A rheological stress prediction method, a hot processing diagram construction method and system

Abstract

This invention relates to the field of hot working process design and microstructure property control technology, specifically to a method for predicting rheological stress, a method for constructing hot working diagrams, and a system. The method includes: creating a constitutive model of the target material, which simulates the rheological stress variation of the target material during hot working; after the constitutive model is created, a residual sample set is created based on the constitutive model; the preset residual network model is optimized based on the residual sample set to obtain an optimized residual network model; the input parameters of the optimized residual network model include: working conditions, and the output parameters include: residuals; the working conditions to be measured are input into the constitutive model to obtain the predicted rheological stress; the working conditions to be measured are input into the optimized residual network model to obtain the residuals; the final rheological stress is calculated based on the predicted rheological stress and the residuals. This invention can reduce the error in rheological stress prediction and improve the repeatability and stability of hot working diagrams.

Description

Technical Field

[0001] This invention relates to the field of thermal processing design and microstructure control technology, specifically to a method for predicting rheological stress, a method for constructing thermal processing diagrams, and a system thereof. Background Technology

[0002] Predicting the rheological stress during the hot deformation process of metallic materials (or alloys) is the foundation for hot working process design and microstructure and property control.

[0003] In the technical field of rheological stress prediction and thermal processing diagram construction, traditional techniques have proposed some methods to improve the stability of thermal processing diagrams.

[0004] For example, patent application CN108595827A proposes a C Mn The invention first describes the evolution mechanism of hot deformation microstructure and the determination method of hot working properties of Al high-strength steel. Mn High-temperature compression tests were conducted on Al high-strength steel to obtain the true stress of the steel. True strain curve data were used to establish a rheological stress prediction model for steel. The chosen model was a constitutive model with a physical basis, based on creep theory and considering the relationship between Young's modulus and the self-diffusion coefficient of austenite with temperature. This constitutive model accurately predicted the rheological stress of steel. A hot deformation processing diagram of the steel was created, and the microstructure evolution mechanism in different regions of the processing diagram was determined by combining it with microstructure analysis. By combining the hot deformation constitutive model and the processing diagram, the hot deformation rheological stress and hot deformation power dissipation efficiency under arbitrary deformation conditions were analyzed, thereby obtaining the corresponding microstructure evolution mechanism and hot working performance information. The results are of great significance for the control of the hot working process of high-strength steel.

[0005] For example, patent application CN114678088A proposes an automated method for fitting a compensated Arrhenius constitutive model to a DMM heat treatment diagram, including the following steps: 1) labeling and removing unreasonable data in the experimental data; 2) sorting the data and outputting the initial rheological curve; 3) performing friction compensation on the data; 4) performing temperature compensation on the data; 5) determining the number of strain compensation operations based on the user-defined minimum compensation strain, maximum compensation strain, and compensation interval, embedding it into the heat treatment diagram module, and performing iterative solution output; 6) establishing corresponding polynomials between all constitutive parameters and strain to obtain the constitutive equation expression established after compensation; 7) using an iterative algorithm to solve for the stress value in reverse and comparing it with the experimental value.

[0006] For example, patent application CN117558374A proposes a method and apparatus for determining parameters of a material thermal deformation constitutive model, comprising: acquiring multiple sets of rheological stress data corresponding to the material under different strain conditions; constructing a material thermal deformation constitutive model based on the different strain conditions and the rheological stress data, wherein the material thermal deformation constitutive model contains multiple different constitutive parameters; obtaining the optimal solution set of constitutive parameters under different strain conditions based on the material thermal deformation constitutive model; determining the fitting polynomial corresponding to the optimal solution of the constitutive parameters in the optimal solution set of the constitutive parameters based on the optimal solution set of the constitutive parameters; and obtaining the constitutive parameter values ​​of the material thermal deformation constitutive model under different strain conditions based on the fitting polynomial.

[0007] However, the aforementioned traditional techniques often suffer from problems such as large global fitting errors and unstable extrapolation in wide temperature, wide strain rate or strongly nonlinear material systems, and may produce prediction results that do not meet physical trends, which in turn leads to problems such as speckled unstable regions and poor repeatability in the processing diagram.

[0008] In response, patent application CN114818437A proposes an optimization method for the isothermal forging process of integral titanium alloy bladed disks. The method includes: establishing a constitutive model of the influence of different compositions of the titanium alloy on relative flow stress; designing the billet size, establishing the material constitutive properties, and using a three-dimensional finite element numerical simulation system for plastic deformation and heat transfer to obtain simulation diagrams of the filling properties, stress field, strain field, and temperature field distribution of the pre-forging and final forging under different billet conditions; and simulation diagrams of the influence of the forging temperature, forging speed, reduction, and friction factor on the distribution of stress field, strain field, and temperature field of the pre-forging and final forging, thereby obtaining the optimal isothermal forging process parameters.

[0009] Furthermore, patent application CN118737331A proposes an optimization method for the superplastic isothermal forging process of titanium alloys. Before superplastic forging, appropriate superplastic experimental parameters are selected, a damage constitutive model is established, and superplastic forming damage simulation is performed. Based on the damage simulation data, the damage prediction of titanium alloy parts during high-temperature forming can be achieved quickly. Finally, the process parameters are optimized based on the damage prediction results, which can effectively optimize the process parameters.

[0010] However, the current method for predicting rheological stress is still ineffective, or rather, its reliability is limited.

[0011] Therefore, a more efficient method for predicting rheological stress is still needed to construct more accurate thermal processing diagrams. Summary of the Invention

[0012] The purpose of this invention is to provide a method for predicting rheological stress, a method for constructing heat treatment diagrams, and a system that partially solves or alleviates the above-mentioned deficiencies in the prior art, and can reduce the error of rheological stress prediction under a wide range of working conditions, and improve the repeatability and stability of heat treatment diagrams.

[0013] To solve the aforementioned technical problems, the present invention specifically adopts the following technical solution: A first aspect of the present invention is to provide a method for predicting rheological stress, comprising the steps of: S101, Create a constitutive model for the target material, which is used to simulate the change law of rheological stress of the target material during hot working; wherein, the constitutive model takes working conditions as input parameters, including: plastic strain ε, strain rate, and temperature T; the output parameters of the constitutive model include: predicted rheological stress σArr; S102, after the constitutive model is created, a residual sample set is created based on the constitutive model; wherein, S102 includes the following steps: S1021, Obtain the actual rheological stress of the target material under multiple sets of the aforementioned working conditions; S1022, input multiple sets of the working conditions into the constitutive model to obtain the predicted rheological stress σArr under the working conditions accordingly; S1023, construct a residual sample set, the residual sample set includes: multiple sets of residual samples, and the residual samples include: the predicted rheological stress σArr, the actual rheological stress, and the operating conditions; S103, optimize the preset residual network model according to the residual sample set to obtain an optimized residual network model; the input parameters of the optimized residual network model include: operating conditions, and the output parameters include: residual; the residual refers to the difference between the predicted rheological stress σArr and the actual rheological stress; S104, Input the working conditions to be measured into the constitutive model to obtain the predicted rheological stress; S105, Input the test conditions into the optimized residual network model to obtain the residual; S106, the final rheological stress is calculated based on the predicted rheological stress and the residual.

[0014] In some embodiments, the constitutive model is an Arrhenius-type constitutive model.

[0015] In some embodiments, S101 includes: S1011, Obtain model sample data, the model sample data including: multiple sets of sample working conditions, and rheological stress corresponding to the sample working conditions; S1012, the model sample data is fitted using a linear regression method to obtain initial simulation values, which include: lnA, α, n, and Q; where lnA, α, and n are material constants, and Q is the deformation activation energy; S1013, Set the constitutive model according to the initial simulation values; S1014, Perform a first optimization operation on the constitutive model to obtain an optimized constitutive model.

[0016] In some embodiments, the first optimization operation is a particle swarm optimization method, and correspondingly, S1014 includes: (1) After performing the optimization operation, the optimization monitoring index is obtained, including: global best fitness, particle swarm diversity or particle average velocity norm. (2) Determine whether the optimized monitoring indicators meet the stagnation warning conditions; wherein, If the result of (2) is yes, then proceed with the following steps: (3) Determine whether the performance index of the constitutive model is less than a preset first threshold; wherein, If the result of (3) is negative, then proceed with the following steps: (4) Generate an alarm signal and start S102.

[0017] In some embodiments, prior to step (4), the method further includes: (5) Obtain the particles located in the current search space and reset some of the particles in the updated search space to obtain the updated particle swarm, wherein the range of the updated search space is larger than the current search space; (6) Perform the optimization operation n times again based on the particle swarm; (7) Determine whether the current global optimal solution has been updated. If yes, then remove the alarm signal and continue to perform the optimization operation. If no, then allow to proceed to step (4).

[0018] In some embodiments, in S1014, a first optimization operation is performed with the goal of reducing the error between the predicted rheological stress and the actual rheological stress.

[0019] In some embodiments, preceding S102 includes: Obtain the performance metrics of the constitutive model; When the performance index is lower than the preset second threshold, then proceeding to S102 is allowed; And / or, the performance metric is the prediction accuracy of the constitutive model; or, the performance metric is the optimization efficiency of the constitutive model within the optimization cycle.

[0020] In some embodiments, it also includes: Verify the first accuracy of the constitutive model in the low temperature range and the second accuracy in the high temperature range; wherein, the low temperature range is the temperature range below the low temperature threshold, and the value range of the low temperature threshold includes: 200℃-400℃; The prediction accuracy is determined based on the first accuracy and the second accuracy.

[0021] The present invention also provides a method for constructing a thermal processing diagram, comprising the following steps: S201, establish a model mesh within the process range; and predict the rheological stress of each mesh, wherein the rheological stress is obtained by the method described in any embodiment of the present invention; S202, calculate the power dissipation parameters under the true strain ε section in the model mesh according to the working conditions and the rheological stress; wherein, the power dissipation parameters include: strain rate sensitivity index m, power dissipation coefficient η, and instability criterion ξ; S203 generates a thermal processing diagram based on power consumption loss parameters.

[0022] The present invention also provides a rheological stress prediction system, comprising: The constitutive model creation module is used to create a constitutive model of the target material. The constitutive model is used to simulate the rheological stress variation law of the target material during hot working. The constitutive model takes working conditions as input parameters, including: plastic strain ε, strain rate, and temperature T. The output parameters of the constitutive model include: predicted rheological stress σArr. A residual sample set creation module is used to create a residual sample set based on the constitutive model after the constitutive model has been created; wherein, the residual sample set creation module includes: The actual rheological stress acquisition unit is used to acquire the actual rheological stress of the target material under multiple sets of working conditions. The predictive rheological stress acquisition unit is used to input multiple sets of the working conditions into the constitutive model to obtain the predicted rheological stress σArr under the working conditions. A residual sample set construction unit is used to construct a residual sample set, which includes multiple sets of residual samples, and the residual samples include the predicted rheological stress σArr, the actual rheological stress, and the operating conditions. The residual network model optimization module is used to optimize a preset residual network model based on the residual sample set to obtain an optimized residual network model; the output parameters of the optimized residual network model include: residuals; The predictive rheological stress acquisition module is used to input the test conditions into the constitutive model to obtain the predicted rheological stress. The residual acquisition module is used to input the test conditions into the optimized residual network model to obtain the residuals; The rheological stress calculation module is used to calculate the final rheological stress based on the predicted rheological stress and the residual.

[0023] Beneficial technical effects: This application provides a progressive dual-model training mechanism, which decouples the training process of the constitutive model and the residual model, thereby avoiding or reducing optimization difficulties that may occur during the coupled optimization process of the constitutive model and the residual model.

[0024] Furthermore, the constitutive model's parameters include the deformation activation energy Q and material constants (α, A, n). Under given temperature and strain rate conditions, the material's flow stress can be solved based on these parameters. A residual model is further introduced to compensate for the nonlinear components. This decoupling of linear and nonlinear components enables more systematic modeling and prediction of material flow stress behavior in complex scenarios such as wide temperature ranges, wide strain rates, or strongly nonlinear material systems. For example, by decoupling and combining the constitutive model and the residual model, it is possible to capture the linear trend of flow stress while compensating for complex nonlinear changes (such as the anomalous hardening of aluminum alloys at lower temperatures) or inflection points, thereby improving the accuracy of flow stress prediction.

[0025] This invention also proposes a dual verification mechanism for constitutive model optimization operations. Specifically, this invention determines whether the optimization operation has stalled based on whether the optimization monitoring index meets the stagnation warning condition. Furthermore, it can also determine whether the optimization operation meets the optimization requirements based on the performance index of the constitutive model, thereby further determining whether and when to introduce a residual model. This approach can provide a more suitable time for introducing the residual model, or in other words, improve the utilization rate of the constitutive model and reduce unnecessary model training costs, thus controlling the model training cost while ensuring the accuracy of rheological stress prediction. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0027] Figure 1 A flowchart illustrating a rheological stress prediction method provided by the present invention; Figure 2This is a schematic diagram of the structure of a rheological stress prediction system provided by the present invention; Figure 3 A schematic block diagram of the structure of a computer device provided by the present invention; Figure 4 This is another flowchart illustrating a rheological stress prediction method provided by the present invention. Figure 5a Nine sets of example graphs showing the variation curves of real strain and real stress provided by this invention; Figure 5b This invention provides six sets of example graphs showing the variation curves of real strain and real stress. Figure 6 Example graph showing the variation of root mean square error (RMSE) with the number of iterations provided by this invention; Figure 7 This is an example of the parameter optimization results of the constitutive model provided by the present invention; Figure 8a The RMSE results provided by this invention are evaluated using 10-fold cross-validation for error assessment. Figure 8b Error evaluation of the MAE results provided by this invention using 10-fold cross-validation; Figure 8c Error evaluation of the MAPE results provided by this invention using 10-fold cross-validation; Figure 8d R provided by the present invention 2 The results were evaluated using 10-fold cross-validation for error assessment. Figure 9a The actual stress surface diagram provided for this invention; Figure 9b The strain rate sensitivity index surface plot provided by this invention; Figure 9c The energy efficiency surface plot provided for this invention; Figure 10 This is an example diagram of the heat treatment process provided by the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0029] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0030] In this document, the terms "upper," "lower," "inner," "outer," "front," "rear," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the present invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0031] In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0032] In this document, "and / or" includes any and all combinations of one or more of the listed related items.

[0033] In this article, "multiple" means two or more, that is, it includes two, three, four, five, etc.

[0034] As used in this specification, the term "about" typically means + / -5% of the value, more typically + / -4% of the value, more typically + / -3% of the value, more typically + / -2% of the value, even more typically + / -1% of the value, and even more typically + / -0.5% of the value.

[0035] In this specification, certain embodiments may be disclosed in a range-bound format. It should be understood that this "range-bound" description is merely for convenience and brevity and should not be construed as a rigid limitation on the disclosed range. Therefore, the description of a range should be considered as having specifically disclosed all possible subranges and the individual numerical values ​​within those ranges. For example, a description of the range 1-6 should be considered as having specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., and the individual numbers within those ranges, such as 1, 2, 3, 4, 5, and 6. This rule applies regardless of the breadth of the range.

[0036] The applicant noted that currently, low-temperature (e.g., 200-400℃) hot working of aluminum alloys is a key process window for automotive lightweighting and aerospace precision forming, but traditional constitutive models have insufficient prediction accuracy in this temperature range, which makes it impossible to meet actual processing requirements.

[0037] To address this issue, this invention proposes a method for predicting the rheological stress of aluminum alloys, particularly suitable for low-temperature hot working scenarios. Specifically, this application proposes a decoupled training approach for the constitutive model and the residual model, which can effectively alleviate the problem of insufficient prediction accuracy for alloys in the low-temperature range.

[0038] Example 1: Please see Figure 1 This invention proposes a method for predicting rheological stress, comprising the following steps: S101, Create a constitutive model for the target material, which is used to simulate the rheological stress variation of the target material during hot working; wherein, the constitutive model uses working conditions as input parameters, and the working conditions include: plastic strain ε, strain rate Temperature T; the output parameters of the constitutive model include: predicted rheological stress σ Arr ; S102, after the constitutive model is created, a residual sample set is created based on the constitutive model; wherein, S102 includes the following steps: S1021, Obtain the actual rheological stress of the target material under multiple sets of the aforementioned working conditions; S1022, input multiple sets of the aforementioned working conditions into the constitutive model to obtain the predicted rheological stress σ under the corresponding working conditions. Arr ; S1023, Construct a residual sample set, the residual sample set including: multiple sets of residual samples, and the residual samples including: the predicted rheological stress σ Arr The actual rheological stress and the operating conditions; S103, optimize the preset residual network model based on the residual sample set to obtain an optimized residual network model; the input parameters of the optimized residual network model include: operating conditions, and the output parameters include: residuals; the residuals refer to the predicted rheological stress σ. Arr The difference between the actual rheological stress and the actual rheological stress; S104, Input the working conditions to be measured into the constitutive model to obtain the predicted rheological stress; S105, Input the test conditions into the optimized residual network model to obtain the residual; S106, the final rheological stress is calculated based on the predicted rheological stress and the residual.

[0039] In some embodiments, the target material may refer to metallic materials, such as aluminum alloys, titanium alloys, magnesium alloys, etc. It should be understood that different target materials have different material properties, therefore, corresponding constitutive models can be created for different types of target materials.

[0040] In some embodiments, the constitutive model is an Arrhenius-type constitutive model.

[0041] In some embodiments, S101 includes: S1011, Obtain model sample data, the model sample data including: multiple sets of sample working conditions, and rheological stress corresponding to the sample working conditions; S1012, the model sample data is fitted using a linear regression method to obtain initial simulation values, which include: lnA, α, n, and Q; where lnA, α, and n are material constants, and Q is the deformation activation energy; S1013, Set the constitutive model according to the initial simulation values; S1014, Perform a first optimization operation on the constitutive model to obtain an optimized constitutive model.

[0042] In some embodiments, model sample data may refer to multiple sets of sample operating conditions corresponding to a specific material, and the rheological stress corresponding to the sample operating conditions.

[0043] In some embodiments, the initial simulation values ​​can be obtained by fitting the model sample data using a linear regression method. Of course, the specific fitting method can refer to existing technologies, and this invention does not limit it.

[0044] As an example, the first optimization operation refers to the global optimization of the baseline parameter θ (i.e. the initial simulation value of the constitutive model) under the constraints of the physical feasible region (A>0, α>0, n>0, Q>0); In some embodiments, the first optimization operation may include the steps of: Step 1: Define the optimization objective: Minimize the error (such as root mean square error) between the flow stress predicted by the constitutive model and the actual flow stress.

[0045] Step 2: Initialize parameter combinations: Under the constraints of the physical feasible region (A>0, α>0, n>0, Q>0), generate multiple sets of initial parameter combinations (A, α, n, Q) as candidate solutions.

[0046] Step 3: Simulation calculation: Substitute each set of candidate parameters into the constitutive model and calculate the predicted value of rheological stress under the corresponding working conditions.

[0047] Step 4: Evaluation and Screening: Compare the errors between the predicted and actual rheological stresses of each candidate parameter combination, and select the combination with the smallest error as the current optimal solution.

[0048] Step 5: Update the global optimal solution: If the error of the current optimal solution is less than the error of the historical global optimal solution, then the current optimal solution is taken as the global optimal solution.

[0049] Step 6: Generate new candidate solutions: Based on the current global optimal solution, generate multiple new sets of candidate parameter combinations through preset adjustment strategies (such as fine-tuning parameters, local perturbations, etc.), and return to step 3.

[0050] Step 7: Iterative convergence: Repeat steps 3 to 6 until the global optimal solution no longer undergoes significant updates in consecutive iterations (i.e., the error tends to stabilize). The constitutive model parameters obtained at this point are the optimized optimal parameters.

[0051] In some embodiments, in S1014, a first optimization operation is performed with the goal of reducing the error between the predicted rheological stress and the actual rheological stress. The effect of the first optimization operation can be found in [reference needed]. Figure 7 .

[0052] In some embodiments, the second optimization operation refers to optimizing the residual model to optimize different input features (i.e., ε) through the residual model. ,T), providing different residual correction values. That is, the conditional inputs of the residual model include: ε, T; and the parameters to be optimized (i.e., hyperparameters); the output includes: residuals; In some embodiments, the second optimization operation may include the steps of: Step 1: Define the optimization objective: Minimize the error between the residual correction value output by the residual model and the true residual (i.e., the difference between the actual rheological stress and the predicted rheological stress).

[0053] Step 2: Construct the residual sample set: This involves combining the operating conditions (ε, (T) and constitutive model predicted value σ Arr Using these as input features and the corresponding true residual values ​​as supervision labels, a dataset for training the residual model is constructed and divided into training, validation, and test sets according to a set ratio.

[0054] Step 3: Set the model architecture and hyperparameter space: Preset the initial architecture of the residual network model (such as the number of hidden layers, number of neurons, or activation function type), and define the hyperparameters to be optimized φ (including learning rate, batch size, regularization coefficient, or optimizer selection), and clarify the candidate range or distribution of each hyperparameter.

[0055] Step 4: Perform hyperparameter optimization: Use a preset optimization strategy (such as grid search, random search, Bayesian optimization or evolutionary algorithm), and use the performance indicators on the validation set (such as root mean square error RMSE, coefficient of determination R², etc.) as the evaluation criteria. Iteratively test different combinations of hyperparameters, gradually narrowing the search space until the hyperparameter configuration that makes the validation set performance optimal is found.

[0056] Step 5: Train the residual network model: Under the optimal hyperparameter configuration, use the training set to supervise the learning of the weight parameters of the residual network model (e.g., iteratively update through the backpropagation algorithm) so that the model can accurately fit the nonlinear mapping relationship from the input features to the residual values.

[0057] Step 6: Model Validation and Evaluation: Use the test set to perform a final performance evaluation on the trained residual network model to verify its ability to generalize and predict residual correction values ​​under unseen operating conditions.

[0058] Step 7: Output the optimized residual model: The residual network model that has undergone hyperparameter optimization and weight training and has the best performance on the validation set is determined as the final optimized residual model for rheological stress correction.

[0059] For example, R² (R squared, Coefficient of determination, or goodness of fit) reflects the accuracy of the model's fit to the data.

[0060] In some embodiments, the preset residual network model can be a neural network model.

[0061] In some embodiments, the optimized residual network model can be a dedicated network model for predicting residuals, which is trained by a preset residual network model and the residual sample set (or a residual sample set) based on the difference sample set between the predicted rheological stress and the actual rheological stress value calculated by the constitutive model.

[0062] In some embodiments, the final rheological stress can be calculated based on the predicted rheological stress obtained from the constitutive model and the residual obtained from the optimized residual network model. For example, the final rheological stress = predicted rheological stress + residual.

[0063] It should be understood that the technical ideas provided in patent applications CN114818437A and CN118737331A both use the residual as part of the optimized constitutive model, that is, they provide a single-model coupled optimization mechanism. For example, the residual between the experimental value of the flow stress of titanium alloy and the calculated value of the flow stress of the constitutive model is used as the objective function. The emphasis is on the optimization of constitutive model parameters based on the residual objective function.

[0064] Therefore, traditional alloy rheological stress prediction models tend to concentrate training resources on the overall training of the constitutive model. However, this application proposes a completely opposite approach to resource adjustment, which is to no longer take the constitutive model as the sole objective, but instead to segment and introduce training resources into the independent training of the constitutive model and the residual model.

[0065] In other words, quite differently, this invention provides a dual-model decoupling optimization mechanism. Specifically: In this embodiment, after the constitutive model is created, a residual sample set is created based on the constitutive model to train and optimize the obtained optimized residual network model. That is to say, the present invention decouples the training process of the constitutive model and the residual model, that is, it divides the training and optimization into two independent stages.

[0066] The applicant noted that traditional constitutive models often struggle to capture the complex changes in material rheological stress (such as abrupt changes, inflection points, and plateaus that occur in aluminum alloys at low temperatures). This application decouples the training process of the constitutive model from that of the residual model, allowing for deeper learning and simulation of linear and nonlinear characteristics using the constitutive and residual models, respectively.

[0067] Specifically, this embodiment uses a decoupled residual model to learn nonlinear variations that are difficult for constitutive models to describe, thereby improving the adaptability and accuracy of rheological stress prediction results.

[0068] Furthermore, this application provides a progressive dual-model training mechanism, which decouples the training processes of the constitutive model and the residual model, thereby avoiding or reducing optimization difficulties that may arise during the coupled optimization process of the constitutive model and the residual model.

[0069] Furthermore, the parameters of the constitutive model include the deformation activation energy Q and material constants (α, A, n). Under given temperature and strain rate conditions, the rheological stress of the material can be solved based on these parameters. A residual model is further introduced to compensate for the nonlinear components. This decoupling of linear and nonlinear components enables more systematic modeling and prediction of the rheological stress behavior of materials in complex scenarios such as wide temperature ranges, wide strain rates, or strongly nonlinear material systems.

[0070] By decoupling and combining the constitutive model and the residual model, the accuracy of rheological stress prediction can be improved by capturing the linear trend of rheological stress while compensating for complex nonlinear variations (such as the anomalous hardening of aluminum alloys at lower temperatures) or inflection points. Please see below. Figure 5a , Figure 5b ,in, Figure 5a , Figure 5b The horizontal axis represents the actual strain value, and the vertical axis represents the actual stress value.

[0071] This article will illustrate the rheological stress prediction and thermal processing diagram construction method provided by the present invention with specific examples: S1 Obtain the thermal deformation experimental dataset D (e.g., including operating conditions) = {(plastic strain ε, strain rate)} Temperature T, predicted rheological stress σ exp )}.

[0072] For example, hot compression experiments can be conducted on the target material under different working conditions, and the actual rheological stress under the corresponding working conditions can be recorded.

[0073] Plastic strain, also known as permanent strain, refers to the strain that remains in a material element after all stress has disappeared.

[0074] Among them, strain rate refers to the linear strain or shear strain that occurs per unit time.

[0075] S2. Establish the Arrhenius constitutive model.

[0076] In some embodiments, multiple sets of sample operating conditions and corresponding rheological stresses can be fitted using a linear regression method to obtain the initial simulation values ​​of the constitutive model.

[0077] For example, the constitutive model σArr=f(ε, ,T;θ), where θ at least includes lnA, α, n, Q (i.e., the initial simulation values).

[0078] Here, A refers to the pre-exponential factor, which is related to the entropy change or vibration frequency of the material and reflects the inherent characteristics of material deformation. lnA is the natural logarithm of A.

[0079] Where α is a correction factor related to the stress level (unit: MPa) - ¹).

[0080] Where n is the stress rate sensitivity exponent, which describes the sensitivity of strain rate to stress.

[0081] Where Q is the deformation activation energy (unit: kJ / mol) during hot deformation, which reflects the energy required for atomic motion or dislocations to overcome energy barriers during high-temperature deformation. The higher the Q value, the more sensitive the deformation is to temperature, that is, the more significant the effect of temperature changes on flow stress.

[0082] S3 First Stage PSO: Automatic optimization of parameters for the constitutive model (i.e., the first optimization operation).

[0083] Particle Swarm Optimization (PSO) is used to search for θ within the feasible region to minimize the prediction error of the constitutive model. The objective function is to minimize the RMSE (Root Mean Square Error) between the experimental values ​​(i.e., actual rheological stress) and the predicted values ​​(i.e., predicted rheological stress). Please refer to [link to relevant documentation]. Figure 6 (Where, the horizontal axis represents the number of iterations; the vertical axis represents the root mean square error).

[0084] Alternatively, in some embodiments, in S1014, a first optimization operation is performed with the goal of reducing the error between the predicted rheological stress and the actual rheological stress.

[0085] S4 calculates the predicted rheological stress σArr of the constitutive model and outputs the initial simulation values ​​lnA, α, n, and Q.

[0086] S5 constructs the residual dataset.

[0087] For example, calculating multiple sets of residuals Δσ=σ exp -σ Arr (i.e., the difference between the predicted rheological stress and the actual rheological stress), resulting in the residual dataset {(x i ,Δσ i The input feature x contains ε. ,T.

[0088] S6. Establish the residual model.

[0089] The residual model is trained using an Artificial Neural Network (ANN): ΔσANN = g(x;φ). Here, the hyperparameter φ includes at least the number of hidden layer neurons H and the regularization coefficient λ. The residual prediction model Δσ is then trained under the hyperparameter φ. ANN .

[0090] S7 Second Stage PSO: Automatic Optimization of ANN Hyperparameters (i.e., the second optimization operation).

[0091] The verification error was calculated using grouped K-fold cross-validation (grouped by temperature and / or strain rate conditions) and used as the PSO evaluation index. The objective function of the second optimization operation was the minimum RMSE between the experimental value (i.e., actual rheological stress) and the predicted value (i.e., predicted rheological stress).

[0092] K-Fold Cross-Validation is a commonly used model evaluation method. It divides the training set into k subsets (folds) of similar size. Each time, one fold is used as the validation set, and the remaining k-1 folds are used as the training set. This process is repeated k times, and the average of the final results is used as the model performance metric. This ensures that each sample is validated exactly once, effectively reducing the randomness of the evaluation and making full use of the data.

[0093] For example, please see Figures 8a-8d .

[0094] Figure 8a The horizontal axis represents the Cross-validation fold, with values ​​from 1 to 10, representing each fold of a 10-fold cross-validation system. The vertical axis represents RMSE (root mean square error) in megapascals (MPa), which measures the deviation between the model's predicted values ​​and the actual values; a smaller value indicates higher prediction accuracy.

[0095] Figure 8b The horizontal axis in the figure refers to the Cross-validation fold, and the vertical axis refers to the MAE (MPa) (mean absolute error), which is measured in megapascals. It is used to measure the average absolute deviation between the predicted value and the actual value. The smaller the value, the more accurate the prediction.

[0096] Figure 8c The horizontal axis in the figure refers to the Cross-validation fold, and the vertical axis refers to MAPE (%) (Mean Absolute Percentage Error), which is expressed as a percentage and is used to measure the relative magnitude of the prediction error. The smaller the value, the higher the relative accuracy of the prediction.

[0097] Figure 8d The horizontal axis in the figure refers to the cross-validation fold, and the vertical axis refers to R² (coefficient of determination), which is used to measure the goodness of fit of the model to the data. The value ranges from 0 to 1. The closer it is to 1, the stronger the model's ability to interpret the data and the better the fit.

[0098] S8 outputs the prediction results.

[0099] σ hyb =σ Arr +Δσ ANN ; Calculate and output RMSE, R 2(Reflects the proportion of the variability in the observed data explained by the model, i.e., goodness of fit), MAPE (Mean Absolute Percentage Error, the average percentage of the absolute value of the prediction error relative to the experimental value, reflecting the relative error level). σ Arr To predict rheological stress, Δσ ANN For the residual, σ hyb This represents the final rheological stress.

[0100] S9 process mesh and continuous stress field generation.

[0101] Establish (ε, within the preset process range) ,T) grid, calculate σ hyb A continuous stress field is formed.

[0102] S10 performs heat treatment diagram calculations based on DMM (Dynamic Material Model) or on the heat treatment diagram theory proposed by Prasad.

[0103] For example, on a fixed ε cross section, the lnσ_hyb-lnedot relationship (which refers to the natural logarithm of the rheological stress (σ_hyb) calculated using a hybrid model under fixed temperature (T) and fixed strain (ε) conditions, versus the strain rate (ε) After fitting, smoothing, or regularizing the functional relationship between the natural logarithms of and , we obtain m = . lnσ / lnedot, and calculate η=2m / (m+1), ξ=( m / ln ) / [m(m+1)]+m; output η diagram and ξ<0 unstable region, thus generating heat treatment diagram.

[0104] Where m refers to the strain rate sensitivity index. η (power dissipation efficiency) represents the proportion of energy used for microstructure evolution (such as dynamic recrystallization and superplasticity) during thermal deformation of the material to the total input energy. ξ (instability criterion) is an indicator for determining whether a material has undergone plastic instability (such as adiabatic shear banding, rheological localization, and cracking).

[0105] The hot working diagram refers to a diagram on a temperature-strain rate plane, displayed in the form of contour lines, showing the efficiency of internal energy dissipation and the risk of plastic instability in the target material under different processing conditions. As a process map, the hot working diagram itself is reliable and repeatable.

[0106] S11 Output Recommended Process Window The recommended processing window is output based on the region with high η and ξ>0.

[0107] In some embodiments, see Figure 4 The present invention also provides a method for predicting rheological stress, comprising the following steps: S201, Create a constitutive model for the target material, which is used to simulate the rheological stress variation of the target material during hot working; wherein, the constitutive model uses working conditions as input parameters, the working conditions including: plastic strain ε, strain rate Temperature T; the output parameters of the constitutive model include: predicted rheological stress σ Arr ; S202, Obtain the actual rheological stress of the target material under multiple sets of working conditions; S203, input multiple sets of the aforementioned working conditions into the constitutive model to obtain the predicted rheological stress σ under the corresponding working conditions. Arr ; S204, construct a residual sample set, the residual sample set includes: multiple sets of residual samples, and the residual samples include: predicted rheological stress, actual rheological stress and operating conditions. S205, using the residual sample set and the neural network model to train a residual model, the input parameters of the residual model include operating conditions, and the output parameters include: residuals; S206, Input the working conditions to be measured into the constitutive model and the residual model respectively to obtain the predicted rheological stress and the residual; S207, the final rheological stress is calculated based on the predicted rheological stress and the residual.

[0108] Example 2: This invention provides a method for predicting rheological stress, including performing a first optimization operation on a constitutive model, wherein the first optimization operation is a particle swarm optimization method, and correspondingly, S1014 includes: (1) After performing the optimization operation, the optimization monitoring index is obtained, including: global best fitness, particle swarm diversity or particle average velocity norm. (2) Determine whether the optimized monitoring indicators meet the stagnation warning conditions; wherein, If the result of (2) is yes, then proceed with the following steps: (3) Determine whether the performance index of the constitutive model is less than a preset first threshold; wherein, If the result of (3) is negative, then proceed with the following steps: (4) Generate an alarm signal and start S102.

[0109] In the process of searching for the optimal combination of parameters in the search space using PSO (Particle Swarm Optimization), each particle has a fitness value, which represents the quality of the solution corresponding to that particle. The global optimal fitness refers to the fitness value corresponding to the best position experienced by all particles in the entire particle swarm, that is, the fitness of the optimal solution found.

[0110] For example, if the global optimal fitness changes very little (e.g., less than a certain threshold) or no longer changes in consecutive optimization operation iterations, then the first optimization operation can be considered to satisfy the stagnation warning condition.

[0111] Among them, particle swarm diversity is used to define the breadth of particle swarm distribution within the search range, reflecting the search's exploratory capabilities.

[0112] For example, high particle swarm diversity indicates that the particles are dispersed, which may mean that the model is still searching extensively; low diversity indicates that the particles are clustered, which may mean that the model is close to convergence or trapped in a local optimum, i.e., it meets the stagnation warning condition.

[0113] The average particle velocity norm is used to define the average value of the norms of all particle velocity vectors.

[0114] For example, the larger the value of the particle's average velocity norm, the more active the particle is, and it is still actively exploring new regions; the smaller the value of the particle's average velocity norm, the slower the particle moves, and it may be close to convergence or stagnate, which means that the stagnation warning condition is met.

[0115] In this embodiment, the present invention proposes a dual verification mechanism for constitutive model optimization operations. Specifically, the present invention determines whether the optimization operation has stalled based on whether the optimization monitoring index meets the stagnation warning condition. Furthermore, it can also determine whether the optimization operation meets the optimization requirements based on the performance index of the constitutive model, thereby further determining whether and when to introduce a residual model. This provides a more suitable time for introducing the residual model, thereby improving the utilization rate of the constitutive model and reducing unnecessary model training costs (such as avoiding the premature introduction of the residual model, which would prevent the constitutive model from achieving its due effectiveness). In other words, the present invention can control the cost of model training while ensuring the accuracy of rheological stress prediction.

[0116] In other words, this dual verification mechanism can stop the optimization of the constitutive model at an appropriate time, that is, allocate computing resources more efficiently, so as to avoid the constitutive model from getting trapped in local optima or overfitting to a certain extent, thereby achieving a better balance between the degree of model optimization and computing cost.

[0117] In some embodiments, prior to step (4), the method further includes: (5) Obtain the particles located in the current search space and reset some of the particles in the updated search space to obtain the updated particle swarm, wherein the range of the updated search space is larger than the current search space; (6) Perform the optimization operation n times again based on the particle swarm; (7) Determine whether the current global optimal solution has been updated. If yes, then remove the alarm signal and continue to perform the optimization operation. If no, then allow to proceed to step (4).

[0118] In this embodiment, it is preferable to dynamically expand the search space of PSO and retry optimizing the constitutive model before introducing the residual model, thereby appropriately restricting the introduction of the residual model and avoiding its premature introduction.

[0119] In other words, this embodiment provides a restrictive mechanism for introducing residual models. It can improve prediction accuracy by training residual models separately, and can also appropriately restrict the introduction of residual models by optimizing three factors: monitoring indicators, model performance indicators, and search space limitations. This allows for the improvement of rheological stress prediction accuracy through residual models while also controlling computational and training costs.

[0120] In some embodiments, preceding S102 includes: Obtain the performance metrics of the constitutive model; When the performance index is lower than the preset second threshold, then proceeding to S102 is allowed; And / or, the performance metric is the prediction accuracy of the constitutive model; or, the performance metric is the optimization efficiency of the constitutive model within the optimization cycle.

[0121] In some embodiments, the preset second threshold and the preset first threshold may be equal or unequal. This embodiment does not impose any restrictions on this.

[0122] In some embodiments, the prediction accuracy of a constitutive model refers to the difference between the model-predicted rheological stress value and the actual experimental measurement value. For example, the larger the difference, the lower the prediction accuracy.

[0123] For example, the lower the prediction accuracy, the lower the performance metric of the constitutive model.

[0124] In some embodiments, the optimization efficiency of a constitutive model within an optimization cycle refers to the ratio between the rate of performance improvement of the constitutive model and the resource consumption during the optimization process. For example, optimization efficiency = fitness improvement value (i.e., the difference between the current fitness value and the initial fitness value) / number of iterations.

[0125] For example, the lower the optimization efficiency, the lower the performance index of the constitutive model.

[0126] The applicant noted that the deformation mechanism of aluminum alloys in key process windows (such as the low-temperature range of 200-400°C) of automotive lightweighting and aerospace precision forming is complex. In this regard, the present invention can determine the model accuracy for different temperature ranges separately. That is, preferably by calculating the first accuracy for the low-temperature range and the second accuracy for the high-temperature range separately. This piecewise calculation mechanism based on the accuracy of temperature range helps to capture the complex deformation mechanism that may exist in a specific temperature range and can clarify in which temperature range the constitutive model is more likely to have prediction accuracy problems.

[0127] In some embodiments, it also includes: Verify the first accuracy of the constitutive model in the low temperature range and the second accuracy in the high temperature range; wherein, the low temperature range is the temperature range below the low temperature threshold, and the value range of the low temperature threshold includes: 200℃-400℃; The prediction accuracy is determined based on the first accuracy and the second accuracy.

[0128] In some embodiments, corresponding weights can be preset for the first accuracy and the second accuracy, and the weighted sum of the accuracy and the corresponding weights can be used as the prediction accuracy.

[0129] In some embodiments, the weights of the first accuracy rate and the second accuracy rate can be set according to the alloy processing specifications (such as artificially set key heating temperature ranges).

[0130] For example, the key heating temperature range of the processed material can be determined to be the low-temperature range based on its physical properties, in which case the weight of the first accuracy rate can be increased.

[0131] In some embodiments, calculating the first accuracy in the low-temperature range and the second accuracy in the high-temperature range separately can help capture the significant defects of the constitutive model for some materials in the low-temperature range.

[0132] Furthermore, if the initial accuracy of the constitutive model is low, the search range of the constitutive model can be updated. If the global optimal solution of the constitutive model is not updated after expanding the search range, this embodiment preferably allows the residual model to focus on learning the residual sample set in the low-temperature range, or assigns higher weights to the sample data in the low-temperature range when training the residual model.

[0133] Alternatively, if the global optimal solution of the constitutive model is updated after expanding the search scope, more data from the low-temperature sample set can be used to further train the constitutive model.

[0134] It should be understood that by calculating the prediction accuracy (i.e., a model performance index of the constitutive model) separately in the high / low temperature range, this invention can provide more targeted directions for improving the training of constitutive and residual models, such as increasing the weight of low temperature sample data, thereby improving the adaptability of rheological stress to different material properties.

[0135] Example 3: Please see Figure 10 In some embodiments, the present invention provides a method for constructing a thermal processing map, comprising the steps of: S201, establish a model mesh within the process range; and predict the rheological stress of each mesh, wherein the rheological stress is obtained by the method described in any embodiment of the present invention; S202, calculate the power dissipation parameters under the true strain ε section in the model mesh according to the working conditions and the rheological stress; wherein, the power dissipation parameters include: strain rate sensitivity index m, power dissipation coefficient η, and instability criterion ξ; For example, Figure 10 In the middle, the gray area represents the unstable region.

[0136] In some embodiments, see Figures 9a-9c X-axis lg( ) (s - ¹) All refer to the logarithm of the strain rate, reflecting the speed of deformation. The Y-axis refers to the deformation temperature, ranging from 580 to 720 K (approximately 307 to 447 °C). The Z-axis refers to the actual strain, reflecting the degree of deformation.

[0137] Figure 9a This is a True Stress surface plot. The True Stress (MPa) ranges from blue (10 MPa) to red (238 MPa), representing the actual stress magnitude of the material under the corresponding process conditions.

[0138] Figure 9b This is a surface plot of the strain rate sensitivity index (m). The m value ranges from 0.04 to 0.31. m is the strain rate sensitivity index, which reflects the stress response characteristics of a material when the strain rate changes. The higher the m value, the better the material's ability to deform uniformly at high temperatures, and the less likely it is to develop local necking.

[0139] Figure 9c This is an energy efficiency (η) surface plot. η is the energy efficiency in the hot working diagram, representing the proportion of energy used for plastic deformation during the material deformation process to the total input energy. The higher the η value, the more fully the energy is utilized, the more efficient the processing, and the fewer internal defects (such as dislocations and pores) in the material, resulting in better performance.

[0140] S203 generates a thermal processing diagram based on power consumption loss parameters.

[0141] It should be understood that the rheological stress predicted by the method described in any embodiment of the present invention can reduce the sensitivity of the derivative to noise / discrete points in the DMM machining diagram calculation, and improve the stability and repeatability of the machining diagram.

[0142] Furthermore, by performing PSO automatic optimization on the constitutive model and residual model respectively, manual parameter tuning can be reduced, which helps to realize the automated process from constitutive fitting to processing diagram output.

[0143] Example 4: Please see Figure 2 In some embodiments, the present invention provides a rheological stress prediction system, comprising: The constitutive model creation module is used to create a constitutive model of the target material. This constitutive model simulates the rheological stress variation of the target material during hot working. The constitutive model uses working conditions as input parameters, including: plastic strain ε and strain rate. Temperature T; the output parameters of the constitutive model include: predicted rheological stress σ Arr ; A residual sample set creation module is used to create a residual sample set based on the constitutive model after the constitutive model has been created; wherein, the residual sample set creation module includes: The actual rheological stress acquisition unit is used to acquire the actual rheological stress of the target material under multiple sets of working conditions. The predictive rheological stress acquisition unit is used to input multiple sets of the working conditions into the constitutive model to obtain the predicted rheological stress σArr under the working conditions. A residual sample set construction unit is used to construct a residual sample set, which includes multiple sets of residual samples, and the residual samples include the predicted rheological stress σArr, the actual rheological stress, and the operating conditions. The residual network model optimization module is used to optimize a preset residual network model based on the residual sample set to obtain an optimized residual network model; the output parameters of the optimized residual network model include: residuals; The predictive rheological stress acquisition module is used to input the test conditions into the constitutive model to obtain the predicted rheological stress. The residual acquisition module is used to input the test conditions into the optimized residual network model to obtain the residuals; The rheological stress calculation module is used to calculate the final rheological stress based on the predicted rheological stress and the residual.

[0144] It should be understood that the rheological stress prediction system provided by the present invention can be used to implement the method steps described in any embodiment of the present invention.

[0145] In some embodiments, this application also provides a schematic block diagram of the structure of a computer device, please see... Figure 3 Computer programs can be used in situations such as Figure 3 It runs on the computer device shown. Figure 3 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory may include non-volatile storage media and internal memory. The non-volatile storage media may store an operating system and computer programs. The computer programs include program instructions that, when executed, cause the processor to perform arbitrary methods. The processor provides computational and control capabilities to support the operation of the entire computer device. The internal memory provides an environment for the execution of the computer programs in the non-volatile storage media; when executed by the processor, these programs cause the processor to perform arbitrary methods. The network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structures shown are merely block diagrams of a portion of the structure related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. It should be understood that the processor may be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0146] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0148] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for predicting rheological stress, characterized in that, Including the following steps: S101, Create a constitutive model for the target material, which is used to simulate the rheological stress variation of the target material during hot working; wherein, the constitutive model uses working conditions as input parameters, and the working conditions include: plastic strain ε, strain rate Temperature T; the output parameters of the constitutive model include: predicted rheological stress σ Arr ; S102, after the constitutive model is created, a residual sample set is created based on the constitutive model; wherein, S102 includes the following steps: S1021, Obtain the actual rheological stress of the target material under multiple sets of working conditions; S1022, input multiple sets of the aforementioned working conditions into the constitutive model to obtain the predicted rheological stress σ under the corresponding working conditions. Arr ; S1023, Construct a residual sample set, the residual sample set including: multiple sets of residual samples, and the residual samples including: the predicted rheological stress σ Arr The actual rheological stress and the operating conditions; S103, optimize the preset residual network model based on the residual sample set to obtain an optimized residual network model; the input parameters of the optimized residual network model include: operating conditions, and the output parameters include: residuals; the residuals refer to the predicted rheological stress σ. Arr The difference between the actual rheological stress and the actual rheological stress; S104, Input the working conditions to be measured into the constitutive model to obtain the predicted rheological stress; S105, Input the test conditions into the optimized residual network model to obtain the residual; S106, the final rheological stress is calculated based on the predicted rheological stress and the residual.

2. The method according to claim 1, characterized in that, The constitutive model is an Arrhenius-type constitutive model.

3. The method according to claim 1, characterized in that, S101 includes: S1011, Obtain model sample data, the model sample data including: multiple sets of sample working conditions, and rheological stress corresponding to the sample working conditions; S1012, the model sample data is fitted using a linear regression method to obtain initial simulation values, which include: lnA, α, n, and Q; where lnA, α, and n are material constants, and Q is the deformation activation energy; S1013, Set the constitutive model according to the initial simulation values; S1014, Perform a first optimization operation on the constitutive model to obtain an optimized constitutive model.

4. The method according to claim 3, characterized in that, The first optimization operation is a particle swarm optimization method, and correspondingly, S1014 includes: (1) After performing the optimization operation, the optimization monitoring index is obtained, including: global best fitness, particle swarm diversity or particle average velocity norm. (2) Determine whether the optimized monitoring indicators meet the stagnation warning conditions; wherein, If the result of (2) is yes, then proceed with the following steps: (3) Determine whether the performance index of the constitutive model is less than a preset first threshold; wherein, If the result of (3) is negative, then proceed with the following steps: (4) Generate an alarm signal and start S102.

5. The method according to claim 4, characterized in that, Before step (4), the following is also included: (5) Obtain the particles located in the current search space and reset some of the particles in the updated search space to obtain the updated particle swarm, wherein the range of the updated search space is larger than the current search space; (6) Perform the optimization operation n times again based on the particle swarm; (7) Determine whether the current global optimal solution has been updated. If yes, then remove the alarm signal and continue to perform the optimization operation. If no, then allow to proceed to step (4).

6. The method according to claim 3, characterized in that, In S1014, a first optimization operation is performed with the goal of reducing the error between the predicted rheological stress and the actual rheological stress.

7. The method according to claim 1, characterized in that, Prior to S102, the following are included: Obtain the performance metrics of the constitutive model; When the performance index is lower than the preset second threshold, then proceeding to S102 is allowed; And / or, the performance metric is the prediction accuracy of the constitutive model; or, the performance metric is the optimization efficiency of the constitutive model within the optimization cycle.

8. The method according to claim 7, characterized in that, Also includes: Verify the first accuracy of the constitutive model in the low temperature range and the second accuracy in the high temperature range; wherein, the low temperature range is the temperature range below the low temperature threshold, and the value range of the low temperature threshold includes: 200℃-400℃; The prediction accuracy is determined based on the first accuracy and the second accuracy.

9. A method for constructing a thermal processing diagram, characterized in that, Including the following steps: S201, establish a model mesh within the process range; and predict the rheological stress of each mesh, wherein the rheological stress is obtained by the method described in any one of claims 1-8; S202, calculate the power dissipation parameters under the true strain ε section in the model mesh according to the working conditions and the rheological stress; wherein, the power dissipation parameters include: strain rate sensitivity index m, power dissipation coefficient η, and instability criterion ξ; S203 generates a thermal processing diagram based on power consumption loss parameters.

10. A rheological stress prediction system, characterized in that, include: The constitutive model creation module is used to create a constitutive model of the target material. This constitutive model simulates the flow stress variation of the target material during hot working. The constitutive model uses working conditions as input parameters, including: plastic strain ε, strain rate, and temperature T. The output parameters of the constitutive model include: predicted flow stress σ. Arr ; A residual sample set creation module is used to create a residual sample set based on the constitutive model after the constitutive model has been created; wherein, the residual sample set creation module includes: The actual rheological stress acquisition unit is used to acquire the actual rheological stress of the target material under multiple sets of working conditions. The predictive rheological stress acquisition unit is used to input multiple sets of the aforementioned operating conditions into the constitutive model to obtain the predicted rheological stress σ under the corresponding operating conditions. Arr ; A residual sample set construction unit is used to construct a residual sample set, which includes multiple sets of residual samples, and the residual samples include the predicted rheological stress σ. Arr The actual rheological stress and the operating conditions; The residual network model optimization module is used to optimize a preset residual network model based on the residual sample set to obtain an optimized residual network model; the output parameters of the optimized residual network model include: residuals; The predictive rheological stress acquisition module is used to input the test conditions into the constitutive model to obtain the predicted rheological stress. The residual acquisition module is used to input the test conditions into the optimized residual network model to obtain the residuals; The rheological stress calculation module is used to calculate the final rheological stress based on the predicted rheological stress and the residual.

Images