A multi-physics iterative coupling prediction method, system, terminal and medium

Abstract

The application belongs to the technical field of engineering simulation data processing, and specifically discloses a multi-physical field iterative coupling prediction method, system, terminal and medium. The method comprises the following steps: based on a prediction batch input, initializing the state of a to-be-predicted object, and constructing a structured feature representation for multi-physical field prediction; constructing a temperature prediction model and a pressure prediction model which are independent of each other; performing multi-round prediction update on the temperature physical field by using the temperature prediction model, and dynamically adjusting relevant state parameters according to the predicted state of the temperature physical field in the prediction update process to obtain a stable temperature physical field prediction state; on this basis, performing prediction processing on the pressure physical field by using the pressure prediction model to obtain a pressure physical field prediction result corresponding to the temperature physical field prediction state, and outputting a multi-physical field steady-state prediction result. By introducing the multi-physical field iterative coupling prediction mechanism, the physical consistency and stability of the multi-physical field prediction result are improved.

Description

Technical Field

[0001] This invention belongs to the field of engineering simulation data processing technology, specifically relating to a multiphysics iterative coupling prediction method, system, terminal, and medium. Background Technology

[0002] As power equipment, energy equipment, and complex industrial systems develop towards higher power density and higher integration, the impact of multi-physics coupling effects on equipment operation is becoming increasingly significant. Taking power equipment such as transformers as an example, multiple physical fields, such as temperature and flow pressure fields, typically exist simultaneously within them. These different physical fields couple and constrain each other, directly affecting the operational safety, reliability, and lifespan of the equipment. Therefore, accurately predicting the distribution of multi-physics fields has become an important technical requirement in equipment design optimization, operational evaluation, and condition monitoring.

[0003] Existing multiphysics analysis methods primarily rely on numerical simulations based on finite element methods or computational fluid dynamics to calculate physical quantities such as temperature and pressure fields by solving governing equations. While these methods offer high physical accuracy, they are typically computationally complex, involve numerous iterations, and are computationally expensive, making them unsuitable for rapid prediction or multi-condition assessment. In recent years, with the development of machine learning technology, the use of data-driven models for multiphysics prediction has gained increasing attention. Among these methods, graph-based prediction methods can better adapt to irregular structures such as finite element meshes, thus improving prediction efficiency to some extent.

[0004] However, existing data-driven multiphysics prediction methods still have significant shortcomings. On the one hand, some methods treat different physical fields as independent prediction targets or only employ simple parallel prediction methods, failing to effectively characterize the coupling relationship between the temperature and pressure fields, resulting in insufficient physical consistency in the prediction results. On the other hand, some methods perform only a single forward calculation during the prediction process, making it difficult to reflect the continuous correction process of the physical field as the state parameters change, thus affecting the prediction accuracy and stability. Summary of the Invention

[0005] This invention addresses the problems in the prior art by providing a multiphysics iterative coupling prediction method, system, terminal, and medium. It solves the problem in the prior art where different physical fields are treated as independent prediction targets, or where only simple parallel prediction methods are used, failing to effectively characterize the coupling relationship between the temperature and pressure fields, resulting in insufficient physical consistency in the prediction results. Furthermore, it solves the problem in the prior art where only a single forward calculation is performed during the prediction process, making it difficult to reflect the continuous correction process of the physical field as state parameters change, thus affecting prediction accuracy and stability.

[0006] The technical solution adopted in this invention is as follows:

[0007] Firstly, this application provides a multiphysics iterative coupling prediction method, which includes the following steps:

[0008] Step S1: Based on the prediction batch input, initialize the state of the object to be predicted and construct a structured feature representation for multiphysics prediction. The structured feature representation is used to characterize the initial state of the object to be predicted under spatial structure, physical properties and operating conditions.

[0009] Step S2: Construct a temperature prediction model for predicting the temperature physical field and a pressure prediction model for predicting the pressure physical field;

[0010] The temperature prediction model and the pressure prediction model are independent of each other;

[0011] Step S3: Based on the structured feature representation, the temperature prediction model is used to perform multiple rounds of prediction updates on the temperature physical field;

[0012] During the prediction update process, the state parameters related to the temperature physical field are dynamically adjusted based on the current prediction state of the temperature physical field. The prediction results of the temperature physical field are continuously corrected based on the adjusted state parameters until the preset iteration termination condition is met, and a stable temperature physical field prediction state is obtained.

[0013] Step S4: Based on the prediction results of the temperature physical field and its corresponding state parameters, the pressure prediction model is used to perform prediction processing on the pressure physical field to obtain the pressure physical field prediction results corresponding to the temperature physical field prediction state.

[0014] Step S5: Output the multi-physics steady-state prediction results obtained from the prediction processing of temperature and pressure physical fields.

[0015] Furthermore, step S1 includes:

[0016] Step S1-1: Based on the prediction batch input, the spatial discrete structure of the object to be predicted is mapped into graph structure data. The graph structure data includes a set of nodes and connection relationships used to represent the spatial adjacency relationship between nodes.

[0017] Step S1-2: Configure node features for each node to characterize the node's spatial location, material physical properties, and operating conditions, forming a structured feature representation required for multiphysics prediction;

[0018] Step S1-3: Initialize the temperature physical field state and pressure physical field state in the node features.

[0019] Furthermore, in step S2, both the temperature prediction model and the pressure prediction model are based on a feature aggregation method using multi-feature map attention.

[0020] The multi-feature map attention structure is as follows: input node features and edge features, and calculate the weighted aggregation information of neighboring nodes through an attention mechanism.

[0021]

[0022] in, This is a learnable attention weight vector. It is a learnable linear transformation matrix. , For node features, As edge features, This indicates a splicing operation.

[0023] Furthermore, step S3 includes:

[0024] Step S3-1: Based on the structured feature representation, call the temperature physical field prediction model to perform temperature physical field prediction on the current node features to obtain the temperature change or temperature update.

[0025] Step S3-2: Update the temperature physical field state corresponding to each node according to the temperature change or temperature update to obtain the temperature physical field prediction state for the current round.

[0026] Step S3-3: Based on the updated temperature physics field prediction state, dynamically adjust the state parameters related to the temperature physics field, and feed the adjusted state parameters back to the node features for the next round of temperature physics field prediction update.

[0027] Repeat steps S3-1 to S3-3 until the preset iteration termination condition is met, and a stable temperature physical field prediction state is obtained.

[0028] Furthermore, in step S3-3, based on the updated predicted state of the temperature physical field, the state parameters related to the temperature physical field are dynamically adjusted using a preset empirical function.

[0029] Furthermore, step S4 includes:

[0030] Step S4-1: Based on the obtained stable temperature physical field prediction state and its corresponding state parameters, construct the input feature representation required for pressure physical field prediction;

[0031] Step S4-2: Using the pressure prediction model, perform pressure physics field prediction on the input feature representation to obtain the pressure physics field prediction result for the current round;

[0032] Step S4-3: Update the pressure physical field state based on the pressure physical field prediction results. Repeat the pressure physical field prediction and update process when the preset conditions are met to obtain stable pressure physical field prediction results.

[0033] Furthermore, the method also includes joint training of the temperature prediction model and the pressure prediction model, the joint training including:

[0034] Obtain the actual temperature and pressure physical field results corresponding to the stable temperature and pressure physical field prediction results;

[0035] The training error is calculated based on the predicted temperature physical field results and the corresponding actual temperature physical field results, as well as the predicted pressure physical field results and the corresponding actual pressure physical field results.

[0036] The model parameters of the temperature prediction model and the pressure prediction model are updated based on the training error.

[0037] Secondly, this application provides a multiphysics iterative coupling prediction system for implementing the multiphysics iterative coupling prediction method as described in the first aspect. The system includes:

[0038] The input processing unit is configured to receive batch prediction inputs, initialize the state of the object to be predicted, and construct a structured feature representation for multiphysics prediction to characterize the initial state of the object under spatial structure, physical properties and operating conditions.

[0039] The model building unit is configured to build a temperature prediction model for predicting the temperature physical field and a pressure prediction model for predicting the pressure physical field. The temperature prediction model and the pressure prediction model are independent of each other.

[0040] The temperature physical field prediction unit is configured to perform multiple rounds of prediction updates on the temperature physical field based on structured feature representation and temperature prediction model. During the prediction update process, the state parameters related to the temperature physical field are dynamically adjusted according to the current prediction state of the temperature physical field until a stable temperature physical field prediction state is obtained.

[0041] The pressure physical field prediction unit is configured to perform prediction processing on the pressure physical field based on the stable temperature physical field prediction state and its corresponding state parameters, and update the pressure physical field when the preset conditions are met, so as to obtain a stable pressure physical field prediction result.

[0042] The result output unit is configured to output the multi-physics steady-state prediction results obtained based on temperature physical field prediction processing and pressure physical field prediction processing.

[0043] Thirdly, this application provides a terminal, including:

[0044] Memory, used to store multiphysics iterative coupling prediction programs;

[0045] A processor is configured to implement the steps of the multiphysics iterative coupling prediction method as described in the first aspect when executing the multiphysics iterative coupling prediction apparatus.

[0046] Fourthly, this application provides a computer-readable storage medium that stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the multiphysics iterative coupling prediction method as described in the first aspect.

[0047] As can be seen from the above technical solutions, the advantages of the present invention are:

[0048] By constructing independent temperature and pressure prediction models and introducing a multiphysics iterative coupling mechanism during the prediction process, multiphysics prediction no longer employs simple parallel or one-time computation methods. Instead, it involves multiple rounds of prediction updates for the temperature physics field, dynamically adjusting temperature-related state parameters based on the temperature prediction status in each update round, thereby gradually obtaining stable temperature physics field prediction results. Based on this stable temperature physics field prediction results and their corresponding state parameters, the pressure physics field is then predicted. This ensures that the dependencies between different physics fields are reasonably reflected in the prediction process, which is beneficial for improving the physical consistency of multiphysics prediction results.

[0049] By mapping the object to be predicted to graph structure data, a unified model is performed on the spatial discrete structure, node adjacency relationship and node physical properties. This enables the prediction method to adapt to complex spatial structures and irregular grid forms, and achieves rapid prediction of multi-physics distribution states without relying on high-cost numerical simulation, thereby improving the efficiency of multi-condition analysis and evaluation.

[0050] By introducing a multi-round iterative update mechanism in the temperature physical field prediction process and continuously correcting the prediction results during the iteration process, this application can simulate the evolution process of the physical field gradually stabilizing. Compared with the method of making only a single prediction, it helps to reduce the sensitivity of prediction error to the initial state or local disturbance, and improve the stability and reliability of prediction results. Attached Figure Description

[0051] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart illustrating the steps of the multiphysics iterative coupling prediction method in the embodiment;

[0053] Figure 2 This is a flowchart of the multiphysics iterative coupling prediction method in the embodiment;

[0054] Figure 3 This is a diagram illustrating the internal architecture of the multi-feature dual-channel graph attention network in the embodiment.

[0055] Figure 4 This is a comparison chart of the predicted physical fields in the examples;

[0056] Figure 5 This is an architecture diagram of the multiphysics iterative coupling prediction system in the embodiment. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0058] Please see Figures 1-4 As shown, this application provides a multiphysics iterative coupling prediction method, including the following steps:

[0059] Step S1: Based on the prediction batch input, initialize the state of the object to be predicted and construct a structured feature representation for multiphysics prediction. The structured feature representation is used to characterize the initial state of the object to be predicted under spatial structure, physical properties and operating conditions.

[0060] Step S1-1: Based on the prediction batch input, the spatial discrete structure of the object to be predicted is mapped into graph structure data. The graph structure data includes a set of nodes and connection relationships used to represent the spatial adjacency relationship between nodes.

[0061] Step S1-2: Configure node features for each node to characterize the node's spatial location, material physical properties, and operating conditions, forming a structured feature representation required for multiphysics prediction;

[0062] Step S1-3: Initialize the temperature physical field state and pressure physical field state in the node features;

[0063] Step S2: Construct a temperature prediction model for predicting the temperature physical field and a pressure prediction model for predicting the pressure physical field;

[0064] The temperature prediction model and the pressure prediction model are independent of each other;

[0065] Both the temperature prediction model and the pressure prediction model are based on a feature aggregation method using multi-feature map attention.

[0066] The multi-feature map attention structure is as follows: input node features and edge features, and calculate the weighted aggregation information of neighboring nodes through an attention mechanism.

[0067]

[0068] in, This is a learnable attention weight vector. It is a learnable linear transformation matrix. , For node features, As edge features, This indicates a splicing operation.

[0069] Step S3: Based on the structured feature representation, the temperature prediction model is used to perform multiple rounds of prediction updates on the temperature physical field;

[0070] During the prediction update process, the state parameters related to the temperature physical field are dynamically adjusted based on the current prediction state of the temperature physical field. The prediction results of the temperature physical field are continuously corrected based on the adjusted state parameters until the preset iteration termination condition is met, and a stable temperature physical field prediction state is obtained.

[0071] Step S3-1: Based on the structured feature representation, call the temperature physical field prediction model to perform temperature physical field prediction on the current node features to obtain the temperature change or temperature update.

[0072] Step S3-2: Update the temperature physical field state corresponding to each node according to the temperature change or temperature update to obtain the temperature physical field prediction state for the current round.

[0073] Step S3-3: Based on the updated temperature physics field prediction state, dynamically adjust the state parameters related to the temperature physics field using a preset empirical function, and feed the adjusted state parameters back to the node features for the next round of temperature physics field prediction update.

[0074] Repeat steps S3-1 to S3-3 until the preset iteration termination condition is met, and a stable temperature physical field prediction state is obtained.

[0075] Step S4: Based on the prediction results of the temperature physical field and its corresponding state parameters, the pressure prediction model is used to perform prediction processing on the pressure physical field to obtain the pressure physical field prediction results corresponding to the temperature physical field prediction state.

[0076] Step S4-1: Based on the obtained stable temperature physical field prediction state and its corresponding state parameters, construct the input feature representation required for pressure physical field prediction;

[0077] Step S4-2: Using the pressure prediction model, perform pressure physics field prediction on the input feature representation to obtain the pressure physics field prediction result for the current round;

[0078] Step S4-3: Update the pressure physical field state based on the pressure physical field prediction results. Repeat the pressure physical field prediction and update process when the preset conditions are met to obtain stable pressure physical field prediction results.

[0079] Step S5: Output the multi-physics steady-state prediction results obtained from the prediction processing of temperature and pressure physical fields.

[0080] In some embodiments, the method further includes jointly training the temperature prediction model and the pressure prediction model, wherein the joint training includes:

[0081] Obtain the actual temperature and pressure physical field results corresponding to the stable temperature and pressure physical field prediction results;

[0082] The training error is calculated based on the predicted temperature physical field results and the corresponding actual temperature physical field results, as well as the predicted pressure physical field results and the corresponding actual pressure physical field results.

[0083] The model parameters of the temperature prediction model and the pressure prediction model are updated based on the training error.

[0084] Joint training employs supervised learning to train the model.

[0085] Training objective: After completing K temperature-driven iterations, calculate the final output temperature T. (K) and pressure P (K) Joint loss between the actual label and the real label.

[0086] Optimization strategy: The two optimizers independently update the parameters of their respective models. During backpropagation, the losses in the temperature and pressure fields propagate to their respective model parameters through a unidirectional dependency chain of T→Properties→P.

[0087] In some embodiments, graph structure data of the object to be predicted is loaded, each graph containing a set of nodes V and a set of edges E.

[0088] Construct an initial feature vector for each node ,Include:

[0089] Geometric features: Normalized coordinates (x, y).

[0090] Physical properties: thermal conductivity k, specific heat capacity C p Density ρ, viscosity μ.

[0091] Operating parameters: grid-side power, valve-side power, ambient temperature (broadcast to each node globally).

[0092] Coupled field placeholder: Reserved dimension to store intermediate values ​​of the physical field to be predicted (initially set to 0 or the state of the previous time step).

[0093] Construct a feature vector Eattr for each edge, which includes edge length weights and boundary type encoding (such as oil-copper interface, oil-paper interface, etc.).

[0094] In some embodiments, two independent graph neural network models are built: a temperature prediction model (TemperatureGNN) and a pressure prediction model (PressureGNN).

[0095] Both models are based on the multi-feature map attention module (GATBlock).

[0096] GATBlock structure: Input node features and edge features, and calculate the weighted aggregation information of neighboring nodes through an attention mechanism.

[0097] The model consists of an input MLP layer, multiple GATBlock layers, and an output MLP layer.

[0098] In some embodiments, T-GAT internal iteration and physical property parameter calculation:

[0099] Set the number of temperature field iterations K (e.g., K=5).

[0100] In the k-th iteration (k=1,2,...,K):

[0101] T-GAT prediction: This involves using the original node features... Current physical properties (Or the previous prediction of field T) is input into TemperatureGNN, and the output is the temperature prediction value T for the current round. (k) .

[0102] Material property parameter update (knowledge guidance): utilizing As input, the set of fluid property parameters is calculated and updated in real time using a preset empirical function (such as a function of density / viscosity as a function of temperature). ,For example , .

[0103] Convergence Propulsion: Update T(k) or This process is repeated K times as input for the next iteration, allowing the temperature field and physical properties to reach a near steady state with rapid convergence.

[0104] P-GAT one-way prediction:

[0105] Feature concatenation: combining the features of the original nodes With the final convergent physical property parameters The data is then pieced together to form the input for the pressure model. .

[0106] Forward propagation: Input PressureGNN, output the pressure prediction value for the current round:

[0107]

[0108] in, This is the predicted pressure value.

[0109] Please see Figure 3 The figure illustrates the dual-channel GAT sub-model architecture for heterogeneous inference. The parallel structures of TemperatureGNN and PressureGNN are highlighted, along with how they utilize edge-weighted attention mechanisms to fuse node and edge features at each layer.

[0110] Please see Figure 4 This figure visually compares the distributions of the temperature and pressure fields predicted by the model of this invention (MFGAT), the benchmark model GraphSage (MFSAGE), Graph Convolutional Neural Network (MFGCN), Multi-Layer Perceptron (MFMLP), and the ablation experimental model with the actual values. The ablation experimental models include the model with global features removed (NG), the model with iterative coupling removed (NI), and the model with material information removed (NM). The figure shows the distribution of absolute errors, intuitively demonstrating the prediction accuracy advantage and robustness of MFGAT in complex regions.

[0111] Table 1 is a comparison table of loss values ​​for generalized prediction of temperature field under multiple operating conditions. It compares the loss function values ​​(mean absolute error MAE, mean relative error MRE, and root mean square error RMSE) of the MFGAT model and the benchmark model in temperature field prediction, demonstrating the optimization effect and performance advantages of the model of this invention in handling temperature prediction tasks in coupled fields.

[0112] Table 1

[0113]

[0114] Table 2 is a comparison table of loss values ​​for generalized prediction of pressure fields under multiple working conditions. It compares the loss function values ​​(MAE, MRE, RMSE) of the MFGAT model and the benchmark model in pressure field prediction, demonstrating the optimization effect and performance advantages of the model of this invention in handling pressure prediction tasks in coupled fields.

[0115] Table 2

[0116]

[0117] Please see Figure 5 As shown, in some embodiments, this application provides a multiphysics iterative coupling prediction system for implementing a multiphysics iterative coupling prediction method, the system comprising:

[0118] The input processing unit is configured to receive batch prediction inputs, initialize the state of the object to be predicted, and construct a structured feature representation for multiphysics prediction to characterize the initial state of the object under spatial structure, physical properties and operating conditions.

[0119] The model building unit is configured to build a temperature prediction model for predicting the temperature physical field and a pressure prediction model for predicting the pressure physical field. The temperature prediction model and the pressure prediction model are independent of each other.

[0120] The temperature physical field prediction unit is configured to perform multiple rounds of prediction updates on the temperature physical field based on structured feature representation and temperature prediction model. During the prediction update process, the state parameters related to the temperature physical field are dynamically adjusted according to the current prediction state of the temperature physical field until a stable temperature physical field prediction state is obtained.

[0121] The pressure physical field prediction unit is configured to perform prediction processing on the pressure physical field based on the stable temperature physical field prediction state and its corresponding state parameters, and update the pressure physical field when the preset conditions are met, so as to obtain a stable pressure physical field prediction result.

[0122] The result output unit is configured to output the multi-physics steady-state prediction results obtained based on temperature physical field prediction processing and pressure physical field prediction processing.

[0123] In some embodiments, this application provides a terminal, including:

[0124] Memory, used to store multiphysics iterative coupling prediction programs;

[0125] A processor is used to execute the steps of the multiphysics iterative coupling prediction method when performing the multiphysics iterative coupling prediction system.

[0126] In some embodiments, this application provides a computer-readable storage medium that stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the multiphysics iterative coupling prediction method.

[0127] The above description is merely a preferred embodiment of one or more embodiments of this specification and is not intended to limit the scope of one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this specification should be included within the protection scope of one or more embodiments of this specification.

Claims

1. A multiphysics iterative coupling prediction method, characterized in that, Includes the following steps: Step S1: Based on the prediction batch input, initialize the state of the object to be predicted and construct a structured feature representation for multiphysics prediction. The structured feature representation is used to characterize the initial state of the object to be predicted under spatial structure, physical properties and operating conditions. Step S2: Construct a temperature prediction model for predicting the temperature physical field and a pressure prediction model for predicting the pressure physical field; The temperature prediction model and the pressure prediction model are independent of each other; Step S3: Based on the structured feature representation, the temperature prediction model is used to perform multiple rounds of prediction updates on the temperature physical field; During the prediction update process, the state parameters related to the temperature physical field are dynamically adjusted based on the current prediction state of the temperature physical field. The prediction results of the temperature physical field are continuously corrected based on the adjusted state parameters until the preset iteration termination condition is met, and a stable temperature physical field prediction state is obtained. Step S4: Based on the prediction results of the temperature physical field and its corresponding state parameters, the pressure prediction model is used to perform prediction processing on the pressure physical field to obtain the pressure physical field prediction results corresponding to the temperature physical field prediction state. Step S4-1: Based on the obtained stable temperature physical field prediction state and its corresponding state parameters, construct the input feature representation required for pressure physical field prediction; Step S4-2: Using the pressure prediction model, perform pressure physics field prediction on the input feature representation to obtain the pressure physics field prediction result for the current round; Step S4-3: Update the pressure physical field state based on the pressure physical field prediction results. Repeat the pressure physical field prediction and update process when the preset conditions are met to obtain stable pressure physical field prediction results. Step S5: Output the multi-physics steady-state prediction results obtained from the prediction processing of temperature and pressure physical fields.

2. The multiphysics iterative coupling prediction method according to claim 1, characterized in that, Step S1 includes: Step S1-1: Based on the prediction batch input, the spatial discrete structure of the object to be predicted is mapped into graph structure data. The graph structure data includes a set of nodes and connection relationships used to represent the spatial adjacency relationship between nodes. Step S1-2: Configure node features for each node to characterize the node's spatial location, material physical properties, and operating conditions, forming a structured feature representation required for multiphysics prediction; Step S1-3: Initialize the temperature physical field state and pressure physical field state in the node features.

3. The multiphysics iterative coupling prediction method according to claim 1, characterized in that, In step S2, both the temperature prediction model and the pressure prediction model are based on a feature aggregation method using multi-feature map attention. The multi-feature map attention structure is as follows: input node features and edge features, and calculate the weighted aggregation information of neighboring nodes through an attention mechanism. in, This is a learnable attention weight vector. It is a learnable linear transformation matrix. , For node features, As edge features, This indicates a splicing operation.

4. The multiphysics iterative coupling prediction method according to claim 1, characterized in that, Step S3 includes: Step S3-1: Based on the structured feature representation, call the temperature physical field prediction model to perform temperature physical field prediction on the current node features to obtain the temperature change or temperature update. Step S3-2: Update the temperature physical field state corresponding to each node according to the temperature change or temperature update to obtain the temperature physical field prediction state for the current round. Step S3-3: Based on the updated temperature physics field prediction state, dynamically adjust the state parameters related to the temperature physics field, and feed the adjusted state parameters back to the node features for the next round of temperature physics field prediction update. Repeat steps S3-1 to S3-3 until the preset iteration termination condition is met, and a stable temperature physical field prediction state is obtained.

5. The multiphysics iterative coupling prediction method according to claim 4, characterized in that, In step S3-3, based on the updated predicted state of the temperature physical field, the state parameters related to the temperature physical field are dynamically adjusted using a preset empirical function.

6. The multiphysics iterative coupling prediction method according to claim 1, characterized in that, The method also includes joint training of the temperature prediction model and the pressure prediction model, the joint training including: Obtain the actual temperature and pressure physical field results corresponding to the stable temperature and pressure physical field prediction results; The training error is calculated based on the predicted temperature physical field results and the corresponding actual temperature physical field results, as well as the predicted pressure physical field results and the corresponding actual pressure physical field results. The model parameters of the temperature prediction model and the pressure prediction model are updated based on the training error.

7. A multiphysics iterative coupling prediction system, used to implement the multiphysics iterative coupling prediction method as described in claim 1, characterized in that, The system includes: The input processing unit is configured to receive batch prediction inputs, initialize the state of the object to be predicted, and construct a structured feature representation for multiphysics prediction to characterize the initial state of the object under spatial structure, physical properties and operating conditions. The model building unit is configured to build a temperature prediction model for predicting the temperature physical field and a pressure prediction model for predicting the pressure physical field. The temperature prediction model and the pressure prediction model are independent of each other. The temperature physical field prediction unit is configured to perform multiple rounds of prediction updates on the temperature physical field based on structured feature representation and temperature prediction model. During the prediction update process, the state parameters related to the temperature physical field are dynamically adjusted according to the current prediction state of the temperature physical field until a stable temperature physical field prediction state is obtained. The pressure physical field prediction unit is configured to perform prediction processing on the pressure physical field based on the stable temperature physical field prediction state and its corresponding state parameters, and update the pressure physical field when the preset conditions are met, so as to obtain a stable pressure physical field prediction result. The result output unit is configured to output the multi-physics steady-state prediction results obtained based on temperature physical field prediction processing and pressure physical field prediction processing.

8. A terminal, characterized in that, include: Memory, used to store multiphysics iterative coupling prediction programs; A processor is configured to implement the steps of the multiphysics iterative coupling prediction method as described in claim 1 when executing the multiphysics iterative coupling prediction device.

9. A computer-readable storage medium, characterized in that, The storage medium stores computer instructions. When the computer reads the computer instructions from the storage medium, the computer executes the multiphysics iterative coupling prediction method as described in claim 1.

Images