Rocket nozzle flow-thermal coupling fast prediction method and system

Abstract

The application discloses a rocket nozzle flow-heat-solid coupling rapid prediction method and system. The method adopts a deep learning reduced-order model to simulate the flow and heat transfer process of high-temperature and high-pressure gas in the nozzle, and rapidly reconstructs the flow field and temperature field; a physically driven reduced-order model is used to solve the heat conduction and deformation of the nozzle structure, and through the bidirectional transmission of the interfacial force-heat parameters and the structural deformation, the flow-heat-solid coupling process is realized; the system uses a database extraction module, a deep learning reduced-order module, a flow characteristic prediction module, a physically driven reduced-order module and a flow-heat-solid coupling prediction module to complete the rapid prediction of the flow-heat-solid coupling. The application significantly improves the calculation efficiency of the traditional full-order simulation method, can support the nozzle geometric parameter scanning, cooling scheme evaluation and material selection, significantly shortens the research and development cycle, and can be widely applied to the field of rocket engine nozzle design and performance evaluation.

Description

Technical Field

[0001] This invention belongs to the field of aerospace engineering performance evaluation technology, and in particular relates to a rapid prediction method and system for rocket nozzle flow thermo-mechanical coupling. Background Technology

[0002] The nozzle is the core component of a rocket engine's energy conversion process. The high-temperature, high-pressure, and high-speed gas jet generated by combustion expands and accelerates through the nozzle to provide propulsion for the rocket. During service, intense convective and radiative heat transfer occurs between the gas jet and the nozzle wall, causing a sharp rise in wall temperature. This high temperature leads to significant thermal deformation and thermal stress in the nozzle, which in turn affects the gas flow state and heat transfer characteristics, forming a strongly coupled physical process involving flow, heat transfer, and structure. Accurately predicting the internal fluid flow, heat transfer processes, and the resulting structural response is crucial for rocket engine nozzle configuration design, cooling scheme formulation, and material selection strategies.

[0003] Currently, high-precision simulations of nozzle flow thermo-mechanical coupling analysis primarily rely on numerical simulation methods, such as fully coupled or weakly coupled simulations based on Computational Fluid Dynamics (CFD) and the Finite Element Method (FEM). While these methods theoretically provide high-fidelity details of multiphysics distributions, their high computational cost limits their widespread application in engineering design. Capturing complex turbulent and thin boundary layer flows requires extremely fine meshes and very small time steps; a single complete coupled simulation can take days or even weeks to complete on a high-performance computer. Furthermore, the design phase requires extensive parameter scanning (such as geometric parameters, inlet conditions, and material properties) or optimization iterations, making traditional high-fidelity numerical simulation methods virtually infeasible.

[0004] A promising technique for improving the computational efficiency of fluid-thermal-structure interaction (FTH) simulations is model reduction. Model reduction is a relative term; a reduced-order model is a low-dimensional approximation built upon a high-fidelity full-order model, achieved by projecting high-dimensional physical data onto a low-dimensional space. For large-scale discrete systems with tens of thousands of degrees of freedom, the projection onto the subspace only has tens or even less than ten degrees of freedom, significantly improving the efficiency of numerical simulations while ensuring accuracy in solving complex physical problems. Current mainstream reduction models are mainly divided into data-driven reduction (deep learning reduction models) and physics-driven reduction models (projective reduction models). Data-driven reduction models combine deep learning and principal component analysis methods, using offline machine learning and online modal superposition to quickly reconstruct complex three-dimensional physical field information. However, data-driven reduction models still rely on data sample acquisition, making extrapolation of operating conditions from training data unreliable. Physics-driven order reduction models are developed from traditional numerical methods. Unlike black-box models, they solve the same physical governing equations as traditional numerical methods, but achieve dimensionality reduction of physical variables through Galerkin projection, significantly improving the computational efficiency of traditional numerical methods for solving large-scale discrete equations. However, physics-driven order reduction models suffer from high invasive development costs and low nonlinearity solution efficiency when dealing with strongly coupled multi-physics and nonlinear problems in rocket engine nozzles. Therefore, the technical problem of efficiently and accurately predicting the fluid-thermal-structure coupling characteristics during rocket engine nozzle service remains unsolved. Summary of the Invention

[0005] The purpose of this invention is to provide a rapid prediction method and system for the thermo-structure coupling of rocket nozzle flow, thereby improving the accuracy and efficiency of nozzle flow characteristics, heat exchange, and structural strength analysis.

[0006] To achieve the objective of this invention, on one hand, this invention provides a rapid prediction method for thermo-mechanical coupling of rocket nozzle flow, comprising the following steps:

[0007] Step 1: Set the range of incoming gas flow conditions in the nozzle, generate parameterized chemical condition samples, establish a high-fidelity computational fluid dynamics model of nozzle flow and heat transfer, and perform numerical simulations on each parameterized chemical condition to obtain a database of gas flow and heat transfer characteristics in the nozzle, and then extract the dominant mode and the projection coefficients corresponding to each parameterized chemical condition.

[0008] Step 2: Build a deep learning model for reducing the order of the gas parameters in the parameterized chemical conditions and train it on the database in Step 1. The input of the deep learning model for reducing the order of the gas parameters in the parameterized chemical conditions is the gas parameters in the parameterized chemical conditions and the output is the projection coefficients corresponding to each parameterized chemical condition.

[0009] Step 3: Based on the deep learning reduced-order model trained in Step 2, predict the projection coefficients of flow characteristics under different operating conditions for gas conditions other than parameterized operating conditions. Then, combine the dominant mode to obtain the nozzle wall temperature vector to achieve the prediction of flow characteristics.

[0010] Step 4: Establish a finite element model of the force-thermal coupling of the nozzle structure, extract the temperature-related heat conduction matrix through numerical integration, extract the discrete interpolation nodes and their nearest neighbor elements from the temperature-related heat conduction matrix, and then construct a physical-driven reduced-order model of the force-thermal coupling of the nozzle.

[0011] Step 5: Use the nozzle wall temperature vector as the boundary condition of the physically driven nozzle force-thermal coupling reduced-order model to complete the fluid-thermal-structure coupling prediction.

[0012] On the other hand, the present invention also provides a system for a rapid prediction method of thermo-mechanical coupling of rocket nozzle flow, comprising the following modules:

[0013] The database extraction module is used to obtain a database of gas flow and heat transfer characteristics inside the nozzle, and then extract the dominant modes and the projection coefficients corresponding to the various parameter conditions.

[0014] The deep learning reduction module is used to build and train deep learning reduction models.

[0015] The flow characteristic prediction module is used to predict the projection coefficient of flow characteristics under different operating conditions for gas conditions other than the parameterized operating conditions, and then combine it with the dominant mode to obtain the nozzle wall temperature vector.

[0016] The physics-driven order reduction module is used to extract discrete interpolation nodes and their nearest neighbor units using a temperature-related heat conduction matrix to construct a physics-driven order reduction model.

[0017] The fluid-thermal-structure coupling prediction module is used to use the nozzle wall temperature vector as the boundary condition of the physical drive reduction module to complete the prediction of fluid-thermal-structure coupling.

[0018] Compared with the prior art, the significant progress of the present invention is as follows: (1) The present invention uses a deep learning reduced-order model to simulate the flow and heat transfer of high-temperature and high-pressure gas jets in the nozzle flow field, and uses a physical-driven reduced-order model to solve the heat conduction and deformation of the nozzle structure. The two are coupled through the transfer of interface force and heat parameters, thereby completing the rapid prediction of the thermo-mechanical coupling behavior of the nozzle flow. Combining the advantages of both, the deep learning reduced-order model has the ability to quickly reconstruct high-dimensional nonlinear flow fields, while retaining the high-precision prediction advantage of the physical-driven reduced-order model for structural mechanical behavior. Tight coupling is achieved through the transfer of interface force and heat parameters, solving the problems of data dependence, unreliable extrapolation and high cost of invasive development of a single reduced-order model; (2) The present invention can compress the prediction time of the thermo-mechanical coupling of the nozzle flow to minutes or even seconds, and ensure the prediction accuracy of key physical field parameters, providing an efficient theoretical tool for the design of rocket engine nozzle configuration, the formulation of cooling schemes and the selection of materials.

[0019] To more clearly illustrate the functional characteristics and structural parameters of the present invention, further explanation is provided below in conjunction with the accompanying drawings and specific embodiments. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0021] Figure 1 This is a flowchart of the nozzle flow thermo-mechanical coupling characteristic prediction for the fusion data physics reduction model of the present invention.

[0022] Figure 2 This is a schematic diagram of the process for obtaining the dominant mode of the nozzle flow field according to the present invention;

[0023] Figure 3 This is a schematic diagram of the deep learning reduced-order model of the present invention predicting the flow characteristics of the flow field;

[0024] Figure 4 This is a schematic diagram of the physical-driven order reduction model construction process of the present invention;

[0025] Figure 5 This is a comparison chart of the nozzle temperature field prediction results of the CFD+FEM model and the reduced-order model of this invention. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This invention provides a rapid prediction method for rocket nozzle flow thermo-mechanical coupling, combining... Figure 1 This includes the following steps:

[0028] Step 1: Analyze parametric simulation and modal analysis: combining Figure 2 The incoming flow conditions of the nozzle gas are set, including pressure and temperature, with ranges of [6MPa, 10MPa] and [3000K, 3500K], respectively. Parametric operating conditions are established using Latin hypercube sampling, and 60 sets of pressure and temperature parameters are sampled to ensure the validity of the sample space. A high-fidelity computational fluid dynamics model of nozzle flow and heat transfer is established, and numerical simulation of nozzle flow characteristics under parametric operating conditions is performed to obtain a database of gas flow characteristics in the nozzle. Then, the dominant modes of gas flow characteristics in the nozzle and the temperature field projection coefficients corresponding to each sample operating condition are extracted using the intrinsic orthogonal decomposition method.

[0029] The dominant mode of the extracted gas flow characteristics inside the nozzle and the temperature field projection coefficients under each sample operating condition As shown in the following formula:

[0030] ;

[0031] ;

[0032] ;

[0033] Where A is the nozzle wall temperature database matrix, It is the transpose of the nozzle wall temperature database matrix. and These are matrices A T The i-th eigenvalue and eigenvector of A;

[0034] The first r dominant modes of the gas flow characteristics inside the nozzle are obtained by analyzing the characteristic energy proportions of each mode. The selection criteria are shown in the following formula:

[0035] ;

[0036] Where j is the j-th order eigenvalue; r is the number of the first r order eigenvalues; n is the total number of eigenvalues; and ε is the eigenenergy percentage, usually taken as 99.9%.

[0037] Step 2: Establish a deep learning model for order reduction: combining Figure 3 A four-layer deep learning model for order reduction is constructed, with the input layer consisting of gas temperature and gas pressure, and the output layer consisting of the temperature field projection coefficients. The deep learning model for order reduction is trained by fitting the nonlinear mapping relationship between the gas parameters and the dominant mode coefficients, as shown in the following formula:

[0038] ;

[0039] Here, w and b are the weights and biases that need to be determined during neural network training. It is the weight corresponding to the k-th gas parameter, c k It is the kth gas parameter (inlet gas pressure and temperature). It is the Sigmoid function, the temperature field projection coefficient In These are the r-th order projection coefficients.

[0040] Step 3: Predict different flow characteristics of the nozzle flow field using the modal superposition method: Based on the deep learning reduced-order model trained in Step 2, the projection coefficient of the temperature field under the gas combustion conditions (3500K, 6.9MPa) not included in the parameterized chemical conditions established in Step 1 is selected for prediction. Then, combined with the dominant mode of the flow characteristics in Step 1, the nozzle wall temperature vector T is obtained. s The flow characteristics can be predicted using the following formula:

[0041] ;

[0042] in, It is a matrix composed of the first r dominant modes of the gas flow characteristics inside the nozzle.

[0043] Step 4: Establish a physics-driven order reduction model:

[0044] Step 4-1, Combining Figure 4 A coupled heat transfer finite element model of the nozzle structure is established, and the governing equations of the finite element model are shown below:

[0045] ;

[0046] Where [T] is the nodal temperature vector, [K(T)] is the temperature-dependent heat conduction matrix, and [P] is the heat load vector;

[0047] According to the finite element method (FEM) calculation format, the temperature-related heat conduction matrix [K(T)] and the heat load vector [P] are respectively assembled by element integration, as shown in the following equation:

[0048] ;

[0049] ;

[0050] ;

[0051] in, It is the element heat conduction matrix of the i-th element; is the element thermal load vector of the i-th element; l is the number of all finite element discrete elements; m is the number of elements on the surface to which the thermal load is applied; The temperature is T e Thermal conductivity coefficient of the unit cell; N T Ω is the transpose of the element shape function N; h is the three-dimensional solid region; h is the convective heat transfer coefficient. It is the surface to which the thermal load is applied; T ∞ It is the ambient temperature for convective heat transfer;

[0052] Step 4-2: By performing feature analysis on the temperature-related heat conduction matrix [K(T)], the discrete interpolation nodes ω of the heat conduction matrix are extracted using the discrete empirical interpolation method. i The calculation formula is as follows:

[0053] ;

[0054] ;

[0055] ;

[0056] in, It is a transformation matrix composed of r unit vectors. It is the first A unit vector with 1 element and 0 elements; It is the modal matrix of the finite element heat conduction matrix. The i-th mode vector, fi is the i-th interpolation coefficient vector; r i It is the i-th interpolation error vector; max is the operator that takes the maximum value and position of the vector. It is the maximum value of the interpolation error vector;

[0057] Step 4-3: Construct a physics-driven force-thermal coupling reduced-order model of the nozzle to calculate the three-dimensional temperature field of the nozzle structure:

[0058] The temperature-dependent heat conduction matrix is ​​obtained by selecting its nearest neighbor cells through discrete empirical interpolation to obtain a new heat conduction matrix. ;

[0059] The new heat conduction matrix The calculation formula is shown below:

[0060] ;

[0061] in, The heat conduction matrix is ​​obtained through discrete empirical interpolation. It is the mode matrix The reduction matrix projected at the interpolation location The inverse matrix; m1 is the number of nearest neighbors of the first interpolation node; Ω 1 It is the region of the nearest neighbor cells of the first interpolation node; m r Ω is the number of nearest neighbor units of the r-th interpolation node; r q is the region of the nearest neighbor cells of the r-th interpolation node; g(q) is the function for calculating the element heat conduction matrix; It is the differential symbol;

[0062] The three-dimensional temperature field of the nozzle structure is obtained by using the Newton-Raphson algorithm in combination with the new heat conduction matrix, as shown in the following equation:

[0063] ;

[0064] in, It is the nozzle temperature vector at the (k+1)th iteration step; γ is the nozzle temperature vector at the k-th iteration step; γ is the convergence factor.

[0065] Step 5: Convert the nozzle wall temperature vector T s As the boundary conditions of the reduced-order model of the nozzle force-thermal coupling driven by the physics, a rapid prediction of fluid-thermal-structure interaction is achieved, combined with... Figure 5 The temperature distribution characteristics of the nozzle were simulated using a CFD+FEM model and a fused data physics reduction model, respectively. The CPU computation time for the two models was 412 s and 12 s, respectively. This shows that the method can compress the prediction time of the nozzle flow thermo-mechanical coupling to the second level, which significantly improves the simulation efficiency of the nozzle flow thermo-mechanical coupling.

[0066] The system of a rapid prediction method for thermo-mechanical coupling of rocket nozzle flow according to the present invention includes the following modules:

[0067] The database extraction module is used to obtain a database of gas flow and heat transfer characteristics inside the nozzle, and then extract the dominant modes and the projection coefficients corresponding to the various parameter conditions.

[0068] The deep learning reduction module is used to build and train deep learning reduction models.

[0069] The flow characteristic prediction module is used to predict the projection coefficient of flow characteristics under different operating conditions for gas conditions other than the parameterized operating conditions, and then combine it with the dominant mode to obtain the nozzle wall temperature vector.

[0070] The physics-driven order reduction module is used to extract discrete interpolation nodes and their nearest neighbor units using a temperature-related heat conduction matrix to construct a physics-driven order reduction model.

[0071] The fluid-thermal-structure coupling prediction module is used to use the nozzle wall temperature vector as the boundary condition of the physical drive reduction module to complete the prediction of fluid-thermal-structure coupling.

[0072] The execution steps of each module are the same as those of the method described above, and will not be repeated here.

[0073] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0074] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A rapid prediction method for rocket nozzle flow thermo-mechanical coupling, characterized in that, Includes the following steps: Step 1: Set the range of incoming gas flow conditions in the nozzle, generate parameterized chemical condition samples, establish a high-fidelity computational fluid dynamics model of nozzle flow and heat transfer, and perform numerical simulations on each parameterized chemical condition to obtain a database of gas flow and heat transfer characteristics in the nozzle, and then extract the dominant mode and the projection coefficients corresponding to each parameterized chemical condition. Step 2: Build a deep learning model for reducing the order of the gas parameters in the parameterized chemical conditions and train it on the database in Step 1. The input of the deep learning model for reducing the order of the gas parameters in the parameterized chemical conditions is the gas parameters in the parameterized chemical conditions and the output is the projection coefficients corresponding to each parameterized chemical condition. Step 3: Based on the deep learning reduced-order model trained in Step 2, predict the projection coefficients of flow characteristics under different operating conditions for gas conditions other than parameterized operating conditions. Then, combine the dominant mode to obtain the nozzle wall temperature vector to achieve the prediction of flow characteristics. Step 4: Establish a finite element model of the force-thermal coupling of the nozzle structure, extract the temperature-related heat conduction matrix through numerical integration, extract the discrete interpolation nodes and their nearest neighbor elements from the temperature-related heat conduction matrix, and then construct a physical-driven reduced-order model of the force-thermal coupling of the nozzle. Step 4-1: Establish a coupled heat transfer finite element model of the nozzle structure to obtain the temperature-related heat conduction matrix; Step 4-2: Perform feature analysis on the temperature-related heat conduction matrix, and then use the discrete empirical interpolation method to extract the discrete interpolation nodes of the heat conduction matrix; Step 4-3: The temperature-related heat conduction matrix is ​​used to construct a physically driven force-thermal coupling reduced-order model of the nozzle through the discrete interpolation nodes, and the three-dimensional temperature field of the nozzle structure is calculated. The temperature-related heat conduction matrix [K(T)] is obtained through the discrete interpolation node ω. i By selecting its nearest neighbor cells, a new heat conduction matrix is ​​obtained. ; The new heat conduction matrix The calculation formula is shown below: ； in, The heat conduction matrix is ​​obtained through discrete empirical interpolation. It is the mode matrix The reduction matrix projected at the interpolation location The inverse matrix; m1 is the number of nearest neighbors of the first interpolation node; Ω 1 It is the region of the nearest neighbor cells of the first interpolation node; m r Ω is the number of nearest neighbor units of the r-th interpolation node; r q is the region of the nearest neighbor cells of the r-th interpolation node; g(q) is the function for calculating the element heat conduction matrix; It is the differential symbol; Using the Newton-Raphson algorithm combined with the new heat conduction matrix The three-dimensional temperature field of the nozzle structure is obtained by solving the equation below: ； in, It is the nozzle temperature vector at the (k+1)th iteration step; γ is the nozzle temperature vector at the k-th iteration step; γ is the convergence factor. Step 5: Use the nozzle wall temperature vector as the boundary condition of the physically driven nozzle force-thermal coupling reduced-order model to complete the fluid-thermal-structure coupling prediction.

2. The rapid prediction method for rocket nozzle flow thermo-mechanical coupling according to claim 1, characterized in that, Step 1 specifically involves: The incoming flow conditions of the nozzle exhaust gas are set, including pressure and temperature. Parametric operating conditions are established using Latin hypercube sampling, and the pressure and temperature parameters are sampled. A high-fidelity computational fluid dynamics model of nozzle flow and heat transfer is established, and numerical simulations of the nozzle flow characteristics under parametric operating conditions are performed to obtain a database of exhaust gas flow characteristics within the nozzle. Then, the dominant modes of exhaust gas flow characteristics within the nozzle and the temperature field projection coefficients corresponding to each sample operating condition are extracted using the intrinsic orthogonal decomposition method. The extracted dominant modes of exhaust gas flow characteristics within the nozzle... and the temperature field projection coefficients under each sample operating condition As shown in the following formula: ； ； ； Where A is the nozzle wall temperature database matrix, It is the transpose of the nozzle wall temperature database matrix. and These are matrices A T The i-th eigenvalue and eigenvector of A; The first r dominant modes of the gas flow characteristics inside the nozzle are obtained by analyzing the characteristic energy proportions of each mode. The selection criteria are shown in the following formula: ； Where j is the j-th order eigenvalue; r is the number of the first r order eigenvalues; n is the total number of eigenvalues; and ε is the proportion of eigenenergy.

3. The rapid prediction method for rocket nozzle flow thermo-mechanical coupling according to claim 2, characterized in that, The deep learning reduction model built in step 2 has the gas temperature and gas pressure as input layers and the temperature field projection coefficient as output layers. The deep learning model for order reduction is trained by fitting the nonlinear mapping relationship between the gas parameters and the dominant mode coefficients, as shown in the following formula: ； Here, w and b are the weights and biases that need to be determined during neural network training. It is the weight corresponding to the k-th gas parameter, c k It is the k-th gas parameter, the temperature field projection coefficient. In These are the r-th order projection coefficients.

4. The rapid prediction method for rocket nozzle flow thermo-mechanical coupling according to claim 3, characterized in that, The nozzle wall temperature vector T in step 3 s The calculation formula is shown below: ； in, It is a matrix composed of the first r dominant modes of the gas flow characteristics inside the nozzle.

5. The rapid prediction method for rocket nozzle flow thermo-mechanical coupling according to claim 4, characterized in that, Step 4-1 is described in detail below: The governing equations of the finite element model are shown below: ； Where [T] is the nodal temperature vector, [K(T)] is the temperature-dependent heat conduction matrix, and [P] is the heat load vector; The temperature-related heat conduction matrix [K(T)] and the heat load vector [P] are respectively assembled by element integration, as shown in the following equation: ； ； ； in, It is the element heat conduction matrix of the i-th element; It is the element thermal load vector of the i-th element; is the total number of all finite element discrete elements; m is the number of elements on the surface to which the thermal load is applied; The temperature is T e Thermal conductivity coefficient of the unit cell; N T Ω is the transpose of the element shape function N; h is the three-dimensional solid region; h is the convective heat transfer coefficient. It is the surface to which the thermal load is applied; T ∞ It is the ambient temperature for convective heat transfer.

6. The rapid prediction method for rocket nozzle flow thermo-mechanical coupling according to claim 5, characterized in that, The discrete interpolation nodes ω of the heat conduction matrix extracted in step 4-2 i As shown in the following formula: ； ； ； in, It is a transformation matrix composed of r unit vectors. It is the first A unit vector with 1 element and 0 elements; It is the modal matrix of the finite element heat conduction matrix. The i-th mode vector, fi is the i-th interpolation coefficient vector; r i It is the i-th interpolation error vector; max is the operator that takes the maximum value and position of the vector. It is the maximum value of the interpolation error vector.

7. A system for rapid prediction of rocket nozzle flow thermo-mechanical coupling according to any one of claims 1-6, characterized in that, Includes the following modules: The database extraction module is used to obtain a database of gas flow and heat transfer characteristics inside the nozzle, and then extract the dominant modes and the projection coefficients corresponding to the various parameter conditions. The deep learning reduction module is used to build and train deep learning reduction models. The flow characteristic prediction module is used to predict the projection coefficient of flow characteristics under different operating conditions for gas conditions other than the parameterized operating conditions, and then combine it with the dominant mode to obtain the nozzle wall temperature vector. The physics-driven order reduction module is used to extract discrete interpolation nodes and their nearest neighbor units using a temperature-related heat conduction matrix to construct a physics-driven order reduction model. The fluid-thermal-structure coupling prediction module is used to use the nozzle wall temperature vector as the boundary condition of the physical drive reduction module to complete the prediction of fluid-thermal-structure coupling.

Images