multi-source heterogeneous sensors. The target sensing data is obtained by preprocessing the sensing data corresponding to a limited number of measurement points selected from the multi-source heterogeneous sensing data. The preprocessing includes at least data cleaning, noise reduction, and format unification. By using regression machine learning algorithms and combining simulation boundary conditions of actual test bench conditions, a physical correction and reconstruction model based on finite measurement points is constructed. The actual working conditions of the test bench are simulated and analyzed to generate simulated physical field data; the simulated physical field data is low-fidelity data. Based on data fusion technology, a mapping relationship is established between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data. By optimizing the fusion weights and fusion strategies, a multi-fidelity fusion model is constructed.

2. The physical field reconstruction method based on finite measurement points according to claim 1, characterized in that, Using regression machine learning algorithms and combined with simulation boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on a finite number of measurement points is constructed, specifically including: By utilizing Kriging, radial basis function neural networks, or polynomial response surfaces, and combining the simulated boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on finite measurement points is constructed. Based on the physical correction and reconstruction model, reconstructed physical field data at a limited number of measurement points are obtained.

3. The physical field reconstruction method based on finite measurement points according to claim 2, characterized in that, Using the Kriging method and combining the simulation boundary conditions of the actual working conditions of the test bench, a physical correction and reconstruction model based on a finite number of measurement points is constructed, specifically including: Based on the actual working conditions of the test bench, determine the boundary conditions of the simulation model; the boundary conditions include at least temperature, pressure, and flow rate. The Kriging model construction method is used to calculate predicted values ​​based on known measurement point data and determine the spatial correlation between data. By using an iterative optimization method, the parameters of the Kriging model are continuously adjusted until the preset conditions are met, thus obtaining the target Kriging model. Based on the differences between the simulation model and the actual working conditions, the physical parameters that need to be corrected are determined; the physical parameters include at least material properties, geometric structure, and boundary conditions. Using the trained target Kriging model, correction values ​​for the parameters to be corrected are generated based on limited measurement point data; The correction values ​​are applied to the simulation model to reconstruct the physical correction and reconstruction model.

4. The physical field reconstruction method based on finite measurement points according to claim 2, characterized in that, The data fusion technology establishes a mapping relationship between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data. A multi-fidelity fusion model is constructed by optimizing the fusion weights and fusion strategy, specifically including: The target sensing data, the reconstructed physical field data, and the simulated physical field data are aligned using a unified grid data alignment. The aligned target sensing data, the reconstructed physical field data, and the simulated physical field data are fused with multi-fidelity. By combining data fusion technology, a mapping relationship is established between the aligned target sensing data, the reconstructed physical field data, and the simulated physical field data to form a multi-fidelity fusion model.

5. The physical field reconstruction method based on finite measurement points according to claim 1, characterized in that, After constructing a multi-fidelity fusion model by optimizing the fusion weights and fusion strategy, the following steps are also included: Establish an objective function that reflects the displacement error of the reconfiguration system under multiple operating conditions; Based on the objective function, an optimization algorithm is applied to optimize the multi-fidelity model.

6. The physical field reconstruction method based on finite measurement points according to claim 1, characterized in that, Constructing a multi-fidelity fusion model specifically includes: Construct a three-dimensional geometric model of the liquid rocket engine, perform finite element method simulation, and build a physical field model; The measurement data collected from the physical test of the liquid rocket engine were used as the boundary conditions for the finite element simulation to calculate the simulation results under different operating conditions. The simulation results under different working conditions are used to train and construct a multi-fidelity physical field model.

7. The physical field reconstruction method based on finite measurement points according to claim 6, characterized in that, After training the simulation results under different operating conditions to construct a multi-fidelity physics model, the following steps are also included: Constructing a digital twin system for liquid rocket engines; During the physical testing of a liquid rocket engine, a digital twin model is driven by a limited number of measurement points to monitor the engine test in real time.

8. A physical field reconstruction device based on a finite number of measurement points, characterized in that, The device includes: The multi-source heterogeneous sensor data acquisition module is used to acquire multi-source heterogeneous sensor data collected by multi-source heterogeneous sensors during the physical test of a liquid rocket engine. The target sensing data determination module is used to select a limited number of sensing data corresponding to measurement points from the multi-source heterogeneous sensing data for preprocessing to obtain target sensing data. The preprocessing includes at least data cleaning, noise reduction, and format unification. The physical correction and reconstruction model building module is used to construct a physical correction and reconstruction model based on a finite number of measurement points by using regression machine learning algorithms and combining the simulation boundary conditions of the actual working conditions of the test bench. The simulated physical field data generation module is used to perform simulation analysis on the actual working conditions of the test bench and generate simulated physical field data; the simulated physical field data is low-fidelity data. The multi-fidelity fusion model construction module is used to establish the mapping relationship between the target sensing data, the mid-fidelity data corresponding to the reconstructed physical field, and the simulated physical field data based on data fusion technology, and to construct a multi-fidelity fusion model by optimizing the fusion weights and fusion strategies.

9. A physical field reconstruction device based on finite measurement points, characterized in that the device... include: Memory, processor, and communication interface coupled to the processor; The memory stores computer programs that can be executed by the processor; When the processor runs the computer program, it executes the physical field reconstruction method based on finite measurement points as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The computer storage medium stores instructions that, when executed, implement the physical field reconstruction method based on finite measurement points as described in any one of claims 1 to 7.

Images