Creep stress relaxation response model parameter calibration method and device based on deep neural network
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- EAST CHINA UNIV OF SCI & TECH
- Filing Date
- 2026-03-24
- Publication Date
- 2026-06-02
Smart Images
Figure CN121920154B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and more specifically, to a method and apparatus for calibrating parameters of a creep stress relaxation response model based on a deep neural network. Background Technology
[0002] In the nuclear power field, the creep stress relaxation behavior of high-temperature pressure-bearing components under complex working conditions such as load holding, cyclic loading, and non-isothermal conditions is affected by tensile and compressive stress states and stress redistribution, exhibiting significant nonlinearity and complexity.
[0003] In the existing technology, creep stress relaxation response models usually rely on creep relaxation test data combined with human experience or repeated iterative calculations for parameter calibration. This process is lengthy, complex, and costly, making it difficult to meet the efficiency requirements of engineering applications. Summary of the Invention
[0004] In view of this, this application provides a method and apparatus for calibrating parameters of a creep stress relaxation response model based on a deep neural network, in order to solve the problem of low efficiency in calibrating creep stress relaxation response model parameters in the prior art.
[0005] Specifically, this application is implemented through the following technical solution:
[0006] In a first aspect, this application provides a method for calibrating parameters of a creep stress relaxation response model based on a deep neural network, including:
[0007] A creep stress relaxation simulation model for high-temperature pressure-bearing components is constructed, and multiple sets of parameter samples corresponding to the creep stress relaxation simulation model are determined.
[0008] Based on the parameter samples and the creep stress relaxation simulation model, the creep stress relaxation response results are calculated to obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data on stress changes over time;
[0009] For each set of parameter samples, based on the creep stress relaxation response results corresponding to the parameter samples, the creep stress relaxation behavior is fitted with parameters to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters.
[0010] Construct a parameter prediction model for each fitting parameter, and train each parameter prediction model based on the parameter values of the fitting parameters corresponding to each group of parameter samples.
[0011] Based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, the parameter values of the target creep stress relaxation response model are determined; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
[0012] In some embodiments, the construction of the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component includes:
[0013] A three-dimensional finite element model of the high-temperature pressure-bearing component is constructed; the data of the three-dimensional finite element model includes at least one of the component geometric information, material parameters, mesh generation, boundary conditions, and load conditions.
[0014] Based on the three-dimensional finite element model, heat transfer analysis and calculation were performed to obtain the steady-state temperature field of the high-temperature pressure-bearing component under the target working conditions.
[0015] Under the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model and creep stress relaxation behavior analysis is performed to obtain the mechanical response of the high-temperature pressure-bearing component in the cyclic loading and load holding stages;
[0016] Generate automated script files corresponding to the above steps of constructing the three-dimensional finite element model, heat transfer analysis calculation, and creep stress relaxation behavior analysis calculation; the automated script files are used to construct the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component.
[0017] In some embodiments, determining the multiple sets of parameter samples corresponding to the creep stress relaxation simulation model includes:
[0018] Based on the requirements of creep stress relaxation behavior analysis, a set of parameters for creep stress relaxation behavior analysis is determined, wherein the set of parameters includes at least one of material parameters, load conditions, and operating condition parameters;
[0019] Determine the value range of each parameter in the parameter set to obtain the parameter sampling space corresponding to each parameter.
[0020] Within the parameter sampling space, the parameter set is sampled using Latin hypercube sampling to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0021] In some embodiments, the parameter prediction model includes a genetic algorithm-optimized backpropagation neural network model; the input of the parameter prediction model includes at least one of material parameters, load conditions, and operating condition parameters; the output of the parameter prediction model includes corresponding fitting parameters.
[0022] In some embodiments, the parameter prediction model is trained through the following steps:
[0023] The parameter values of each group of parameter samples and their corresponding fitted parameters are normalized, and training sample sets and test sample sets are constructed based on the normalization results.
[0024] A genetic algorithm is used to globally optimize the initial weights and thresholds of the parameter prediction model; the optimization objects of the genetic algorithm include the connection weights and bias parameters of each layer of the parameter prediction model; the optimization objective of the genetic algorithm includes individual performance evaluation based on the fitness function.
[0025] Based on the intermediate model parameters obtained from the global optimization, the backpropagation algorithm is used to train the parameter prediction model. The intermediate model parameters of the parameter prediction model are iteratively updated by minimizing the function error between the predicted output and the true output until the convergence condition is met.
[0026] Based on the model parameters obtained when the convergence condition is met, as well as the input normalized parameters and the output denormalized parameters, the trained parameter prediction model is determined.
[0027] In some embodiments, the convergence condition includes an error index satisfying a preset condition; the error index includes root mean square error and / or coefficient of determination.
[0028] In some embodiments, determining the parameter values of the target creep stress relaxation response model based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed includes:
[0029] Based on the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, an input parameter vector is constructed; the input parameter vector includes at least one of material parameters, load conditions, and working condition parameters.
[0030] The input parameter vector is input into each of the trained parameter prediction models to obtain the parameter values output by each parameter prediction model.
[0031] Based on the parameter values output by each of the parameter prediction models, the parameter values of the target creep stress relaxation response model are determined.
[0032] Secondly, this application provides a parameter calibration device for a creep stress relaxation response model based on a deep neural network, comprising:
[0033] The first construction module is used to construct a creep stress relaxation simulation model corresponding to a high-temperature pressure-bearing component and determine multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0034] The calculation module is used to calculate the creep stress relaxation response results based on the parameter samples and the creep stress relaxation simulation model, and obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data of stress changes over time;
[0035] The fitting module is used to fit the creep stress relaxation behavior to each set of parameter samples based on the creep stress relaxation response results corresponding to the parameter samples, so as to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters.
[0036] The second construction module is used to construct a parameter prediction model corresponding to each fitting parameter, and to train each parameter prediction model based on the parameter values of the fitting parameters corresponding to each group of parameter samples.
[0037] The third construction module is used to determine the parameter values of the target creep stress relaxation response model based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
[0038] In some embodiments, the first building module is used to:
[0039] A three-dimensional finite element model of the high-temperature pressure-bearing component is constructed; the data of the three-dimensional finite element model includes at least one of the component geometric information, material parameters, mesh generation, boundary conditions, and load conditions.
[0040] Based on the three-dimensional finite element model, heat transfer analysis and calculation were performed to obtain the steady-state temperature field of the high-temperature pressure-bearing component under the target working conditions.
[0041] Under the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model and creep stress relaxation behavior analysis is performed to obtain the mechanical response of the high-temperature pressure-bearing component in the cyclic loading and load holding stages;
[0042] Generate automated script files corresponding to the above steps of constructing the three-dimensional finite element model, heat transfer analysis calculation, and creep stress relaxation behavior analysis calculation; the automated script files are used to construct the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component.
[0043] In some embodiments, the first building module is used to:
[0044] Based on the requirements of creep stress relaxation behavior analysis, a set of parameters for creep stress relaxation behavior analysis is determined, wherein the set of parameters includes at least one of material parameters, load conditions, and operating condition parameters;
[0045] Determine the value range of each parameter in the parameter set to obtain the parameter sampling space corresponding to each parameter.
[0046] Within the parameter sampling space, the parameter set is sampled using Latin hypercube sampling to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0047] In some embodiments, the parameter prediction model includes a genetic algorithm-optimized backpropagation neural network model; the input of the parameter prediction model includes at least one of material parameters, load conditions, and operating condition parameters; the output of the parameter prediction model includes corresponding fitting parameters.
[0048] In some embodiments, the second building module is used to:
[0049] The parameter values of each group of parameter samples and their corresponding fitted parameters are normalized, and training sample sets and test sample sets are constructed based on the normalization results.
[0050] A genetic algorithm is used to globally optimize the initial weights and thresholds of the parameter prediction model; the optimization objects of the genetic algorithm include the connection weights and bias parameters of each layer of the parameter prediction model; the optimization objective of the genetic algorithm includes individual performance evaluation based on the fitness function.
[0051] Based on the intermediate model parameters obtained from the global optimization, the backpropagation algorithm is used to train the parameter prediction model. The intermediate model parameters of the parameter prediction model are iteratively updated by minimizing the function error between the predicted output and the true output until the convergence condition is met.
[0052] Based on the model parameters obtained when the convergence condition is met, as well as the input normalized parameters and the output denormalized parameters, the trained parameter prediction model is determined.
[0053] In some embodiments, the convergence condition includes an error index satisfying a preset condition; the error index includes root mean square error and / or coefficient of determination.
[0054] In some embodiments, the third building module is used for:
[0055] Based on the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, an input parameter vector is constructed; the input parameter vector includes at least one of material parameters, load conditions, and working condition parameters.
[0056] The input parameter vector is input into each of the trained parameter prediction models to obtain the parameter values output by each parameter prediction model.
[0057] Based on the parameter values output by each of the parameter prediction models, the parameter values of the target creep stress relaxation response model are determined.
[0058] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the creep stress relaxation response model parameter calibration method based on a deep neural network as described in the first aspect or any embodiment of the first aspect.
[0059] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method for calibrating parameters of a creep stress relaxation response model based on a deep neural network as described in the first aspect or any embodiment of the first aspect.
[0060] The method and apparatus for calibrating parameters of a creep stress relaxation response model based on a deep neural network provided in this application constructs a creep stress relaxation simulation model and generates multiple sets of parameter samples. First, it calculates and fits the creep stress relaxation response results corresponding to each set of parameter samples. Then, it trains a parameter prediction model based on the fitted parameters. This eliminates the need to repeat the finite element simulation calculation and parameter fitting process in subsequent analysis stages, allowing the creep stress relaxation response under the target working condition to be obtained. This significantly reduces computation time and resource consumption, improving analysis efficiency. By uniformly fitting the creep stress relaxation response results corresponding to multiple sets of parameter samples and training a parameter prediction model based on the fitted parameters, it achieves automatic prediction of creep stress relaxation response model parameters. This reduces the uncertainty caused by repeated manual adjustment of model parameters in traditional methods, improving the stability and consistency of analysis results and increasing the efficiency of creep stress relaxation response model parameter calibration. Attached Figure Description
[0061] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present application of a method for calibrating parameters of a creep stress relaxation response model based on a deep neural network;
[0062] Figure 2 This is one of the schematic diagrams illustrating fitting parameters in an exemplary embodiment of this application;
[0063] Figure 3 This is a second schematic diagram illustrating fitting parameters in an exemplary embodiment of this application;
[0064] Figure 4 This is a third schematic diagram illustrating fitting parameters in an exemplary embodiment of this application;
[0065] Figure 5This is a fourth schematic diagram illustrating fitting parameters in an exemplary embodiment of this application;
[0066] Figure 6 This is a schematic diagram of a fitting curve shown in an exemplary embodiment of this application;
[0067] Figure 7 This is a schematic diagram of a parameter calibration device for a creep stress relaxation response model based on a deep neural network, as shown in an exemplary embodiment of this application;
[0068] Figure 8 This is a schematic diagram of a computer device illustrated in an exemplary embodiment of this application. Detailed Implementation
[0069] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.
[0070] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.
[0071] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."
[0072] Research has found that existing creep stress relaxation response models typically rely on creep relaxation test data combined with human experience or iterative calculations for parameter calibration. This process is lengthy, complex, and costly, making it difficult to meet the efficiency requirements of engineering applications.
[0073] In view of this, the embodiments of this application provide a method and apparatus for calibrating parameters of a creep stress relaxation response model based on a deep neural network. By constructing a creep stress relaxation simulation model and generating multiple sets of parameter samples, the creep stress relaxation response results corresponding to each set of parameter samples are first calculated and parameters are fitted. Then, a parameter prediction model is trained based on the fitted parameters. This eliminates the need to repeat the finite element simulation calculation and parameter fitting process in subsequent analysis stages, thus obtaining the creep stress relaxation response under the target working condition. This significantly reduces computation time and computational resource consumption, and improves analysis efficiency. By uniformly fitting the creep stress relaxation response results corresponding to multiple sets of parameter samples and training a parameter prediction model based on the fitted parameters, the automatic prediction of creep stress relaxation response model parameters is achieved. This reduces the uncertainty caused by repeated adjustments of model parameters based on manual experience in traditional methods, improves the stability and consistency of analysis results, and increases the efficiency of creep stress relaxation response model parameter calibration.
[0074] The deficiencies of the existing technical solutions are the result of the inventor's practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed in this application below should be considered as the inventor's contributions to this application.
[0075] To facilitate understanding of this embodiment, the application scenario of the deep neural network-based creep stress relaxation response model parameter calibration method disclosed in this application embodiment will first be introduced. The execution entity of the deep neural network-based creep stress relaxation response model parameter calibration method provided in this application embodiment can be a computer device. In some possible implementations, this deep neural network-based creep stress relaxation response model parameter calibration method can be implemented by a processor calling computer-readable instructions stored in memory.
[0076] See Figure 1 The diagram shown is a flowchart illustrating a method for calibrating parameters of a creep stress relaxation response model based on a deep neural network, as demonstrated in an exemplary embodiment of this application. The method includes steps S101 to S105, wherein:
[0077] S101. Construct a creep stress relaxation simulation model for high-temperature pressure-bearing components, and determine multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0078] High-temperature pressure-bearing components refer to structural parts that bear internal pressure, external pressure, or mechanical loads for extended periods in high-temperature environments. During service, they are simultaneously subjected to temperature fields and mechanical loads, and may be under conditions of load maintenance, cyclic loading, or non-isothermal conditions. Under these conditions, the mechanical response of high-temperature pressure-bearing components is not only related to the instantaneous load but also closely related to time, temperature, load path, and material creep characteristics. Their stress and strain will continuously evolve over time.
[0079] Creep stress relaxation behavior refers to the mechanical behavior of a material or structure under high temperature conditions, where its strain remains basically constant or is constrained, and the internal stress gradually decreases over time due to creep deformation.
[0080] During the service life of high-temperature pressure-bearing components, the components are subjected to continuous or cyclic loads in a high-temperature environment, and the material undergoes creep deformation. Under the action of constraint conditions, the internal stress of the component relaxes over time, thus forming the creep stress relaxation behavior of high-temperature pressure-bearing components.
[0081] In the nuclear power field, the creep stress relaxation behavior of high-temperature pressure-bearing components under complex conditions such as holding load, cyclic loading, and non-isothermal conditions is significantly nonlinear and complex, influenced by tensile and compressive stress states and stress redistribution. When the primary and secondary stresses have opposite signs, alternating compressive and tensile creep strains occur during the holding load stage, leading to structural creep recovery behavior. However, existing creep stress relaxation models typically rely on creep relaxation test data combined with manual experience or iterative calculations for parameter calibration. This process is lengthy, complex, and costly, failing to meet the efficiency requirements of engineering applications. Furthermore, these methods often require individual parameter adjustments during parameter inversion, making them highly sensitive to initial parameter selection and compromising the stability and consistency of calibration results. Moreover, creep stress relaxation behavior is typically influenced by multiple factors such as temperature, stress state, holding time, and loading path. Traditional parameter fitting methods struggle to effectively characterize the coupling effects between these multiple parameters, especially when considering tensile and compressive effects and complex load paths, significantly limiting their applicability.
[0082] To this end, this application first constructs a creep stress relaxation simulation model for high-temperature pressure-bearing components and determines multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0083] In some embodiments, the construction of the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component includes:
[0084] A three-dimensional finite element model corresponding to the high-temperature pressure-bearing component is constructed. The data of the three-dimensional finite element model includes at least one of the component's geometric information, material parameters, mesh generation, boundary conditions, and load conditions. Based on the three-dimensional finite element model, heat transfer analysis calculations are performed to obtain the steady-state temperature field of the high-temperature pressure-bearing component under target operating parameters. Under the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model, and creep stress relaxation behavior analysis calculations are performed to obtain the mechanical response of the high-temperature pressure-bearing component during cyclic loading and load holding stages. An automated script file is generated corresponding to the above steps of constructing the three-dimensional finite element model, heat transfer analysis calculations, and creep stress relaxation behavior analysis calculations. The automated script file is used to construct the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component.
[0085] In this step, a three-dimensional finite element model of the high-temperature pressure-bearing component is first constructed. This three-dimensional finite element model is used to numerically analyze the mechanical behavior of the high-temperature pressure-bearing component under high temperature and load conditions. Its model data includes at least one or more of the following: component geometric information, material parameters, mesh generation, boundary conditions, and load conditions. Specifically, the component geometric information describes the structural morphology of the high-temperature pressure-bearing component; the material parameters characterize the mechanical and thermal properties of the component material; the mesh generation discretizes the component; and the boundary conditions and load conditions simulate the stress state of the component under actual working conditions.
[0086] After constructing the three-dimensional finite element model, heat transfer analysis is performed based on the model to obtain the steady-state temperature field of the high-temperature pressure-bearing component under target operating parameters. This steady-state temperature field characterizes the temperature distribution of the component under thermal load and serves as the fundamental input condition for considering temperature effects in subsequent mechanical analyses.
[0087] After obtaining the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model under the steady-state temperature field conditions, and creep stress relaxation behavior analysis and calculation are performed. Through this calculation process, the mechanical response of the high-temperature pressure-bearing component during the cyclic loading stage and the load holding stage is obtained, which reflects the creep stress relaxation behavior of the component under the combined action of high temperature and load over time.
[0088] After completing the construction of the three-dimensional finite element model, heat transfer analysis calculation, and creep stress relaxation behavior analysis calculation, the modeling process, analysis flow, and calculation control process corresponding to each of the above steps are generated into an automated script file. This automated script file is used to automatically construct and call the creep stress relaxation simulation model of the high-temperature pressure-bearing component, supporting batch analysis calculations under multiple working conditions.
[0089] The above method can be used to form a creep stress relaxation simulation model for subsequent parameter sample calculation and creep stress relaxation response analysis.
[0090] In some embodiments, determining the multiple sets of parameter samples corresponding to the creep stress relaxation simulation model includes:
[0091] Based on the requirements for creep stress relaxation behavior analysis, a parameter set for creep stress relaxation behavior analysis is determined. The parameter set includes at least one of material parameters, load conditions, and operating condition parameters. The parameter value range corresponding to each parameter in the parameter set is determined to obtain the parameter sampling space corresponding to each parameter. Within the parameter sampling space, the parameter set is sampled using Latin hypercube sampling to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0092] In this step, firstly, based on the requirements of creep stress relaxation behavior analysis, a set of parameters for creep stress relaxation behavior analysis is determined. This set of parameters characterizes the service status of high-temperature pressure-bearing components under different analysis scenarios, and its parameter types include at least one of material parameters, load conditions, and operating condition parameters. By combining these parameters, analysis inputs under different operating conditions can be constructed.
[0093] After determining the parameter set, the parameter value range for each parameter in the parameter set is determined to obtain the parameter sampling space for each parameter. The parameter value range can be set based on material performance parameters, design operating conditions, or engineering experience to limit the reasonable value range of the parameter samples.
[0094] After obtaining the parameter sampling space, the parameter set is sampled within the parameter sampling space using the Latin hypercube sampling method to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model. The Latin hypercube sampling method can achieve uniform coverage of each parameter value within the parameter sampling space, thereby improving the representativeness of the parameter samples in the multidimensional parameter space.
[0095] The multiple sets of parameter samples generated in the above manner are used for the calculation of subsequent creep stress relaxation response results, parameter fitting, and training of parameter prediction models, thereby providing a stable and reliable parameter input basis for creep stress relaxation behavior analysis.
[0096] S102. Based on the parameter samples and the creep stress relaxation simulation model, calculate the creep stress relaxation response results to obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data on stress changes over time.
[0097] In this step, for each set of parameter samples, the material parameters, load conditions, and working condition parameters in the parameter sample can be input into the creep stress relaxation simulation model to construct the analysis working condition corresponding to the parameter sample. Under the analysis working condition, mechanical analysis calculations are performed on the high-temperature pressure-bearing component, causing the component to undergo creep deformation under high-temperature environment and load, and causing the evolution of internal stress over time.
[0098] During the calculation process, the stress variation over time of the high-temperature pressure-bearing component during creep stress relaxation is obtained by solving the mechanical response of the component during the loading, holding, and cyclic loading stages. The creep stress relaxation response results include historical data on the stress variation over time at key locations of the component, which reflects the creep stress relaxation behavior of the component under the combined action of high temperature and load.
[0099] The historical data on stress changes over time can be recorded in the form of a discrete time series, where each time point corresponds to the stress state of the component at that moment. In this way, a set of creep stress relaxation response results can be obtained for each set of parameter samples, thus forming a one-to-one correspondence between multiple sets of parameter samples and creep stress relaxation response results.
[0100] The above calculation process of creep stress relaxation response results provides a data foundation for subsequent parameter fitting of creep stress relaxation behavior based on historical stress data.
[0101] S103. For each set of parameter samples, based on the creep stress relaxation response results corresponding to the parameter samples, perform parameter fitting on the creep stress relaxation behavior to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters.
[0102] In this step, for each set of parameter samples, historical data on the stress variation over time corresponding to that parameter sample are obtained, and a pre-defined creep stress relaxation response model is selected as the fitting model. The creep stress relaxation response model is used to describe the evolution of internal stress over time in high-temperature pressure-bearing components under high temperature and load conditions.
[0103] In this embodiment, based on the historical data of stress variation over time, the unknown parameters in the creep stress relaxation response model are fitted and calculated to ensure that the response equation has good fitting accuracy to the historical stress data throughout the entire analysis time range. The fitting process can be achieved by minimizing the error between the calculated values of the response equation and the historical stress data.
[0104] Through the above parameter fitting process, a creep stress relaxation response model is obtained for each set of parameter samples. The creep stress relaxation response model includes multiple fitting parameters, which are used to characterize the amplitude characteristics, evolution rate, or time-related properties of creep stress relaxation behavior.
[0105] By constructing corresponding creep stress relaxation response models for different parameter samples, we can establish the correspondence between parameter samples and creep stress relaxation behavior characteristics, thereby providing basic data support for subsequent analysis and processing based on fitted parameters.
[0106] In some implementations, the creep stress relaxation response models corresponding to each set of parameter samples may have the same terms, but the coefficients of each term may be different.
[0107] For example, the creep stress relaxation response model can be expressed as:
[0108] ;
[0109] in, for Stress value at time 10:00 For initial stress, and These are the parameters (i.e., fitting parameters) of the creep stress relaxation response model to be fitted.
[0110] S104. Construct a parameter prediction model for each fitting parameter, and train each parameter prediction model based on the parameter values of the fitting parameters corresponding to each group of parameter samples.
[0111] In this step, a parameter prediction model is constructed for each of the multiple fitting parameters included in the creep stress relaxation response model. By separating the prediction tasks for different fitting parameters, each parameter prediction model is used to establish the mapping relationship between the input parameters and a single fitting parameter, thereby reducing the impact of multi-parameter coupling on prediction accuracy.
[0112] Specifically, for each fitted parameter in the creep stress relaxation response model, a parameter prediction model is constructed. The parameter prediction model takes at least one of the material parameters, load conditions, and operating condition parameters as model inputs, and the corresponding fitted parameter as model output, to predict the value of the fitted parameter under given parameter sample conditions.
[0113] After constructing the parameter prediction model, each parameter prediction model is trained based on the parameter values of the fitted parameters corresponding to each group of parameter samples. Specifically, the input parameters in each group of parameter samples are mapped one-to-one with the fitted parameter values obtained through parameter fitting to construct a training dataset for model training.
[0114] During model training, the parameter prediction model learns the correspondence between input parameters and fitted parameters through iterative learning on the training dataset, thereby obtaining a model capable of predicting the corresponding fitted parameters based on the input parameters. Through this training process, the parameter prediction model corresponding to each fitted parameter is determined, enabling it to predict the fitted parameters under different analytical scenarios.
[0115] Through the above methods, independent modeling and training of each fitting parameter in the creep stress relaxation response model were achieved, providing a model foundation for subsequent creep stress relaxation behavior analysis based on the prediction model.
[0116] In some implementations, the parameter prediction model includes a genetic algorithm-optimized backpropagation neural network model; the input of the parameter prediction model includes at least one of material parameters, load conditions, and operating condition parameters; the output of the parameter prediction model includes corresponding fitting parameters.
[0117] In some implementations, the parameter prediction model is trained through the following steps:
[0118] The parameter values of each group of parameter samples and their corresponding fitted parameters are normalized, and training sample sets and test sample sets are constructed based on the normalization results.
[0119] A genetic algorithm is used to globally optimize the initial weights and thresholds of the parameter prediction model; the optimization objects of the genetic algorithm include the connection weights and bias parameters of each layer of the parameter prediction model; the optimization objective of the genetic algorithm includes individual performance evaluation based on the fitness function.
[0120] Based on the intermediate model parameters obtained from the global optimization, the backpropagation algorithm is used to train the parameter prediction model. The intermediate model parameters of the parameter prediction model are iteratively updated by minimizing the function error between the predicted output and the true output until the convergence condition is met.
[0121] Based on the model parameters obtained when the convergence condition is met, as well as the input normalized parameters and the output denormalized parameters, the trained parameter prediction model is determined.
[0122] In this step, the parameter samples and their corresponding fitted parameter values are normalized to eliminate the impact of differences in parameter dimensions and numerical ranges on model training, and training and testing sample sets are constructed. Normalization makes the model training process more stable and accelerates convergence; optimizing initial parameters using a genetic algorithm avoids the problem of traditional neural network training easily getting trapped in local optima, thus improving prediction accuracy.
[0123] Through the above training process, the obtained parameter prediction model can predict the fitting parameters of the corresponding creep stress relaxation response model under different input conditions, providing reliable model parameters for subsequent analysis of creep stress relaxation behavior of high-temperature pressure-bearing components.
[0124] In some embodiments, the convergence condition includes an error metric satisfying a preset condition. The error metric is used to evaluate the accuracy of the parameter prediction model training and includes the root mean square error (RMSE) and / or the coefficient of determination (CQD). ).
[0125] The root mean square error (RMSE) measures the average error between the model's predicted output and the actual output. Its calculation formula is as follows:
[0126] ;
[0127] in, Indicates the first The actual output value of each sample Indicates the first The predicted output value for each sample, where N represents the number of samples.
[0128] Coefficient of determination The formula used to evaluate the goodness of fit of the prediction results is as follows:
[0129] ;
[0130] in, Indicates the first The mean of the true output values of each sample.
[0131] In some implementations, when the RMSE is less than a first preset threshold and the When the value exceeds the second preset threshold, the neural network model training can be considered complete.
[0132] S105. Based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, determine the parameter values of the target creep stress relaxation response model; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
[0133] In this step, corresponding input parameters can be constructed based on the material parameters, load conditions, and operating parameters of the scenario to be analyzed. These input parameters are then input into a trained parameter prediction model to obtain the parameter values of the target creep stress relaxation response model corresponding to the scenario to be analyzed. Furthermore, based on the determined parameter values, a target creep stress relaxation response model for the scenario to be analyzed can be constructed. This model can then be used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component under the scenario to be analyzed, thereby enabling various mechanical analyses.
[0134] In some embodiments, determining the parameter values of the target creep stress relaxation response model based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed includes:
[0135] Based on the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, an input parameter vector is constructed; the input parameter vector includes at least one of material parameters, load conditions, and working condition parameters.
[0136] The input parameter vector is input into each of the trained parameter prediction models to obtain the parameter values output by each parameter prediction model.
[0137] Based on the parameter values output by each of the parameter prediction models, the parameter values of the target creep stress relaxation response model are determined.
[0138] Through the above steps, it is possible to quickly construct creep stress relaxation response models for high-temperature pressure-bearing components under different working conditions without repeating finite element calculations and parameter fitting, providing reliable analysis results for structural integrity assessment and engineering applications.
[0139] For example, the target creep stress relaxation response model can be used for long-term service performance analysis of high-temperature pressure-bearing components. For instance, during the operation of equipment such as steam generator pipelines in nuclear power plants, high-temperature pressure vessels in chemical plants, or high-temperature reactor pipelines, the relevant components are subjected to high temperatures and continuous loads for a long time. Creep and stress relaxation phenomena will gradually occur inside the material, which will lead to changes in the structural stress state over time.
[0140] When conducting equipment safety assessments, the scenario to be analyzed can be determined based on specific operating conditions (such as temperature level, load magnitude, and duration), and the parameter values of the target creep stress relaxation response model can be determined using a trained parameter prediction model. Subsequently, the target creep stress relaxation response model is introduced into structural mechanics analysis or finite element simulation calculations to simulate the stress evolution process of high-temperature pressure-bearing components during long-term operation, obtaining the response results of stress changes over time.
[0141] Based on the analysis results, the stress relaxation degree, potential stress redistribution, and structural safety margin of high-temperature pressure-bearing components under long-term operating conditions can be further evaluated, thereby providing a basis for equipment life prediction, maintenance cycle formulation, and structural reliability assessment.
[0142] For example, the target creep stress relaxation response model can also be used for stress redistribution analysis of high-temperature piping systems. For instance, in the main steam piping system of a nuclear power unit or thermal power unit, the pipes will experience significant creep deformation and stress relaxation during long-term high-temperature operation, resulting in a gradual redistribution of the stress state within the structure.
[0143] When conducting safety analysis of pipeline structures, the analysis scenario can be determined based on operating parameters such as pipeline operating temperature, internal pressure load, and constraint conditions. The parameter values of the target creep stress relaxation response model can then be determined using a trained parameter prediction model. Subsequently, the target creep stress relaxation response model is introduced into the finite element analysis process of the pipeline structure to simulate the stress change process of the pipeline under long-term operating conditions.
[0144] The above analysis can provide the stress relaxation level and stress redistribution at key locations in the pipeline, thus providing a technical basis for pipeline support structure design optimization, stress assessment, and operational safety management.
[0145] For example, the target creep stress relaxation response model can also be used for creep failure risk analysis of welded joints in high-temperature structures. For instance, in high-temperature pressure pipelines or pressure vessels, the welded joint area is often more prone to creep damage due to differences in material structure and stress concentration.
[0146] When assessing the reliability of welded joints, the analysis scenario can be determined based on the operating temperature, load conditions, and material properties of the structure where the welded joint is located. The parameter prediction model is then used to determine the parameter values of the target creep stress relaxation response model. Subsequently, the creep stress relaxation behavior of the welded joint region is simulated in the finite element analysis model.
[0147] The above analysis can identify high-risk stress areas and potential creep damage locations in the welded joint area, thereby providing technical support for welded structure optimization design, operation monitoring, and maintenance planning.
[0148] The following will take a perforated square plate structure as an example of a high-temperature pressure-bearing component to introduce the specific implementation method of the embodiments of this application.
[0149] S1. In this specific embodiment, a perforated square plate structure is taken as the research object. Considering the geometric symmetry of the structure, only a quarter of its geometric model is established for analysis. The ratio of the aperture D to the plate length L of the perforated square plate is 0.2, and the ratio of the plate thickness T to the plate length L is 0.05, where the plate length L is taken as an example value of 50. In terms of load application, the perforated square plate is subjected to a constant mechanical tensile load in the horizontal direction; the thermal load is applied to the structural surface in the form of the temperature difference between the inner surface temperature of the hole and the outer surface temperature of the plate. After the finite element model is established, heat transfer analysis is first performed on the model to obtain the steady-state temperature field of the structure; then, cyclic analysis is carried out based on the temperature field. The cyclic load is set to four cycles, where the thermal load adopts a loading-holding-unloading cyclic form, and the mechanical load remains constant throughout the analysis process. After completing the above analysis process, the finite element calculation job is submitted and the calculation results are obtained; at the same time, the above finite element modeling and analysis process is organized and an automated modeling and calculation script file based on Python language is generated to realize the automatic calling and batch calculation of the finite element model.
[0150] S2. Select multiple parameters related to the creep stress relaxation behavior of the perforated square plate as the input parameter set. This parameter set includes one or more of the following: temperature, pressure, holding time, Young's modulus, coefficient of thermal expansion, yield stress, and creep constitutive model parameters (such as material constant A and stress exponent n). Given the sampling range for each parameter, the Latin hypercube sampling method is used to sample within the multi-parameter space to obtain uniformly covered finite element calculation samples. In this embodiment, a total of 200 sets of parameter samples are generated, each set corresponding to a set of finite element analysis conditions.
[0151] S3. After generating multi-parameter samples, for each set of parameter samples, a Python script is used to batch submit and automate the finite element calculation job, obtain the corresponding creep stress relaxation response, and construct a dataset for machine learning training. The specific process includes the following:
[0152] S31. Key Location Identification and Stress History Extraction
[0153] For each finite element calculation result file, the element and its integration point that reach the maximum value of creep damage related indicators are first searched in the target analysis step, and this position is used as the representative position of creep stress relaxation behavior; then the historical data of equivalent stress and principal stress at the integration point over time are automatically extracted to form a stress-time series.
[0154] S32. Fitting and Evaluation of Creep Stress Relaxation Parameters
[0155] Based on the extracted stress history data, the parameters of the creep stress relaxation response equation are fitted, and the fitted equation is as follows:
[0156] .
[0157] in, For initial stress, and The parameters (i.e., fitting parameters) are those of the creep stress relaxation response model to be fitted. The parameters are fitted using the nonlinear least squares method. and Perform the inversion and use the coefficient of determination. The fitting accuracy is evaluated to ensure the reliability of the fitting results.
[0158] S33. Training Dataset Construction
[0159] The input parameters corresponding to each set of finite element calculations (including temperature, pressure, holding time, material parameters, creep parameters, and one or more of the creep initiation and termination stresses) are compared with the creep stress relaxation parameters obtained by fitting. and One-to-one correspondence, batch writing to data files, ultimately forming a multi-parameter-response dataset for neural network training and prediction.
[0160] S4. Based on the constructed multi-parameter creep stress relaxation response dataset, a neural network model optimized by a genetic algorithm is used to learn and predict the creep stress relaxation parameters. The specific implementation process is as follows:
[0161] S41. Training Sample Construction and Data Preprocessing
[0162] The obtained dataset is divided into training and testing sets. The input vector includes multi-dimensional features such as temperature, pressure, holding time, material, creep parameters, and stress. The output targets are the parameters in the creep stress relaxation response model. and To improve the stability and convergence speed of network training, the input and output data are normalized.
[0163] S42. Genetic Algorithm for Optimizing Neural Network Structure and Initial Parameters
[0164] For parameters With parameters To meet the prediction requirements, single-output feedforward neural network models were constructed, and a genetic algorithm was used to globally optimize the network weights and thresholds. The genetic algorithm uses minimizing prediction error as its objective function, and searches for optimal initial network parameters through selection, crossover, and mutation operations, thereby avoiding the problem of getting trapped in local optima during the training process of traditional neural networks.
[0165] S43. Model Training, Validation, and Prediction Output
[0166] Based on the optimal initial parameters obtained by the genetic algorithm, the neural network is further trained using the backpropagation algorithm until the error convergence condition is met. The model's prediction accuracy is validated using a test set, and the root mean square error (RMSE) and coefficient of determination are used. The prediction results were evaluated. The results show that the parameters... The prediction model on the training and test sets The values are 0.99953 and 0.99764 respectively, corresponding to RMSE values of 1.4192 and 3.5793; parameters The prediction model on the training and test sets The values are 0.99305 and 0.99056, respectively, corresponding to RMSEs of 0.017239 and 0.020182. These results demonstrate that the model possesses high prediction accuracy and good generalization performance. Ultimately, a model capable of rapidly predicting creep stress relaxation parameters based on given working conditions and material parameters is established. and Machine learning models.
[0167] S5. After completing the genetic algorithm optimization and neural network model training described in step S4, the input parameters of the structure to be analyzed under given working conditions are input into the trained neural network model. These input parameters include temperature, pressure, holding time, and material and creep-related parameters. Specifically, a set of input data, including internal temperature, internal pressure, holding time, material elastic modulus, coefficient of thermal expansion, yield stress, and creep model-related parameters, is input into the neural network model as a sample to be predicted. An example set of input data is as follows: internal temperature 501.389, pressure -71.859, holding time 257.002, elastic modulus 205532.88, and coefficient of thermal expansion... The yield stress is 213.988, and the creep parameter A is... The creep parameter n is 4.501.
[0168] After normalizing the input data in the same way as during the model training phase, the parameters in the creep stress relaxation model are then processed using the neural network model. With parameters Predictions are made, and the prediction results are inversely normalized to obtain the corresponding physical quantity parameter values. The predicted parameters... The value is 333.269, parameter The value is 0.058984. Therefore, the method of this invention can achieve rapid inversion of creep stress relaxation model parameters under different working conditions without repeating finite element calculations and parameter fitting processes, improving parameter acquisition efficiency and providing reliable input for subsequent structural creep behavior analysis and engineering evaluation.
[0169] See Figures 2 to 5 The figures shown are one to four schematic diagrams illustrating fitting parameters according to an exemplary embodiment of this application. Wherein, Figure 2 For parameters The comparison between the predicted values of the training set and the actual values of the training set; Figure 3 For parameters Comparison of predicted values on the test set with actual values on the test set; Figure 4 For parameters The comparison between the predicted values of the training set and the actual values of the training set; Figure 5 For parameters The comparison between the predicted values of the test set and the actual values of the test set.
[0170] See Figure 6 The image shown is a schematic diagram of a fitting curve illustrated in an exemplary embodiment of this application. Figure 6 The horizontal axis x represents time, in hours, and the vertical axis y represents the signed von Mises stress, in megapascals (MPa).
[0171] The creep stress relaxation response model parameter calibration method based on deep neural networks provided in this application constructs a creep stress relaxation simulation model and generates multiple sets of parameter samples. First, it calculates and fits the creep stress relaxation response results corresponding to each set of parameter samples. Then, it trains a parameter prediction model based on the fitted parameters. This eliminates the need to repeat the finite element simulation calculation and parameter fitting process in subsequent analysis stages, allowing the creep stress relaxation response under the target working condition to be obtained. This significantly reduces computation time and resource consumption, improving analysis efficiency. By uniformly fitting the creep stress relaxation response results corresponding to multiple sets of parameter samples and training a parameter prediction model based on the fitted parameters, it achieves automatic prediction of creep stress relaxation response model parameters. This reduces the uncertainty caused by repeated manual adjustment of model parameters in traditional methods, improving the stability and consistency of analysis results and increasing the efficiency of creep stress relaxation response model parameter calibration.
[0172] Corresponding to the aforementioned embodiments of the creep stress relaxation response model parameter calibration method based on deep neural networks, this application also provides embodiments of the creep stress relaxation response model parameter calibration device based on deep neural networks.
[0173] See Figure 7 The diagram shown is a schematic representation of a parameter calibration device for a creep stress relaxation response model based on a deep neural network, according to an exemplary embodiment of this application. The device includes:
[0174] The first construction module 710 is used to construct a creep stress relaxation simulation model corresponding to a high-temperature pressure-bearing component and determine multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0175] The calculation module 720 is used to calculate the creep stress relaxation response results based on the parameter samples and the creep stress relaxation simulation model, and obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data of stress changes over time;
[0176] The fitting module 730 is used to fit the creep stress relaxation behavior to each set of parameter samples based on the creep stress relaxation response results corresponding to the parameter samples, so as to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters.
[0177] The second construction module 740 is used to construct a parameter prediction model corresponding to each fitting parameter, and to train each parameter prediction model based on the parameter values of the fitting parameters corresponding to each group of parameter samples.
[0178] The third construction module 750 is used to determine the parameter values of the target creep stress relaxation response model based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
[0179] In some embodiments, the first building module 710 is used to:
[0180] A three-dimensional finite element model of the high-temperature pressure-bearing component is constructed; the data of the three-dimensional finite element model includes at least one of the component geometric information, material parameters, mesh generation, boundary conditions, and load conditions.
[0181] Based on the three-dimensional finite element model, heat transfer analysis and calculation were performed to obtain the steady-state temperature field of the high-temperature pressure-bearing component under the target working conditions.
[0182] Under the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model and creep stress relaxation behavior analysis is performed to obtain the mechanical response of the high-temperature pressure-bearing component in the cyclic loading and load holding stages;
[0183] Generate automated script files corresponding to the above steps of constructing the three-dimensional finite element model, heat transfer analysis calculation, and creep stress relaxation behavior analysis calculation; the automated script files are used to construct the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component.
[0184] In some embodiments, the first building module 710 is used to:
[0185] Based on the requirements of creep stress relaxation behavior analysis, a set of parameters for creep stress relaxation behavior analysis is determined, wherein the set of parameters includes at least one of material parameters, load conditions, and operating condition parameters;
[0186] Determine the value range of each parameter in the parameter set to obtain the parameter sampling space corresponding to each parameter.
[0187] Within the parameter sampling space, the parameter set is sampled using Latin hypercube sampling to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0188] In some embodiments, the parameter prediction model includes a genetic algorithm-optimized backpropagation neural network model; the input of the parameter prediction model includes at least one of material parameters, load conditions, and operating condition parameters; the output of the parameter prediction model includes corresponding fitting parameters.
[0189] In some embodiments, the second building module 740 is used to:
[0190] The parameter values of each group of parameter samples and their corresponding fitted parameters are normalized, and training sample sets and test sample sets are constructed based on the normalization results.
[0191] A genetic algorithm is used to globally optimize the initial weights and thresholds of the parameter prediction model; the optimization objects of the genetic algorithm include the connection weights and bias parameters of each layer of the parameter prediction model; the optimization objective of the genetic algorithm includes individual performance evaluation based on the fitness function.
[0192] Based on the intermediate model parameters obtained from the global optimization, the backpropagation algorithm is used to train the parameter prediction model. The intermediate model parameters of the parameter prediction model are iteratively updated by minimizing the function error between the predicted output and the true output until the convergence condition is met.
[0193] Based on the model parameters obtained when the convergence condition is met, as well as the input normalized parameters and the output denormalized parameters, the trained parameter prediction model is determined.
[0194] In some embodiments, the convergence condition includes an error index satisfying a preset condition; the error index includes root mean square error and / or coefficient of determination.
[0195] In some embodiments, the third building module 750 is used for:
[0196] Based on the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, an input parameter vector is constructed; the input parameter vector includes at least one of material parameters, load conditions, and working condition parameters.
[0197] The input parameter vector is input into each of the trained parameter prediction models to obtain the parameter values output by each parameter prediction model.
[0198] Based on the parameter values output by each of the parameter prediction models, the parameter values of the target creep stress relaxation response model are determined.
[0199] The creep stress relaxation response model parameter calibration device based on deep neural networks provided in this application constructs a creep stress relaxation simulation model and generates multiple sets of parameter samples. It first calculates and fits the creep stress relaxation response results corresponding to each set of parameter samples, and then trains a parameter prediction model based on the fitted parameters. This eliminates the need to repeat the finite element simulation calculation and parameter fitting process in subsequent analysis stages, allowing the creep stress relaxation response under the target working condition to be obtained. This significantly reduces computation time and resource consumption, improving analysis efficiency. By uniformly fitting the creep stress relaxation response results corresponding to multiple sets of parameter samples and training a parameter prediction model based on the fitted parameters, the device achieves automatic prediction of creep stress relaxation response model parameters. This reduces the uncertainty caused by repeated manual adjustment of model parameters in traditional methods, improving the stability and consistency of analysis results and increasing the efficiency of creep stress relaxation response model parameter calibration.
[0200] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.
[0201] This application also provides a computer device, such as... Figure 8 The diagram shown is a schematic representation of a computer device according to an exemplary embodiment of this application. The computer device includes:
[0202] A processor 81 and a memory 82; the memory 82 stores machine-readable instructions executable by the processor 81, and the processor 81 executes the machine-readable instructions stored in the memory 82. When the machine-readable instructions are executed by the processor 81, the processor 81 performs the following steps:
[0203] A creep stress relaxation simulation model for high-temperature pressure-bearing components is constructed, and multiple sets of parameter samples corresponding to the creep stress relaxation simulation model are determined.
[0204] Based on the parameter samples and the creep stress relaxation simulation model, the creep stress relaxation response results are calculated to obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data on stress changes over time;
[0205] For each set of parameter samples, based on the creep stress relaxation response results corresponding to the parameter samples, the creep stress relaxation behavior is fitted with parameters to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters.
[0206] Construct a parameter prediction model for each fitting parameter, and train each parameter prediction model based on the parameter values of the fitting parameters corresponding to each group of parameter samples.
[0207] Based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, the parameter values of the target creep stress relaxation response model are determined; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
[0208] In some embodiments, the construction of the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component includes:
[0209] A three-dimensional finite element model of the high-temperature pressure-bearing component is constructed; the data of the three-dimensional finite element model includes at least one of the component geometric information, material parameters, mesh generation, boundary conditions, and load conditions.
[0210] Based on the three-dimensional finite element model, heat transfer analysis and calculation were performed to obtain the steady-state temperature field of the high-temperature pressure-bearing component under the target working conditions.
[0211] Under the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model and creep stress relaxation behavior analysis is performed to obtain the mechanical response of the high-temperature pressure-bearing component in the cyclic loading and load holding stages;
[0212] Generate automated script files corresponding to the above steps of constructing the three-dimensional finite element model, heat transfer analysis calculation, and creep stress relaxation behavior analysis calculation; the automated script files are used to construct the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component.
[0213] In some embodiments, determining the multiple sets of parameter samples corresponding to the creep stress relaxation simulation model includes:
[0214] Based on the requirements of creep stress relaxation behavior analysis, a set of parameters for creep stress relaxation behavior analysis is determined, wherein the set of parameters includes at least one of material parameters, load conditions, and operating condition parameters;
[0215] Determine the value range of each parameter in the parameter set to obtain the parameter sampling space corresponding to each parameter.
[0216] Within the parameter sampling space, the parameter set is sampled using Latin hypercube sampling to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
[0217] In some embodiments, the parameter prediction model includes a genetic algorithm-optimized backpropagation neural network model; the input of the parameter prediction model includes at least one of material parameters, load conditions, and operating condition parameters; the output of the parameter prediction model includes corresponding fitting parameters.
[0218] In some embodiments, training the parameter prediction model includes:
[0219] The parameter values of each group of parameter samples and their corresponding fitted parameters are normalized, and training sample sets and test sample sets are constructed based on the normalization results.
[0220] A genetic algorithm is used to globally optimize the initial weights and thresholds of the parameter prediction model; the optimization objects of the genetic algorithm include the connection weights and bias parameters of each layer of the parameter prediction model; the optimization objective of the genetic algorithm includes individual performance evaluation based on the fitness function.
[0221] Based on the intermediate model parameters obtained from the global optimization, the backpropagation algorithm is used to train the parameter prediction model. The intermediate model parameters of the parameter prediction model are iteratively updated by minimizing the function error between the predicted output and the true output until the convergence condition is met.
[0222] Based on the model parameters obtained when the convergence condition is met, as well as the input normalized parameters and the output denormalized parameters, the trained parameter prediction model is determined.
[0223] In some embodiments, the convergence condition includes an error index satisfying a preset condition; the error index includes root mean square error and / or coefficient of determination.
[0224] In some embodiments, determining the parameter values of the target creep stress relaxation response model based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed includes:
[0225] Based on the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, an input parameter vector is constructed; the input parameter vector includes at least one of material parameters, load conditions, and working condition parameters.
[0226] The input parameter vector is input into each of the trained parameter prediction models to obtain the parameter values output by each parameter prediction model.
[0227] Based on the parameter values output by each of the parameter prediction models, the parameter values of the target creep stress relaxation response model are determined.
[0228] The aforementioned memory 82 includes a main memory 821 and an external memory 822. The main memory 821, also known as internal memory, is used to temporarily store the computational data in the processor 81, as well as the data exchanged with external memory 822 such as a hard disk. The processor 81 exchanges data with the external memory 822 through the main memory 821.
[0229] The specific execution process of the above instructions can be referred to the steps of the parameter calibration method for the creep stress relaxation response model based on deep neural networks described in the embodiments of this application, and will not be repeated here.
[0230] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.
[0231] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of the creep stress relaxation response model parameter calibration method based on a deep neural network described in the above-described method embodiments. The storage medium can be a volatile or non-volatile computer-readable storage medium.
[0232] This application also provides a computer program product, including a computer program / instruction, which, when executed by the computer program / instruction processor, implements the creep stress relaxation response model parameter calibration method based on a deep neural network as provided in the various embodiments of this application.
[0233] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.
[0234] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.
[0235] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0236] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0237] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0238] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0239] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A method for calibrating parameters of a creep stress relaxation response model based on a deep neural network, characterized in that, The method includes: A creep stress relaxation simulation model for high-temperature pressure-bearing components is constructed, and multiple sets of parameter samples corresponding to the creep stress relaxation simulation model are determined. Based on the parameter samples and the creep stress relaxation simulation model, the creep stress relaxation response results are calculated to obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data on stress changes over time; For each set of parameter samples, based on the creep stress relaxation response results corresponding to the parameter samples, the creep stress relaxation behavior is fitted with parameters to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters. A parameter prediction model is constructed for each fitting parameter, and each parameter prediction model is trained based on the parameter values of the fitting parameters corresponding to each group of parameter samples. The parameter prediction model includes a genetic algorithm to optimize a backpropagation neural network model. The input of the parameter prediction model includes at least one of material parameters, load conditions, and working condition parameters. The output of the parameter prediction model includes the corresponding fitting parameters. Based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, the parameter values of the target creep stress relaxation response model are determined; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
2. The method of claim 1, wherein, The simulation model for creep stress relaxation corresponding to high-temperature pressure-bearing components includes: A three-dimensional finite element model of the high-temperature pressure-bearing component is constructed; the data of the three-dimensional finite element model includes at least one of the component geometric information, material parameters, mesh generation, boundary conditions, and load conditions. Based on the three-dimensional finite element model, heat transfer analysis and calculation were performed to obtain the steady-state temperature field of the high-temperature pressure-bearing component under the target working conditions. Under the steady-state temperature field, cyclic loading is applied to the three-dimensional finite element model and creep stress relaxation behavior analysis is performed to obtain the mechanical response of the high-temperature pressure-bearing component in the cyclic loading and load holding stages; Generate automated script files corresponding to the above steps of constructing the three-dimensional finite element model, heat transfer analysis calculation, and creep stress relaxation behavior analysis calculation; the automated script files are used to construct the creep stress relaxation simulation model corresponding to the high-temperature pressure-bearing component.
3. The method of claim 1, wherein, The determination of the multiple sets of parameter samples corresponding to the creep stress relaxation simulation model includes: Based on the requirements of creep stress relaxation behavior analysis, a set of parameters for creep stress relaxation behavior analysis is determined, wherein the set of parameters includes at least one of material parameters, load conditions, and operating condition parameters; Determine the value range of each parameter in the parameter set to obtain the parameter sampling space corresponding to each parameter. Within the parameter sampling space, the parameter set is sampled using Latin hypercube sampling to generate multiple sets of parameter samples corresponding to the creep stress relaxation simulation model.
4. The method of claim 1, wherein, The parameter prediction model is trained using the following steps: The parameter values of each group of parameter samples and their corresponding fitted parameters are normalized, and training sample sets and test sample sets are constructed based on the normalization results. A genetic algorithm is used to globally optimize the initial weights and thresholds of the parameter prediction model; the optimization objects of the genetic algorithm include the connection weights and bias parameters of each layer of the parameter prediction model; the optimization objective of the genetic algorithm includes individual performance evaluation based on the fitness function. Based on the intermediate model parameters obtained from the global optimization, the backpropagation algorithm is used to train the parameter prediction model. The intermediate model parameters of the parameter prediction model are iteratively updated by minimizing the function error between the predicted output and the true output until the convergence condition is met. Based on the model parameters obtained when the convergence condition is met, as well as the input normalized parameters and the output denormalized parameters, the trained parameter prediction model is determined.
5. The method of claim 4, wherein, The convergence condition includes that the error index meets the preset condition; the error index includes the root mean square error and / or the coefficient of determination.
6. The method of claim 1, wherein, Based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, the parameter values of the target creep stress relaxation response model are determined, including: Based on the creep stress relaxation behavior analysis requirements of the scenario to be analyzed, an input parameter vector is constructed; the input parameter vector includes at least one of material parameters, load conditions, and working condition parameters. The input parameter vector is input into each of the trained parameter prediction models to obtain the parameter values output by each parameter prediction model. Based on the parameter values output by each of the parameter prediction models, the parameter values of the target creep stress relaxation response model are determined.
7. A device for calibrating parameters of a creep stress relaxation response model based on a deep neural network, characterized in that, The device includes: The first construction module is used to construct a creep stress relaxation simulation model corresponding to a high-temperature pressure-bearing component and determine multiple sets of parameter samples corresponding to the creep stress relaxation simulation model. The calculation module is used to calculate the creep stress relaxation response results based on the parameter samples and the creep stress relaxation simulation model, and obtain the creep stress relaxation response results corresponding to each set of parameter samples; the creep stress relaxation response results include historical data of stress changes over time; The fitting module is used to fit the creep stress relaxation behavior to each set of parameter samples based on the creep stress relaxation response results corresponding to the parameter samples, so as to obtain the creep stress relaxation response model corresponding to the parameter samples; the creep stress relaxation response model includes multiple fitting parameters. The second construction module is used to construct a parameter prediction model corresponding to each fitting parameter, and to train each parameter prediction model based on the parameter values of the fitting parameters corresponding to each group of parameter samples. The parameter prediction model includes a genetic algorithm-optimized backpropagation neural network model. The input of the parameter prediction model includes at least one of material parameters, load conditions, and working condition parameters. The output of the parameter prediction model includes the corresponding fitting parameters. The third construction module is used to determine the parameter values of the target creep stress relaxation response model based on the trained parameter prediction model and the creep stress relaxation behavior analysis requirements of the scenario to be analyzed; the target creep stress relaxation response model is used to characterize the creep stress relaxation response of the high-temperature pressure-bearing component in the scenario to be analyzed.
8. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 6.
9. A computer readable storage medium having stored thereon a computer program, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.