Valve performance simulation and optimization method and system based on hybrid deep learning

Abstract

The invention discloses a valve performance simulation and optimization method and system based on mixed deep learning, and the method comprises the steps: S1, carrying out the finite element simulation analysis of a valve, and solving through the thermosetting coupling of a multi-physics field, stress, deformation and temperature distribution data of the valve under different working conditions and maximum thermal stress and maximum thermal deformation displacement of key parts of the valve are obtained; s2, after data processing, taking stress, deformation and temperature distribution data of the valve under different working conditions as input parameters, taking maximum thermal stress and maximum thermal deformation displacement as output parameters, training a neural network model, and establishing a thermal-structural performance prediction model; s3, on the basis of a prediction result, a reference point-based non-dominated genetic algorithm is adopted, a reference point generation and self-adaptive agent model is combined, and multi-objective optimization is carried out to output an optimal parameter combination so as to adjust valve design; according to the invention, deep learning and multi-objective optimization are carried out on the finite element simulation result by using the neural network, and the valve performance prediction precision and optimization efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment design and optimization, and more specifically, to a valve performance simulation and optimization method and system based on hybrid deep learning. Background Art

[0002] Currently, the existing valve performance estimation mainly includes the following technologies:

[0003] Physical experiments and empirical formulas: Data is obtained through laboratory tests such as pressure tests and flow tests, and the performance is estimated by combining relevant empirical formulas such as the Darcy-Weisbach formula. However, the experimental cost is high, the cycle is long, and the generalization ability of the empirical formula for complex working conditions is insufficient;

[0004] Finite element simulation technology (FEA): Tools such as ANSYS and COMSOL are used to perform stress analysis on the valve structure, simulate material deformation, fatigue life, etc.; the internal flow field distribution of the valve (such as flow velocity, pressure gradient) is simulated through computational fluid dynamics (CFD). However, high-precision simulation requires dense mesh division, and the calculation is time-consuming. A single simulation may take several hours to several days;

[0005] Traditional optimization methods: Valve parameters are adjusted based on the trial-and-error method or the response surface method, lacking the effective processing ability for multi-objective and non-linear problems, relying on manual experience to adjust parameters, and it is difficult to achieve automated optimization.

[0006] In summary, traditional valve design relies on experience and experiments, with repeated trial-and-error, low efficiency and high cost. Moreover, finite element analysis consumes a lot of calculation time and resources under high-precision meshes, making it difficult to meet the requirements of rapid iteration; the existing machine learning models lack sufficient high-quality training data support, resulting in limited prediction accuracy and generalization ability, and it is difficult to achieve a rapid mapping from target performance to geometric parameters, unable to meet the design requirements for specific performance requirements; the coupling between finite element analysis and deep learning models is not tight enough, and the advantages of both cannot be fully utilized; in terms of optimization, parameter optimization relies on manual parameter adjustment, lacking automated and multi-objective collaborative optimization capabilities.

[0007] Therefore, how to provide a valve performance simulation and optimization method and system based on hybrid deep learning to quickly simulate and optimize the performance of valves under different working conditions and improve the accuracy of valve performance prediction is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0008] In view of this, the present invention provides a valve performance simulation and optimization method and system based on hybrid deep learning, which adopts multi-physical field coupling simulation technology to simultaneously simulate multiple physical fields under the target working conditions of the valve, capture the dynamic response of the valve; constructs input parameter and output parameter models in finite element analysis for predicting the performance indicators of the valve under any working conditions, and at the same time adopts transfer learning technology to transfer the pre-trained weights of the general valve model to the new type of valve, reducing data requirements; the finite element simulation generates high-precision data to train the neural network, and the prediction results of the neural network are fed back to the optimization module, and the optimized design parameters re-trigger the data model to ensure the self-correction ability of the system, so as to solve some of the technical problems mentioned in the background technology.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A valve performance simulation and optimization method based on hybrid deep learning, comprising the following steps:

[0011] S1. Perform finite element simulation analysis on the valve and use the multi-physical field thermo-solid coupling solution method to obtain the stress, deformation and temperature distribution data of the valve under different working conditions, as well as the maximum thermal stress and the maximum displacement of the thermal deformation of the key components of the valve.

[0012] S2. After processing the simulation data, use the stress, deformation and temperature distribution data of the valve under different working conditions as input parameters, and the maximum thermal stress and the maximum displacement of the thermal deformation as output parameters to train the neural network model and establish a thermal-structural performance prediction model.

[0013] S3. Based on the prediction results of the thermal-structural performance prediction model, adopt a reference point-based non-dominated genetic algorithm, combine the reference point generation and the adaptive surrogate model for multi-objective optimization, and output the optimal parameter combination to adjust the valve design.

[0014] Preferably, the specific content of step S1 includes:

[0015] S11. Establish a three-dimensional geometric model of the valve and define the material properties, including thermal parameters and structural parameters.

[0016] S12. Adopt a hybrid adaptive mesh division of tetrahedrons and triangles for the established three-dimensional geometric model of the valve.

[0017] S13. Adopt multi-physical field coupling to set the thermodynamic boundary conditions and structural mechanics boundaries, and configure the solver and convergence criteria.

[0018] S14. Solve and perform post-processing to output the parameters of the maximum thermal stress and the maximum displacement of the thermal deformation.

[0019] Preferably, in step S11, the thermal parameters include the thermal conductivity k, specific heat capacity cp, and coefficient of thermal expansion α; the structural parameters include the elastic modulus E, Poisson's ratio ν, and input pressure σ in ;

[0020] In step S13, the thermodynamic boundary conditions are specifically: the heat transfer parameters of the temperature field, the fluid inlet temperature T in , the coefficient of thermal expansion α, and the convective heat transfer coefficient h on the outer wall conv and the ambient temperature T ref ;

[0021] The structural mechanics boundary is specifically: the input pressure σ is set according to the pressure load applied in the actual working condition in , the fixed constraints are set, and the fluid dynamics boundary conditions are applied;

[0022] The configured solver is the direct coupling solver in the thermal stress module using COMSOL Multiphysics multi-physics simulation; the residual and the maximum number of iterations are used as the convergence criteria.

[0023] Preferably, in step S2, the specific structure of the thermal-structural performance prediction model established is:

[0024] The input layer is the material parameters and working condition parameters. Among them, the material parameters include the thermal conductivity k, coefficient of thermal expansion α, elastic modulus E, and Poisson's ratio ν; the working condition parameters include the input temperature T in , the convective heat transfer coefficient h on the outer wall conv and the input pressure σ in ;

[0025] The hidden layer uses 4 convolutional layers with a kernel size of 3×3×3 to extract the three-dimensional temperature and structural characteristics of the valve, encodes the geometric and material parameters into a 128-dimensional vector, splices the CNN output with the encoder vector, and passes through 3 fully connected layers to obtain the prediction result;

[0026] The activation function of the hidden layer uses LeakyReLU, and linearly activates the output layer;

[0027] The output layer: outputs the thermal stress and the maximum displacement of thermal deformation.

[0028] Preferably, the specific content of training the neural network model is:

[0029] Generate steady-state data through parametric thermo-solid coupling simulation, perform data augmentation, and add Gaussian noise to the temperature field;

[0030] Set the loss function to reduce the difference between the predicted value and the true value:

[0031] L = λ1∣T max,pre -T max,sim|+λ2|σ max,pre -σ max,sim |

[0032] Among them, T max,pre is the predicted maximum temperature, which is the predicted value of the highest temperature during the steady-state operation of the valve output by the neural network model. T max,sim is the simulated maximum temperature, which is the highest temperature value of the valve obtained by finite element simulation and used as the training label. σ max,pre is the predicted maximum thermal stress, which is the predicted value of the peak thermal stress of the valve output by the neural network model. σ max,sim is the simulated maximum thermal stress, which is the maximum thermal stress value obtained by finite element simulation and used as the training label. λ1 is the temperature error weight coefficient, and λ2 is the stress error weight coefficient.

[0033] Preferably, the specific content of step S3 includes:

[0034] S31. Initialization stage: Define the design variables as the material parameters and working condition parameters design variables. The material parameters include the elastic modulus E and Poisson's ratio ν. The working condition parameters include the inlet temperature T in and the input pressure σ in , and generate reference points based on the material parameters and working condition parameter variables;

[0035] S32. Surrogate model construction:

[0036] Train the CNN-LSTM hybrid model based on historical finite element data and conduct model verification. The input is the design variable, and the output is the maximum allowable stress σ max and the allowable maximum deformation u max ;

[0037] Define the surrogate model confidence threshold for dynamic confidence evaluation. If the variance of the predicted value Var of σ max ≤ 10 MPa, skip the finite element verification; otherwise, calibrate it with the simulation data;

[0038] S33. Use initial full-parameter space exploration and later local refined search for iterative optimization;

[0039] S34. Dynamically adjust the material plan: Build an internal material database and record the elastic modulus E and Poisson's ratio ν. Check the E and ν of the current optimal solution every 10 iterations, match the closest engineering material from the library, and ensure that the constraints are not violated, that is, the maximum allowable stress and the allowable maximum deformation need to meet the material allowable range.

[0040] Preferably, for step S33, the specific content of the full-parameter space exploration is:

[0041] Initial population generation: Randomly generate n groups of design parameters, covering material parameters: elastic modulus E, Poisson's ratio ν, and coefficient of thermal expansion α, and operating conditions parameters: inlet temperature T in and input pressure σ in ; Call the surrogate model to predict performance and screen non-dominated solutions as the initial population;

[0042] Adaptive sampling: Generate new parameters in the sparse region to supplement the sampling points;

[0043] The specific content of local refined search is as follows:

[0044] Focus on the hot spot area: Select a preset proportion of individuals with the best performance in the current Pareto solution set and conduct local search in the defined neighborhood;

[0045] Hybrid optimization strategy:

[0046] Gradient-assisted mutation: Calculate the objective function gradient for the continuous variables of input pressure and input temperature;

[0047] Directed crossover: Cross the individuals with high temperature and low stress with those with low temperature and high stress to explore potential balance points.

[0048] A valve performance simulation and optimization system based on hybrid deep learning, based on the described valve performance simulation and optimization method based on hybrid deep learning, includes: a finite element simulation module, a thermal-structural performance prediction module, and a multi-objective optimization module;

[0049] The finite element simulation module is used to perform finite element simulation analysis on the valve and use the multi-physics thermo-solid coupling solution method to obtain the stress, deformation, and temperature distribution data of the valve under different operating conditions, as well as the maximum thermal stress and the maximum displacement of thermal deformation of the key components of the valve;

[0050] The thermal-structural performance prediction module is used to, after processing the simulation data, use the stress, deformation, and temperature distribution data of the valve under different operating conditions as input parameters and the maximum thermal stress and the maximum displacement of thermal deformation as output parameters to train the neural network model and establish a thermal-structural performance prediction model;

[0051] The multi-objective optimization module is used to, based on the prediction results of the thermal-structural performance prediction model, adopt a reference point-based non-dominated genetic algorithm, combine with the reference point generation and the adaptive surrogate model to perform multi-objective optimization, and output the optimal parameter combination to adjust the valve design.

[0052] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the described valve performance simulation and optimization method based on hybrid deep learning.

[0053] A processing terminal includes a memory and a processor. A computer program that can run on the processor is stored in the memory. When the processor executes the computer program, the described method for valve performance simulation and optimization based on hybrid deep learning is implemented.

[0054] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method and system for valve performance simulation and optimization based on hybrid deep learning. By combining finite element analysis and deep learning technology, and using a neural network to perform deep learning on the finite element simulation results, the accuracy of valve performance prediction can be improved. Especially when facing complex working conditions, it can better capture non-linear relationships. Compared with traditional manual optimization or optimization methods based solely on physical models, the optimization module of the present invention can greatly shorten the design cycle and improve the optimization efficiency. Through the training of the deep learning model, the system can handle valve performance prediction under different working conditions, with strong adaptability and flexibility. The system has a user-friendly interface, which is convenient for engineers to get started quickly, reduces manual intervention, and improves work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0056] Figure 1 Schematic diagram of a method for valve performance simulation and optimization based on hybrid deep learning provided by the present invention;

[0057] Figure 2 Schematic diagram of the finite element simulation analysis process provided by the present invention;

[0058] Figure 3 Schematic diagram of the neural network model structure provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0060] The embodiments of the present invention disclose a method for valve performance simulation and optimization based on hybrid deep learning, as Figure 1 , including the following steps:

[0061] S1. Perform finite element simulation analysis on the valve and use the multi-physics thermo-solid coupling solution method to obtain the stress, deformation, and temperature distribution data of the valve under different working conditions, as well as the maximum thermal stress and the maximum displacement of thermal deformation of the key components of the valve.

[0062] S2. After processing the simulation data, use the stress, deformation, and temperature distribution data of the valve under different working conditions as input parameters, and the maximum thermal stress and the maximum displacement of thermal deformation as output parameters to train the neural network model and establish a thermal-structural performance prediction model.

[0063] S3. Based on the prediction results of the thermal-structural performance prediction model, adopt the reference point-based non-dominated genetic algorithm, combine the reference point generation with the adaptive surrogate model for multi-objective optimization, and output the optimal parameter combination to adjust the valve design.

[0064] To further implement the above technical solution, as Figure 2 , the specific content of step S1 includes:

[0065] S11. Establish a three-dimensional geometric model of the valve and define the material properties, including thermal parameters and structural parameters.

[0066] In this embodiment, a three-dimensional geometric model of the valve is established based on CAD and SOLIDWORKS software.

[0067] S12. Use a mixed adaptive mesh of tetrahedrons and triangles for the established three-dimensional geometric model of the valve to improve the calculation accuracy and stability, and perform encrypted mesh processing on some thermally sensitive regions and connection regions.

[0068] S13. Use multi-physics coupling to set the thermodynamic boundary conditions and structural mechanics boundaries, and configure the solver and convergence criteria.

[0069] S14. Solve and perform post-processing, and output the parameters of the maximum thermal stress and the maximum displacement of thermal deformation.

[0070] To further implement the above technical solution, in step S11, the thermal parameters include the thermal conductivity k, specific heat capacity cp, and thermal expansion coefficient α; the structural parameters include the elastic modulus E, Poisson's ratio ν, and input pressure σ in ;

[0071] In step S13, the thermodynamic boundary conditions are specifically: the heat transfer parameters of the temperature field, the fluid inlet temperature T in , the thermal expansion coefficient α, the convective heat transfer coefficient h of the outer wall conv and the ambient temperature T ref ;

[0072] In this embodiment, the basis for setting the thermodynamic boundary conditions is:

[0073] Heat conduction equation (steady state): , where T is the temperature field and Q is the internal heat source;

[0074] Thermoelastic equation (steady state stress): , where is the strain tensor, T ref is the reference temperature, and σ is the steady state stress;

[0075] Apply the thermal contact boundary condition;

[0076] The specific structural mechanics boundary is: Set the input pressure σ according to the actual working condition by applying the pressure load in , set the fixed constraint, and apply the hydrodynamic boundary condition;

[0077] In this embodiment, the basis for setting the structural mechanics boundary is to set the flange fixed constraint ux = uy = uz = 0; Apply different pressure loads to the inner flow channel;

[0078] Configure the solver: Use the direct coupling solver in the thermal stress module of COMSOL Multiphysics multi-physics simulation;

[0079] Convergence criterion: Residual R < 10 -6 , and the maximum number of iterations is 200.

[0080] In this embodiment, finite element simulation is used to conduct a detailed analysis of the mechanical properties (such as stress, deformation, etc.) of the valve under different working conditions, and the simulation results provide a data basis for the training of the neural network.

[0081] To further implement the above technical solution, as Figure 3 , in step S2, the specific structure of the thermal-structural performance prediction model established is:

[0082] Input layer, which is material parameters and working condition parameters. Among them, the material parameters include thermal conductivity k, coefficient of thermal expansion α, elastic modulus E, and Poisson's ratio ν; The working condition parameters include input temperature T in , convective heat transfer coefficient h on the outer wall conv and input pressure σ in ;

[0083] Hidden layer, using 4 convolutional layers with a kernel size of 3×3×3, used to extract the three-dimensional temperature and structural characteristics of the valve, encode the geometric and material parameters into a 128-dimensional vector, splice the CNN output with the encoder vector, and obtain the prediction result through 3 fully connected layers; The number of neurons in the 3 fully connected layers is 256 - 128 - 64;

[0084] The activation function of the hidden layer uses LeakyReLU, α = 0.1, and linearly activates the output layer;

[0085] Output layer: Output the thermal stress and the maximum displacement of thermal deformation as parameters.

[0086] To further implement the above technical solution, the specific content of training the neural network model is as follows:

[0087] Generate steady-state data through parametric thermo-solid coupling simulation (generate 500 groups of steady-state data, covering extreme working conditions, such as T in = 500 °C, σ in = 30 MPa), and perform data augmentation. Add Gaussian noise to the temperature field (C μ = 0, σ = 2 °C, to improve the robustness of the model);

[0088] Set the loss function to reduce the difference between the predicted value and the true value:

[0089] L = λ1∣T max,pre - T max,sim ∣ + λ2∣σ max,pre - σ max,sim ∣

[0090] Where, T max,pre is the predicted maximum temperature, which is the predicted value of the highest temperature when the valve is in steady-state operation output by the neural network model, T max,sim is the simulated maximum temperature, which is the highest temperature value of the valve obtained by finite element simulation calculation and used as the training label, σ max,pre is the predicted maximum thermal stress, which is the predicted value of the peak thermal stress of the valve output by the neural network model, σ max,sim is the simulated maximum thermal stress, which is the maximum thermal stress value obtained by finite element simulation and used as the training label. λ1 is the temperature error weight coefficient, set to 0.6, and λ2 is the stress error weight coefficient, set to 0.4.

[0091] In this embodiment, before training the neural network model, it also includes preprocessing the data obtained from finite element analysis, such as cleaning and normalizing, removing noise and unifying the scale to ensure the high quality of the data. Subsequently, fuse multiple data sources to improve the generalization ability and prediction accuracy of the model; design a suitable neural network model, combine the finite element analysis results and experimental data, and train a high-precision performance prediction model that can predict the performance of the valve under different working conditions, including parameters such as flow rate, pressure, and sealing performance.

[0092] To further implement the above technical solution, in step S3, adopt the non-dominated sorting genetic algorithm NSGA-III algorithm based on reference points, combine reference points to generate an adaptive surrogate model, and solve the valve design optimization problem under high-dimensional objectives (temperature, stress, deformation). The specific content includes:

[0093] S31. Initialization stage: Define the design variables as material parameters and working condition parameters. The material parameters include the elastic modulus E and Poisson's ratio ν, and the working condition parameters include the inlet temperature T in and the input pressure σ in , and generate reference points based on the material parameters and working condition parameter variables;

[0094] S32. Surrogate model construction:

[0095] Train a CNN-LSTM hybrid model based on historical finite element data (500 groups) and conduct model verification. The input is the design variables, and the outputs are the maximum allowable stress σ max allowed for the material and the maximum allowable deformation u max ; Model verification: The stress error in the test set < 5%, and the deformation error < 5%;

[0096] Define a surrogate model confidence threshold (η = 0.9) for dynamic confidence evaluation. If the variance of the predicted value of σ max Var ≤ 10 MPa, skip the finite element verification; otherwise, calibrate with the simulation data;

[0097] S33. Use initial full parameter space exploration and later local refined search for iterative optimization;

[0098] S34. Dynamically adjust the material scheme: Build a built-in material database (including stainless steel, titanium alloy, ceramics, etc.) and record the elastic modulus E and Poisson's ratio ν. Check the E and ν of the current optimal solution every 10 iterations, match the closest engineering material from the library, and at the same time ensure that the constraints are not violated, that is, the maximum allowable stress and the maximum allowable deformation need to meet the allowable range of the material.

[0099] In this embodiment, after obtaining the valve performance prediction, use an optimization algorithm to adjust the valve design, quickly find the optimal design scheme, and improve the working efficiency and reliability of the valve.

[0100] To further implement the above technical solution, the specific content of the full parameter space exploration in step S33 is:

[0101] Initial population generation: Randomly generate n groups of design parameters (50 groups), covering material parameters: elastic modulus E, Poisson's ratio ν, and thermal expansion coefficient α, and working condition parameters: inlet temperature T in and input pressure σ in ; Call the surrogate model to predict the performance, and screen the non-dominated solutions (Pareto solutions) as the initial population;

[0102] Adaptive sampling: Generate new parameters in the sparse area to supplement the sampling points. The sparse area is the area in the target space that has not been covered; for example, if the target value in a certain area (σ max = 100 MPa, umax If there is no solution (e.g., σ ≤ 100 MPa, u ≤ 0.1 mm), 10 groups of new parameters are generated in this neighborhood;

[0103] The specific content of local refined search is as follows:

[0104] Focus on the hot spot area: Select a preset proportion of individuals with the best performance in the current Pareto solution set and conduct local search in the defined neighborhood;

[0105] In this embodiment, 50% of the individuals with the best performance in the current Pareto solution set are selected. For example, if σ ≤ 100 MPa, u ≤ 0.1 mm, local search is conducted in its neighborhood; The defined neighborhood is the inlet temperature T ± 20 °C, and the input pressure σ ± 10 MPa; max ≤ 100 MPa, u max ≤ 0.1 mm, and local search is conducted in its neighborhood; The defined neighborhood is the inlet temperature T in ± 20 °C, and the input pressure σ in ± 10 MPa;

[0106] Hybrid optimization strategy:

[0107] Gradient-assisted mutation: Calculate the gradient of the objective function for the continuous variables of input pressure and input temperature;

[0108] In this embodiment, the input pressure is generated along an input gradient of 5 MPa, and the input temperature is generated along the direction of the input gradient of a 20 °C decrease;

[0109] Directed crossover: Cross the high-temperature and low-stress individuals with the low-temperature and high-stress individuals to explore potential balance points.

[0110] In this embodiment, during the performance optimization process, if the predicted values are directly used, 1 finite element calculation (about 2 hours) can be saved, the surrogate model reduces the finite element calls by 70%, dynamically matches the actual engineering materials, and avoids the problem that the theoretical optimal solution cannot be manufactured.

[0111] A valve performance simulation and optimization system based on hybrid deep learning, based on a valve performance simulation and optimization method based on hybrid deep learning, includes: a finite element simulation module, a thermal-structural performance prediction module, and a multi-objective optimization module;

[0112] The finite element simulation module is used to perform finite element simulation analysis on the valve and use the multi-physics thermo-solid coupling solution method to obtain the stress, deformation, and temperature distribution data of the valve under different working conditions, as well as the maximum thermal stress and the maximum displacement of thermal deformation of the key components of the valve;

[0113] The thermal-structural performance prediction module is used to, after processing the simulation data, use the stress, deformation, and temperature distribution data of the valve under different working conditions as input parameters, and the maximum thermal stress and the maximum displacement of thermal deformation as output parameters to train the neural network model and establish a thermal-structural performance prediction model;

[0114] The multi-objective optimization module is used to perform multi-objective optimization based on the prediction results of the thermal-structural performance prediction model, adopt a reference point-based non-dominated genetic algorithm, generate a reference point and an adaptive surrogate model, and output an optimal parameter combination to adjust the valve design.

[0115] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, a valve performance simulation and optimization method based on hybrid deep learning is implemented.

[0116] A processing terminal includes a memory and a processor. A computer program that can run on the processor is stored in the memory. When the processor executes the computer program, a valve performance simulation and optimization method based on hybrid deep learning is implemented.

[0117] The various embodiments in this specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple. For the relevant parts, refer to the descriptions in the method part.

[0118] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A valve performance simulation and optimization method based on hybrid deep learning, characterized in that, It includes the following steps: S1. Conduct finite element simulation analysis on the valve and use the multi-physics thermo-solid coupling solution method to obtain the stress, deformation, and temperature distribution data of the valve under different working conditions, as well as the maximum thermal stress and the maximum displacement of thermal deformation of the key components of the valve; S2. After processing the simulation data, use the stress, deformation, and temperature distribution data of the valve under different working conditions as input parameters, and the maximum thermal stress and the maximum displacement of thermal deformation as output parameters to train the neural network model and establish a thermal-structural performance prediction model; S3. Based on the prediction results of the thermal-structural performance prediction model, adopt the reference point-based non-dominated genetic algorithm, combine the reference point generation with the adaptive surrogate model for multi-objective optimization, and output the optimal parameter combination to adjust the valve design.

2. The valve performance simulation and optimization method based on hybrid deep learning according to claim 1, characterized in that, The specific content of step S1 includes: S11. Establish a three-dimensional geometric model of the valve and define the material properties, including thermal parameters and structural parameters; S12. Use a mixed adaptive mesh of tetrahedrons and triangles for the established three-dimensional geometric model of the valve; S13. Adopt multi-physics coupling to set the thermodynamic boundary conditions and structural mechanics boundaries, and configure the solver and convergence criteria; S14. Solve and perform post-processing, and output the parameters of the maximum thermal stress and the maximum displacement of thermal deformation.

3. The method for valve performance simulation and optimization based on hybrid deep learning according to claim 2, characterized in that In step S11, the thermal parameters include the thermal conductivity k, the specific heat capacity cp, and the coefficient of thermal expansion α; the structural parameters include the elastic modulus E, the Poisson's ratio ν, and the input pressure σ in ; In step S13, the thermodynamic boundary conditions are specifically: the heat transfer parameters of the temperature field, the fluid inlet temperature T in , the coefficient of thermal expansion α, the convective heat transfer coefficient h on the outer wall cov and the ambient temperature T ref ; The specific boundary conditions of structural mechanics are as follows: Apply a pressure load according to the actual working conditions to set the input pressure σ in , set fixed constraints, and apply hydrodynamic boundary conditions; The configured solver is the direct coupling solver in the thermal stress module of COMSOL Multiphysics multi-physics simulation; the residual and the maximum number of iterations are the convergence criteria.

4. A valve performance simulation and optimization method based on hybrid deep learning according to claim 1, characterized in that In step S2, the specific structure of the established thermal-structural performance prediction model is: The input layer is the material parameters and operating conditions parameters. Among them, the material parameters include the thermal conductivity k, the coefficient of thermal expansion α, the elastic modulus E, and the Poisson's ratio ν; the operating conditions parameters include the input temperature T in , the convective heat transfer coefficient h on the outer wall conv , and the input pressure σ in ; Hidden layer: Use 4 convolutional layers with a kernel size of 3×3×3 to extract the three-dimensional temperature and structural characteristics of the valve, encode the geometric and material parameters into 128-dimensional vectors, splice the CNN output with the encoder vector, and pass through 3 fully connected layers to obtain the prediction result; The activation function of the hidden layer uses LeakyReLU and linearly activates the output layer; Output layer: Output the parameters of thermal stress and the maximum displacement of thermal deformation.

5. A valve performance simulation and optimization method based on hybrid deep learning according to claim 1, characterized in that, The specific content of training the neural network model is: Generate steady-state data through parametric thermo-solid coupling simulation and perform data augmentation by adding Gaussian noise to the temperature field; Set the loss function to reduce the difference between the predicted value and the true value: L = λ1∣T max,pre -T max,sim ∣ + λ2∣σ max,pre -σ max,sim ∣; Among them, T max,pre is the predicted maximum temperature, which is the predicted value of the highest temperature when the valve is in steady-state operation output by the neural network model. T max,sim is the simulated maximum temperature, which is the highest temperature value of the valve obtained by finite element simulation and used as the training label. σ max,pre is the predicted maximum thermal stress, which is the predicted value of the peak thermal stress of the valve output by the neural network model. σ max,sim is the simulated maximum thermal stress, which is the maximum thermal stress value obtained by finite element simulation and used as the training label. λ1 is the temperature error weight coefficient, and λ2 is the stress error weight coefficient.

6. A method for valve performance simulation and optimization based on hybrid deep learning according to claim 1, characterized in that The specific content of step S3 includes: S31. Initialization stage: Define the design variables as the material parameters and the working condition parameters. The material parameters include the elastic modulus E and the Poisson's ratio ν, and the working condition parameters include the inlet temperature T in and the input pressure σ in , and generate reference points based on the material parameters and the working condition parameter variables; S32. Surrogate model construction: Train a CNN-LSTM hybrid model based on historical finite element data and perform model verification, with the design variables as the input and the maximum allowable stress σ and the maximum allowable deformation u that the material permits as the output max and the maximum allowable deformation u max ; Define the confidence threshold of the surrogate model for dynamic confidence evaluation. If σ max The variance of the predicted value Var ≤ 10 MPa, then skip the finite element verification; otherwise, calibrate with the simulation data. S33. Adopt initial full-parameter space exploration and subsequent local refinement search for iterative optimization; S34. Dynamically adjust the material plan: Build an internal material database and record the elastic modulus E and Poisson's ratio ν. Check the E and ν of the current optimal solution every 10 iterations, match the closest engineering material from the library, and ensure that the constraints are not violated, that is, the maximum allowable stress and the allowable maximum deformation need to meet the material allowable range.

7. A method for valve performance simulation and optimization based on hybrid deep learning according to claim 6, characterized in that, For step S33, the specific content of the full-parameter space exploration is: Initial population generation: Randomly generate n groups of design parameters, covering material parameters: elastic modulus E, Poisson's ratio ν, and coefficient of thermal expansion α, and operating conditions parameters: inlet temperature T in and input pressure σ in ; Call the surrogate model to predict the performance, and screen the non-dominated solutions as the initial population; Adaptive sampling: Generate new parameters in the sparse area to supplement the sampling points; The specific content of the local refinement search is: Focus on the hot spot area: Select a preset proportion of individuals with the best performance in the current Pareto solution set and conduct local search in the defined neighborhood; Hybrid optimization strategy: Gradient-assisted mutation: Calculate the objective function gradient for the continuous variables of the input pressure and input temperature; Directed crossover: Cross high-temperature and low-stress individuals with low-temperature and high-stress individuals to explore potential balance points.

8. A valve performance simulation and optimization system based on hybrid deep learning, characterized in that, A valve performance simulation and optimization method based on hybrid deep learning according to any one of claims 1-7, comprising: a finite element simulation module, a thermal-structural performance prediction module, and a multi-objective optimization module; The finite element simulation module is used to perform finite element simulation analysis on the valve and use a multi-physics thermo-solid coupling solution method to obtain stress, deformation, and temperature distribution data of the valve under different working conditions, as well as the maximum thermal stress and the maximum displacement of thermal deformation of the key components of the valve; The thermal-structural performance prediction module is used to, after processing the simulation data, use the stress, deformation, and temperature distribution data of the valve under different working conditions as input parameters, and the maximum thermal stress and the maximum displacement of thermal deformation as output parameters to train a neural network model and establish a thermal-structural performance prediction model; The multi-objective optimization module is used to, based on the prediction results of the thermal-structural performance prediction model, adopt a reference-point-based non-dominated genetic algorithm, combine reference points with an adaptive surrogate model for multi-objective optimization, and output an optimal parameter combination to adjust the valve design.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements a valve performance simulation and optimization method based on hybrid deep learning according to any one of claims 1-7.

10. A processing terminal includes a memory and a processor, and a computer program that can run on the processor is stored in the memory, characterized in that, When the processor executes the computer program, it implements a valve performance simulation and optimization method based on hybrid deep learning according to any one of claims 1-7.

Images