A method, system, device, and medium for temperature control of titanium alloy forging

Abstract

The present application relates to the technical field of control adjustment, and particularly relates to a temperature control method, system, device and medium for titanium alloy forging, which constructs a model library describing the thermodynamic behavior of a forging, and dynamically generates a double-target temperature trajectory; a state observer is run in real time to estimate the internal temperature field of the forging online, and a feedforward control set value is calculated based on a model predictive control framework; at the same time, a feedback correction loop is used to generate a feedback correction control amount and correct the model online; finally, the feedforward and feedback control amounts are adaptively fused to generate and apply a final driving signal; by combining a physical model and a data-driven method, the accuracy of the modeling of deformation heat and the estimation of core temperature is improved, the metallurgical and cost constraints are balanced, the forging is ensured to be formed within an optimal temperature window, and the problem that the existing technology is difficult to control the microstructure of the forging within the optimal temperature window is solved.

Description

Technical Field

[0001] This invention relates to the field of control and regulation technology, and in particular to a temperature control method, system, equipment and medium for titanium alloy forging. Background Technology

[0002] Titanium alloys are widely used in aerospace and other fields with extremely high requirements for structural reliability due to their excellent specific strength and heat resistance. However, titanium alloys are quite sensitive to temperature, and their final mechanical properties and fatigue life during the forging process depend almost entirely on the microstructure formed during forging.

[0003] Existing technologies for temperature control typically employ indirect or macroscopic methods, such as measuring furnace temperature or die temperature, or directly measuring the surface temperature of the forging, to predict workpiece temperature using experience or offline simulation. However, in actual forging operations, especially under high strain rates and complex stress conditions, a large amount of deformation heat is generated inside the workpiece. This internal heat is instantaneous and non-uniformly distributed, which may make it difficult to obtain and accurately model the actual temperature of the core deformation region of the forging in real time.

[0004] In general, existing control systems struggle to control the core temperature of forgings in real time and with precision, making it difficult to determine whether the temperature has exceeded or fallen below the critical temperature window required for the formation of the target microstructure. Summary of the Invention

[0005] The main objective of this invention is to provide a temperature control method for titanium alloy forging, which aims to solve the problem that existing temperature control methods are unable to control the formation of the microstructure of forgings within the optimal temperature window.

[0006] To achieve the above objectives, the present invention provides a temperature control method for titanium alloy forging, the control method comprising the following steps:

[0007] The thermophysical parameters of the forging are obtained, and a model library is constructed based on the thermophysical parameters. The model library is used to describe the thermodynamic behavior of the forging during the heating and forging process.

[0008] Obtain the alloy constraints and cost constraints of the forging, and obtain the dual-objective temperature control trajectory based on the alloy constraints and cost constraints;

[0009] Obtain the temperature state correction gain matrix, and obtain the distribution estimate of the temperature field of the forging based on the temperature state correction gain matrix and the model library;

[0010] A coupled prediction model is constructed based on the distribution estimation of the temperature field of the forging and the target temperature control trajectory, and the feedforward control setpoint is obtained based on the coupled prediction model.

[0011] The measured surface temperature of the forging and the predicted value of the coupled prediction model are obtained. Based on the measured surface temperature of the forging, the optimal control sequence and the predicted value of the coupled prediction model, the feedback correction control quantity is obtained, and the coupled prediction model is corrected by the feedback correction control quantity.

[0012] The actual control signal is obtained based on the feedforward control setpoint and the feedback correction control quantity.

[0013] To achieve the above objectives, the present invention also provides a control system, the control system comprising:

[0014] An offline modeling module is used to acquire the thermophysical parameters of forgings and construct a model library based on the thermophysical parameters. The model library is used to describe the thermodynamic behavior of forgings during heating and forging processes.

[0015] The trajectory generation module is used to obtain the alloy constraints and cost constraints of the forging, and to obtain a dual-objective temperature control trajectory based on the alloy constraints and cost constraints.

[0016] The state estimation module is used to obtain the temperature state correction gain matrix and obtain the distribution estimate of the temperature field of the forging based on the temperature state correction gain matrix and the model library.

[0017] The feedforward control module is used to construct a coupled prediction model based on the distribution estimation of the temperature field of the forging and the target temperature control trajectory, and to obtain the feedforward control setpoint based on the coupled prediction model.

[0018] The feedback control module is used to obtain the measured temperature of the forging surface and the predicted value of the coupled prediction model, obtain the feedback correction control quantity based on the measured temperature of the forging surface, the optimal control sequence and the predicted value of the coupled prediction model, and correct the coupled prediction model through the feedback correction control quantity.

[0019] A temperature control module is used to obtain the actual control signal based on the feedforward control setpoint and the feedback correction control quantity.

[0020] To achieve the above objectives, the present invention also provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program and the processor executes the computer program to implement the above-described method.

[0021] To achieve the above objectives, the present invention also provides a computer-readable storage medium storing a computer program, wherein a processor executes the computer program to implement the above-described method.

[0022] The beneficial effects that this invention can achieve are as follows:

[0023] This invention improves the accuracy of temperature field prediction, especially the accurate modeling of deformation heat under high strain rate conditions, by combining physical models and data-driven methods. The dual-objective temperature control trajectory generation mechanism balances metallurgical and cost constraints, ensuring that forgings are formed within the optimal temperature window. The feedforward-feedback composite control architecture improves the system's adaptability to process parameter fluctuations and reduces the deviation between core temperature and surface temperature trajectory tracking. The state estimation method based on dual correction improves the estimation accuracy of core temperature, providing a reliable basis for microstructure control. Thus, it solves the problem in existing temperature control methods that make it difficult to control the formation of microstructure in forgings within the optimal temperature window. Attached Figure Description

[0024] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific 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. In the drawings, the elements or parts are not necessarily drawn to scale.

[0025] Figure 1 This is a flowchart illustrating the control method in Embodiment 1 of the present invention;

[0026] Figure 2 This is a structural block diagram of the control system in Embodiment 10 of the present invention.

[0027] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0028] 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0029] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a specific posture. If the specific posture changes, the directional indication will also change accordingly.

[0030] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0031] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the meaning of "and / or" throughout the text includes three parallel solutions; for example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0032] Example 1

[0033] Reference Figure 1 This embodiment provides a temperature control method for titanium alloy forging, the control method including the following steps:

[0034] Obtain the thermophysical property parameters of the forging, and construct a model library based on the thermophysical property parameters. The model library is used to describe the thermodynamic behavior of the forging during the heating and forging process; it is understood that this is step 1.

[0035] Obtain the alloy constraints and cost constraints of the forging, and obtain the dual-objective temperature control trajectory based on the alloy constraints and cost constraints; this is understood to be step 2.

[0036] Obtain the temperature state correction gain matrix, and obtain the distribution estimate of the temperature field of the forging based on the temperature state correction gain matrix and the model library; this is understood to be step 3.

[0037] Based on the estimated distribution of the temperature field of the forging and the target temperature control trajectory, a coupled prediction model is constructed, and the feedforward control setpoint is obtained based on the coupled prediction model; this is understood to be step 4.

[0038] Obtain the measured surface temperature of the forging and the predicted value of the coupled prediction model. Based on the measured surface temperature of the forging, the optimal control sequence, and the predicted value of the coupled prediction model, obtain the feedback correction control quantity, and correct the coupled prediction model through the feedback correction control quantity; this is understood to be step 5.

[0039] Based on the feedforward control setpoint and the feedback correction control quantity, the actual control signal is obtained; this is understood to be step 6.

[0040] In the traditional temperature control process of titanium alloy forging, due to the time-varying and spatial heterogeneity of the deformation heat source, conventional modeling methods cannot effectively identify the dynamic characteristics of the internal heat source, resulting in a systematic deviation between the predicted temperature field and the actual working conditions. This deviation is particularly significant when the forging process parameters change drastically, directly affecting the synchronization of the core temperature and surface temperature trajectory tracking.

[0041] Due to the aforementioned problems, a non-uniform recrystallized structure will form inside the forging, with abnormal grain growth occurring in localized areas. These microstructural defects cannot be eliminated in subsequent heat treatment processes, leading to a two-order-of-magnitude reduction in fatigue crack initiation life. More seriously, inaccurate temperature control may cause an imbalance in the competition mechanism between dynamic and static recrystallization during forging, resulting in batches of products exhibiting mechanical property dispersion exceeding product certification standards, thereby directly affecting the safe performance of the product.

[0042] Based on the above problems, this embodiment explores the feasibility of combining multi-objective optimization and predictive control. Due to the conflict between metallurgical constraints and cost constraints, conventional single-objective optimization is difficult to balance the two. By establishing a dual-objective temperature control trajectory generation mechanism, discrete trajectory points that satisfy the constraints are searched in the simulation environment, and then a continuous trajectory is generated through curve fitting. However, there are still dynamic errors between the simulation model and the actual system, so a feedforward-feedback composite control architecture needs to be designed. In this architecture, the feedforward control generates the setpoint based on the predictive model, while the feedback control dynamically corrects the model deviation based on the measured surface temperature.

[0043] It should be noted that this embodiment constructs a model library that integrates physical models and data-driven residual compensation, combines dual-objective optimization to generate temperature control trajectories, and adopts a feedforward-feedback composite control architecture to achieve coordinated and precise control of the core temperature and surface temperature of the forging, thereby ensuring that the microstructure is formed within the target temperature window.

[0044] It should also be noted that this embodiment first obtains the thermophysical parameters of the forging, including thermal conductivity, specific heat capacity and density, etc., and constructs a model library based on these parameters. This model library is used to describe the thermodynamic behavior of the forging during the heating and forging process, including temperature distribution, heat conduction, etc.

[0045] Next, the alloy constraints and cost constraints of the forging are obtained. The alloy constraints include requirements such as temperature window, thermal stress gradient and soaking time, while the cost constraints consider factors such as energy consumption and production efficiency. Based on these constraints, a dual-objective temperature control trajectory is generated through an optimization algorithm, including target change curves for core temperature and surface temperature.

[0046] Then, the temperature state correction gain matrix is ​​obtained, and the temperature field distribution of the forging is estimated by combining it with the model library. This step improves the accuracy of temperature field estimation by fusing model predictions and measured data.

[0047] Based on the temperature field distribution estimation and the target temperature control trajectory, a coupled prediction model is constructed. This model considers the influence of factors such as temperature and strain on the microstructure evolution. By solving the optimization problem, the feedforward control setpoint is obtained.

[0048] In the actual control process, the measured temperature of the forging surface is obtained and compared with the predicted value of the coupled prediction model. The feedback correction control quantity is calculated based on the deviation, and the coupled prediction model is dynamically modified to adapt to changes in actual working conditions.

[0049] Finally, the feedforward control setpoint and the feedback correction control input are fused to generate the actual control signal. This feedforward-feedback composite control strategy can balance the foresight of model predictions with the robustness of real-time correction.

[0050] Through the above scheme, this embodiment achieves precise control of the core temperature of titanium alloy forgings. Specifically, by combining physical models and data-driven methods, the accuracy of temperature field prediction is improved, especially the accurate modeling of deformation heat under high strain rate conditions. The dual-objective temperature control trajectory generation mechanism balances metallurgical constraints and cost constraints, ensuring that the forging is formed within the optimal temperature window. The feedforward-feedback composite control architecture improves the system's adaptability to process parameter fluctuations and reduces the deviation between core temperature and surface temperature trajectory tracking. The state estimation method based on dual correction improves the estimation accuracy of core temperature, providing a reliable basis for microstructure control. The comprehensive application of these technologies effectively solves the problem of difficulty in real-time and precise control of forging core temperature in existing technologies, laying the foundation for obtaining a uniform and stable microstructure.

[0051] Example 2:

[0052] In this embodiment, obtaining the thermophysical parameters of the forging and constructing a model library based on the thermophysical parameters includes the following sub-steps:

[0053] The thermal properties of the forging are obtained, and a reference physical model based on Fourier's law of heat conduction is established based on the thermal properties. The reference physical model includes an internal heat source term to be identified.

[0054] A residual dynamics neural network is constructed based on the thermal property parameters in historical data and the benchmark physical model. The residual dynamics neural network is used to learn and model the residual between the benchmark physical model and the real data. The residual corresponds to the internal heat source term.

[0055] The parameters of the residual dynamic neural network are globally optimized using the differential evolution algorithm to obtain a refined model.

[0056] The refined model is embedded as the internal heat source item into the baseline physical model to form a model library;

[0057] The thermophysical parameters include the thermal conductivity, specific heat capacity, and density of the forging.

[0058] It should be noted that the baseline physical model describes the heat conduction process through partial differential equations. However, the heat source terms in the equations are difficult to express analytically due to material deformation. The residual dynamics neural network receives temperature gradient and strain rate as input and outputs residual quantities related to the internal heat source. The differential evolution algorithm generates multiple candidate solutions in the parameter space, each corresponding to a set of neural network weights. By calculating the fitting error of the candidate solutions on historical data through simulation, the parameter combination with the smallest error is selected. For example, in the forging case of titanium alloy TA15, the optimized neural network reduced the temperature field prediction error from ±25℃ to ±8℃. The baseline physical model embedded in the refining model can accurately reflect the coupling effect of deformation heat and external heating, providing a high-precision prediction basis for subsequent temperature control.

[0059] It should also be noted that, based on the above, this embodiment achieves accurate modeling of the thermodynamic behavior of forgings. By introducing a residual dynamics neural network, the gap between the baseline physical model and actual data is bridged, improving the model's prediction accuracy. The application of the differential evolution algorithm ensures the global optimization of the neural network parameters, further enhancing the model's accuracy and robustness. The model library constructed in this way provides a reliable theoretical basis for subsequent temperature control, helping to achieve precise control of the microstructure of forgings and improve the performance and quality of titanium alloy forgings.

[0060] To make the technical solution of step 1 of the present invention clearer, the processing procedure of step 1 is described in detail here, which includes the following four sub-steps, specifically:

[0061] For sub-step 1.1:

[0062] The core steps are: to obtain the thermal properties of the forging and to establish a benchmark physical model based on Fourier's law of heat conduction based on the thermal properties. The benchmark physical model includes an internal heat source term to be identified.

[0063] In this embodiment, by establishing a three-dimensional transient partial differential equation for heat conduction based on Fourier's law of heat conduction, starting from the first principles of physics, the physical basis of heat transfer within an object is described by the three-dimensional transient partial differential equation for heat conduction, and the expression satisfies:

[0064] ;

[0065] in, Let be the temperature field function, representing the temperature at time t at coordinates (x, y, z) inside the forging.

[0066] ρ is the density of the titanium alloy, derived from material handbooks or actual measurements;

[0067] The specific heat capacity of titanium alloy as a function of temperature T is derived from a material properties database.

[0068] The thermal conductivity of titanium alloy as a function of temperature T is derived from a materials property database.

[0069] t is time;

[0070] It is a divergence operator used to describe the convergence or divergence of heat flow;

[0071] This is a gradient operator used to describe the rate of change of temperature in space;

[0072] For the internal heat source term, in this embodiment, it can be represented as the heat generation power per unit volume, and in this embodiment, it is regarded as an unknown function to be identified.

[0073] It is understandable that the above expression describes any point within the forging. The rate of change of temperature T at any given time t depends on the heat inflow / outflow at that point and the presence of a heat source within it, so that the model has physical interpretability and extrapolation capability.

[0074] It is also understandable that, based on the above expression, the partial differential equations can be discretized using finite element analysis (FEA) software (such as ANSYS, ABAQUS) or by self-programming to form a numerically solvable simulation model; then boundary conditions can be set, including thermal convection and thermal radiation between the furnace / mold / air; finally, a parameterized benchmark physical model can be output, which can accurately simulate the heat transfer process without an internal heat source.

[0075] For sub-step 1.2:

[0076] Its core steps are:

[0077] A residual dynamics neural network is constructed based on the thermal property parameters in historical data and the benchmark physical model. The residual dynamics neural network is used to learn and model the residual between the benchmark physical model and the real data. The residual corresponds to the internal heat source term.

[0078] In this embodiment, a residual dynamics neural network is designed to learn and model the residual between the baseline physical model and the real data, wherein the residual physically corresponds to the internal heat source term;

[0079] In this embodiment, instead of directly using a neural network to fit the elusive temperature field, a neural network is used to learn and model the residual between the baseline physical model and the real data. This residual is used to represent the internal deformation heat source of the forging during the forging process. .

[0080] Specifically, the main idea of ​​the above process is to use historical datasets to perform reverse calculations, taking the actual, measured temperature field changes (or temperature changes at key points) as known data and substituting them into the baseline physical model to solve in reverse for the internal heat source terms that enable the physical equations to hold at each time and location. Understandably, the solution obtained by reverse engineering... This refers to the target residual data that needs to be modeled. Finally, a Residual Dynamics Neural Network (RDNN) is constructed to establish the mapping relationship between the local state of the forging and the local heat generation rate. In some preferred embodiments, the RBF neural network is chosen as the basic structure of the RDNN, and its expression satisfies:

[0081] ;

[0082] in, The internal heat source value predicted by the neural network at time t is trained on the residual heat source data obtained by reverse engineering in step 1.1.

[0083] For the mapping function of the RBF neural network;

[0084] , , These represent the local strain rate, cumulative strain, and current temperature at time t, respectively. These data are derived from historical datasets or intermediate variables during the simulation process.

[0085] Using spatial coordinates, the model can learn the non-uniform distribution characteristics of deformation heat in space.

[0086] More specifically, the above expression defines a neural network model whose input is the real-time physical state (strain, strain rate, temperature, position) of a point on the forging, and whose output is the instantaneous heat source generation rate at that point in that state. .

[0087] For sub-step 1.3:

[0088] Its core steps are:

[0089] The parameters of the residual dynamic neural network are globally optimized using a differential evolution algorithm to obtain a refined model.

[0090] In this embodiment, a fitness function oriented towards the hybrid model is constructed, and the hyperparameters of the residual dynamics neural network are globally optimized using the differential evolution algorithm to ensure optimal collaborative operation between the baseline physical model and the residual dynamics neural network.

[0091] It is understandable that simply training an RDNN may result in poor coupling between it and the baseline physical model. Therefore, the steps in this embodiment optimize the hybrid system as a whole to ensure that the final prediction accuracy of the two working together is maximized.

[0092] This embodiment designs a fitness function for hybrid models and uses the differential evolution (DE) algorithm to globally optimize the hyperparameters of the residual dynamics neural network, ensuring optimal collaboration between the baseline physical model and the residual dynamics neural network. The expression satisfies:

[0093] ;

[0094] In the formula, The fitness value of the hybrid model can also be expressed as mean squared error in some implementations.

[0095] N is the total number of sample points in the validation dataset;

[0096] For the i-th sampling point in the dataset, the specific discrete-time instant is given.

[0097] The data represents the actual temperature measurement of the forging at time t, and is derived from the dataset.

[0098] This is the predicted output of the hybrid model;

[0099] The set of RDNN network parameters that the DE algorithm is currently evaluating is, in some implementations, described as all centers and width A set of.

[0100] It is understandable that the above expression calculates a set of RDNN parameters given by the DE algorithm. The mean square error of the temperature prediction of the entire hybrid model for all real data points is used as the sole evaluation criterion for the iterative optimization of the DE algorithm. By evaluating the final output of the hybrid model, rather than the intermediate output of the RDNN, it can be ensured that the parameters found in the optimization process enable the physical model and the neural network model to work together optimally, thus making up for the coupling mismatch problem that may be caused by separate optimization. In this way, a refined RDNN model that has been globally optimized and is highly coupled with the physical model is obtained.

[0101] More specifically, the calculation process for the predicted output of the mixture model is as follows:

[0102] Using parameters to be evaluated The RDNN calculates the heat source term. ;

[0103] Will Substitute into the baseline physical model of sub-step 1.1 ;

[0104] Numerical solution to the equation to obtain the predicted temperature at time t. .

[0105] For sub-step 1.4:

[0106] The core step is to embed the refined model as the internal heat source into the baseline physical model to form a model library.

[0107] In this embodiment, the refined RDNN model is formally embedded into the baseline physics model as a heat source module, forming the final hybrid physics-information model, whose governing equation expression is:

[0108] ;

[0109] in, The optimal parameters found by the DE algorithm.

[0110] Understandably, for the different deformation behaviors of titanium alloys in different temperature ranges (such as the α+β two-phase region and the β phase region), sub-steps 1.2 to 1.4 are repeated to establish a dedicated refined RDNN model for each key working condition range, and finally form a hybrid model library.

[0111] Example 3:

[0112] In this embodiment, obtaining the alloy constraints and cost constraints of the forging, and obtaining the dual-objective temperature control trajectory based on the alloy constraints and cost constraints, includes the following sub-steps:

[0113] Obtain the metallurgical constraints and construct them into a set of constraint conditions;

[0114] Obtain cost constraints and construct them into a multi-objective optimization function;

[0115] By employing a global optimization algorithm and performing simulations based on a model library, the optimal discrete trajectory point that minimizes the metallurgical constraints and cost constraints is searched and determined.

[0116] The optimal discrete trajectory points are obtained through curve fitting technology to generate the target temperature control trajectory.

[0117] The target temperature control trajectory includes a core temperature target trajectory and a surface temperature target trajectory;

[0118] The alloy constraints include temperature window constraints, thermal stress gradient constraints, and soaking time constraints.

[0119] Understandably, temperature window constraints are achieved by setting upper and lower temperature limits for different stages of the forging process, such as maintaining the core temperature at 950±15℃ in the β phase transformation region; thermal stress gradient constraints are achieved by limiting the temperature difference between adjacent regions in the temperature field to no more than 50℃ / mm. The soaking time constraint requires the workpiece to remain within the target temperature range for at least 120 seconds. In the multi-objective optimization function, heating energy cost is calculated based on the product of heating power and time, while time cost is calculated based on the total forging time. The global optimization algorithm generates a set of candidate trajectory points in each iteration, calls the model library to simulate and calculate its temperature evolution process, verifies whether the constraints are met, and calculates the total cost as the fitness value. The trajectory point that satisfies all constraints and has the lowest cost is retained, and finally, the core temperature target trajectory and the surface temperature target trajectory are generated through cubic spline interpolation. For example, in the initial heating stage, the surface temperature trajectory rises to the target range at a rate of 10℃ / s, while the core temperature trajectory achieves synchronous heating through hysteresis compensation to avoid excessive internal and external temperature differences. Thus, the dual-objective temperature control trajectory minimizes heating costs while meeting metallurgical quality requirements.

[0120] In some embodiments, the above steps can also be described as:

[0121] The preset metallurgical constraints, including temperature window, thermal stress gradient and soaking time, are constructed into a set of mathematical constraints.

[0122] Construct a multi-objective optimization function that integrates energy consumption cost, equipment depreciation cost, production time cost, and product quality penalty. ;

[0123] A global optimization algorithm is employed to perform simulations within the safe and feasible region defined by the mathematical constraint set, using the hybrid physical-information model to search for and determine the multi-objective optimization function. Minimize the optimal discrete trajectory points;

[0124] The optimal discrete trajectory points are used to generate a continuous and smooth core temperature target trajectory through curve fitting technology. and the surface temperature target trajectory .

[0125] In summary, step 2 can be divided into the following four sub-steps:

[0126] For sub-step 2.1:

[0127] The preset metallurgical constraints, including temperature window, thermal stress gradient and soaking time, are constructed into a set of mathematical constraints.

[0128] Its specific process includes:

[0129] Define temperature window constraints: ensure that the temperature is within a safe process window at any time and at any location within the forging, avoiding overheating or entering unwanted phase regions.

[0130] Define thermal stress gradient constraint: To prevent internal cracks in forgings caused by excessive internal and external temperature differences, the internal temperature gradient needs to be limited. The expression satisfies:

[0131] ;

[0132] In the formula, It represents the temperature field gradient norm at a given time and location, i.e., the rate at which the temperature at that point changes most rapidly in space;

[0133] The maximum permissible temperature gradient, expressed in °C / mm, is determined by the material's fracture toughness, coefficient of thermal expansion, and other mechanical properties through finite element analysis or empirical formulas. Understandably, directly calculating a three-dimensional stress field is extremely complex, while the temperature gradient is the direct cause of thermal stress and is easier to calculate. By constraining the temperature gradient, the goal of controlling thermal stress can also be achieved.

[0134] Define the soaking time constraint: To ensure the homogenization of microstructure and temperature before forging, the forging must remain in the range close to the target forging temperature for a sufficiently long time, as expressed in the following expression:

[0135]

[0136] In the formula, Total duration of the entire heating process;

[0137] is the Heaviside step function, which has a value of 1 when its independent variable is greater than or equal to 0, and a value of 0 otherwise.

[0138] The core temperature of the forging at time t;

[0139] The ultimate target forging temperature;

[0140] To define the temperature width of the homogenization zone;

[0141] The minimum soaking time required is derived from the forging process specifications.

[0142] Understandably, the above expression accurately calculates the core temperature of the forging. The total time spent within the homogenization temperature range is ensured to be no less than a minimum value. By employing integration and Heaviside functions, the homogenization time can be defined rigorously and accurately, making it more robust than a simple timer, especially in scenarios with small temperature fluctuations.

[0143] For sub-step 2.2:

[0144] In this embodiment, the fuzzy concept of the "optimal selection" heating trajectory is quantified into a multi-dimensional, calculable cost function that integrates energy consumption, equipment wear and tear, production efficiency, and product quality. The input cost parameters include energy consumption cost, equipment wear and tear cost, production time cost, and product quality penalty. Specifically, in some embodiments, these cost parameters are derived from the factory's cost accounting department, equipment maintenance manuals, and quality control standards, including:

[0145] Cost coefficient per unit of energy (electricity, natural gas) ;

[0146] Equivalent loss cost coefficient of forging press under unit load ;

[0147] Production cost per unit time (labor, depreciation, etc.) ;

[0148] Mass penalty coefficient for deviation from the final target temperature .

[0149] The specific processing procedure is represented as follows:

[0150] ;

[0151] In the formula, Let the total cost objective function be the function to be optimized.

[0152] This represents the time integration over the entire process from start to finish, summing up the total cost;

[0153] The heating power applied at time t is one of the control variables for optimization.

[0154] Let be the forging force function, representing the force at which the core temperature is... strain rate The forging pressure at that time, which is given by the constitutive model of the material, establishes the key physical relationship between temperature and equipment load;

[0155] , , , All of these represent weighting coefficients for cost parameters, which are dimensionless adjustment parameters set according to production priorities.

[0156] This is a terminal quality penalty item used to penalize deviations between the final core temperature and the target value.

[0157] Understandably, the above expression represents the total cost of the entire heating forging process as a single value. The goal of the optimization algorithm is to find a temperature control path that minimizes this total cost and unifies multiple conflicting production objectives into a single framework. For example, heating up too quickly can save time but increases energy costs; forging too low a temperature can save energy but leads to a surge in forging force, thereby increasing equipment wear and tear costs. The above function provides a mathematical basis for balancing these contradictions.

[0158] For sub-step 2.3:

[0159] A global optimization algorithm is employed to perform simulations within the safe and feasible region defined by the mathematical constraint set, using the hybrid physical-information model to search for and determine the multi-objective optimization function. Minimize the optimal discrete trajectory points;

[0160] In this embodiment, by selecting a global optimization algorithm suitable for handling such complex constrained nonlinear optimization problems, such as a genetic algorithm or a particle swarm optimization algorithm, a complete heating-forging-soaking temperature trajectory is represented as a series of temperature setpoints with key control time points.

[0161] In some preferred embodiments, the iterative optimization process is as follows:

[0162] A population containing multiple candidate control trajectories is randomly generated for one generation;

[0163] For each candidate trajectory, a complete simulation is performed using the hybrid physical-information model established in step one to obtain the full temperature evolution of the forging core and surface under that trajectory. and ;

[0164] Check whether the temperature evolution process obtained from the simulation satisfies all the metallurgical constraints in sub-step 2.1 throughout the entire process. If any violation occurs, impose a large penalty value on the candidate trajectory, or eliminate it directly;

[0165] For a trajectory that satisfies all constraints, its total cost is calculated using the optimization function in sub-step 2.2, and used as its fitness (the lower the cost, the higher the fitness).

[0166] Based on fitness, a better candidate trajectory population for the next generation is generated through operations such as selection, crossover, mutation, or velocity-position update;

[0167] Repeat the above process until the algorithm converges or reaches the maximum number of iterations.

[0168] It is also understandable that, based on the above process, after iteration to convergence, the obtained result represents the discrete time point-temperature value sequence of the optimal process path, corresponding to the core and surface temperatures respectively.

[0169] For sub-step 2.4:

[0170] The optimal discrete trajectory points are used to generate a continuous and smooth core temperature target trajectory through curve fitting technology. and the surface temperature target trajectory .

[0171] In this embodiment, the objective is to transform the discrete, raw optimization results obtained in the previous step into a smooth, continuous mathematical function form that is easy for the real-time control system to call. Then, using curve fitting techniques such as B-spline or piecewise polynomial interpolation, the discrete trajectory points are fitted into two continuous and smooth mathematical curves. In this embodiment, B-spline curve fitting is preferred, and its expression satisfies:

[0172] ;

[0173] In the formula, To fit the generated continuous target temperature trajectory function;

[0174] These are the control points for the B-spline curve, and their positions are determined by the discrete trajectory points output from sub-step 2.3 through a fitting algorithm.

[0175] The B-spline basis functions are a set of piecewise polynomials determined by the nodal vectors and have local support properties.

[0176] n and k are the number of control points and the spline order, respectively.

[0177] Understandably, the above expression uses a set of smoothed basis functions to weighted average a series of discrete control points, ultimately obtaining two dynamically generated optimal target trajectory functions that can be directly used in subsequent control steps, namely:

[0178] Core temperature target trajectory ;

[0179] Surface temperature target trajectory .

[0180] Example 4:

[0181] In this embodiment, obtaining the temperature state correction gain matrix and estimating the distribution of the forging temperature field based on the temperature state correction gain matrix and the model library includes the following sub-steps:

[0182] Based on the state estimate of the previous moment in the model library, obtain the prior temperature field state at the current moment;

[0183] Obtain the surface temperature measurement value of the forging, and then obtain the temperature measurement innovation based on the surface temperature measurement value of the forging and the prior temperature field state;

[0184] Obtain the temperature state correction gain matrix;

[0185] Perform the first correction: correct the prior temperature field state using the temperature measurement innovation and temperature state correction gain matrix;

[0186] Perform a second correction: use the temperature measurement innovation to obtain a correction factor, and obtain an intermediate temperature state estimate based on the correction factor;

[0187] The intermediate temperature state estimate is used as the final posterior state estimate at the current moment, and the best estimate of the core temperature and the distribution estimate of the forging temperature field are obtained.

[0188] It should be noted that within each control cycle, the model library calculates the prior temperature field distribution based on thermophysical parameters and deformable heat sources. Real-time measurement data from the surface temperature sensor is compared with the prior predicted values ​​to generate a temperature measurement innovation. This innovation is used to initially correct each node in the temperature field through a gain matrix. Subsequently, a correction factor is calculated based on the statistical characteristics of the measurement innovation, and the intermediate temperature state after the initial correction is adjusted a second time. For example, when the deviation between the measured and predicted surface temperature values ​​exceeds a threshold, the correction factor dynamically increases the correction amplitude of internal nodes to compensate for the modeling error of the deformable heat sources. The two-stage correction process improves the accuracy of temperature field estimation from the perspectives of linear error compensation and nonlinear dynamic compensation, respectively. The final output posterior state estimate includes the optimal estimate of the core temperature and the temperature distribution across the entire space, providing high-precision initial conditions for subsequent predictive control.

[0189] Based on the above, accurate estimation of the temperature field of forgings is achieved. Due to the adoption of a dual correction mechanism, the accuracy of temperature field estimation is significantly improved. In addition, by introducing a correction factor, the technical solution in this embodiment can better adapt to the temperature change characteristics under different working conditions.

[0190] Similarly, the above summary steps can also be expressed as:

[0191] The steps of the real-time operational status observer, employing a dual-calibration physical information observer, specifically include:

[0192] By combining the physical-information model with the state estimate from the previous time step, the prior temperature field state at the current time step is predicted.

[0193] Innovative Temperature Measurement Method for Calculating the Relationship Between Measured Temperature of Forging Surface and Predicted Temperature of Corresponding Surface in Prior State ;

[0194] Perform the first calibration: utilizing the aforementioned temperature measurement innovation Temperature State Correction Gain Matrix This directly corrects the prior temperature field state.

[0195] Perform a second calibration: utilizing the aforementioned temperature measurement innovation Online update of a heat source model correction factor used to characterize the uncertainty of the internal heat source model. The correction factor is used in subsequent predictions to dynamically adjust the internal heat source terms predicted by the neural network in the hybrid physical-information model.

[0196] It should be noted that traditional state observers (such as Kalman filters or Romberg observers) usually treat the model as a whole and uniformly correct all internal states based on the output error. However, in this embodiment, the model of a dual-correction physical information observer is used not only to correct the temperature state itself, but also to correct the part of the model with the greatest uncertainty, namely the real-time correction of the internal deformation heat source estimated by the RDNN, so as to provide real-time feedback and adjust its understanding of the internal physical processes (especially deformation heat) based on the actual measurement of the surface temperature.

[0197] For sub-step 3.1:

[0198] Based on the refined RDNN model in step 1, and the current measurable input, the deformation heat source distribution of each node inside the forging at the current moment is predicted. Then, the predicted heat source terms are... Substituting the discretized physical model (finite element form) from step one, perform a single-step time integration from the state at time t-1. Predict the prior state estimate at time t.

[0199] The expression satisfies:

[0200] ;

[0201] In the formula, The prior state estimation vector at time t contains the predicted temperature values ​​of all discrete nodes of the forging at time t.

[0202] This is the posterior state estimation vector at time t-1, which is the best estimate output by the observer loop in the previous round.

[0203] , The discretized state transition matrix and input matrix are obtained from the baseline physical model (finite element discretization) in step one, and are temperature-related. Functions to handle nonlinear problems;

[0204] This is a generalized input vector that includes all heat inputs, encompassing not only external heating / cooling power but also internal heat source terms predicted by the RDNN. .

[0205] The above expression is the matrix form of the discretized heat conduction equation. Based on the temperature field at the previous moment and the current heat input, the temperature field distribution at the next moment is calculated. The state-space matrix form is used for seamless integration with the observer framework in modern control theory. , It can change with temperature, making the prediction step adaptable to nonlinear systems.

[0206] For sub-step 3.2:

[0207] From the high-dimensional prior state vector In the process, the predicted temperature value corresponding to the surface measurement point is extracted to calculate the temperature measurement innovation. The expression satisfies:

[0208] ;

[0209] In the formula, The temperature measurement innovation at time t represents a scalar or small-dimensional vector, which in this embodiment can also be the deviation signal of the model prediction;

[0210] This refers to the actual reading of the sensor, such as an infrared thermometer, which represents the actual measured value of the surface temperature of the forging.

[0211] The measurement matrix is ​​used to filter out the state components corresponding to the measurement points from the complete state vector. For a single surface temperature measurement point, it is a row vector with 1 at the position of the corresponding surface node and 0 at the other positions.

[0212] For sub-step 3.3:

[0213] Traditional observers directly correct the entire temperature state using a gain matrix with the bias signal. In this embodiment, however, a dual correction mechanism divides this correction process into two parts, simultaneously updating the uncertainties of the temperature state and the heat source model. This includes a first correction and a second correction, specifically:

[0214] For the first level of correction:

[0215] The direct correction of the temperature state vector, similar to the update steps of a standard Kalman filter, is used to correct common prediction biases caused by model simplification, parameter errors, etc., and the expression satisfies:

[0216] ;

[0217] In the formula, This is an estimate of the intermediate temperature state after the first correction.

[0218] The temperature-state correction gain matrix is ​​obtained by solving the Riccati equation or by offline optimization tuning to balance response speed and noise immunity.

[0219] For the second correction:

[0220] The correction for uncertainties in the internal heat source model can be understood as follows: the deviation signal not only reflects the error in temperature prediction but also represents a potential systematic bias in the prediction of the internal heat source. Therefore, an internal variable representing this uncertainty, namely the heat source model correction factor, is updated through temperature innovation (deviation signal), with the expression satisfying:

[0221] ;

[0222] In the formula, This is the best estimate of the heat source model correction factor after the update at time t;

[0223] The heat source correction gain is a scalar parameter that needs to be tuned, which determines the speed at which the heat source model is adjusted based on measurement errors.

[0224] In the above expression, an integral feedback loop is established. If the measured surface temperature is consistently higher than the predicted temperature, it means that the internal deformation heat source may be underestimated. This formula will gradually increase the correction factor (greater than 1), and vice versa, and finally obtain the intermediate temperature state estimate after the first correction and the updated heat source model correction factor.

[0225] For sub-step 3.4:

[0226] The intermediate temperature state estimate after the first correction is used as the final posterior state estimate at the current time t, i.e. ,

[0227] In other embodiments, the final state vector can also be derived from a higher dimension. In this process, the core physical quantities of most interest in subsequent control steps are extracted, specifically including:

[0228] Best estimate of core temperature ;

[0229] Complete internal temperature field distribution estimation ;

[0230] Current heat source model correction factor .

[0231] Example 5:

[0232] In this embodiment, obtaining the optimal control sequence based on the distribution estimation of the forging temperature field and the target temperature control trajectory includes the following sub-steps:

[0233] Define a multi-temporal framework;

[0234] Obtain the microstructure evolution model of the forging grains, and construct a coupled prediction model in a multi-time domain framework based on the microstructure evolution model, the distribution estimation of the forging temperature field, and the target temperature control trajectory after dimensionality increase.

[0235] Obtain tactical and strategic costs, and construct a composite cost function based on tactical and strategic costs;

[0236] At each control time, within a multi-time-domain framework, the optimal control sequence is obtained with the goal of minimizing the composite cost function, and the first element of the optimal control sequence is used as the feedforward control setpoint.

[0237] It should be noted that the multi-time-domain framework is divided into three nested optimization levels: the first level covers the heating control sequence for the next five minutes, the second level predicts the grain size evolution for the next thirty minutes, and the third level evaluates the microstructure uniformity throughout the forging cycle. When constructing the coupled prediction model, the core temperature and surface temperature from the temperature field distribution estimation are used as boundary conditions input to the microstructure evolution model, while the grain size output from the model is fed back to the temperature control module to form a closed loop. Through dimensionality increase, the state variables of the prediction model are expanded from a single temperature field to a multi-physics set including strain energy, phase fraction, and grain size. In the composite cost function, the weight of tactical cost decreases linearly with the forging stage, while the weight of strategic cost increases exponentially in the phase transition sensitive temperature range. At each control moment, a sequential quadratic programming algorithm is used to solve the multi-time-domain optimization problem, generating an optimal sequence containing the next ten control steps. The first control variable directly acts on the heating system, and subsequent control variables participate in rolling optimization as prediction benchmarks.

[0238] It should also be noted that the above scheme can simultaneously consider temperature control and microstructure evolution across multiple time scales, achieving precise control over the forging process. Therefore, while ensuring short-term temperature control accuracy, it optimizes the medium- and long-term microstructure evolution process, improving the final mechanical properties and fatigue life of the forging. Furthermore, by introducing a composite cost function of tactical and strategic costs, the short-term control effect and long-term microstructure evolution goals can be balanced during the control process, achieving more comprehensive and efficient temperature control. Specifically, this method can effectively reduce temperature fluctuations, lower the thermal stress gradient, and ensure that the forging forms an ideal microstructure within the optimal temperature window.

[0239] In some embodiments, the above-described generalized steps can be represented as:

[0240] Expand the dimension of the system's state vector so that it simultaneously includes the temperature field state and the state of key microstructure parameters.

[0241] Construct a composite cost function consisting of tactical costs and strategic costs. The tactical cost is used to evaluate the short-term tactical prediction time domain. The tracking error of the dual-target temperature trajectories, and the strategic cost used to evaluate the long-term strategic forecast time domain throughout the process end. Internally, the impact on the final micro-organization and overall economic benefits;

[0242] At each control time, the composite cost function is minimized by solving a solution. For the constrained optimization problem with the objective, an optimal control sequence is obtained, and the first element of the sequence is used as the feedforward control setpoint.

[0243] Understandably, at each control instant in this step, it is no longer a simple calculation of an optimal control power or optimal control temperature, but rather a hierarchical optimization process with far-reaching predictive power to calculate a control command that can both satisfy short-term accurate temperature tracking and ensure optimal long-term tissue performance. Specifically, this step includes the following four interrelated sub-steps:

[0244] For sub-step 4.1:

[0245] Extending the system's state from a purely thermodynamic state (temperature) to a coupled thermodynamic-metallurgical state, the new state vector... It includes not only temperature, but also key microstructure parameters, and the expression satisfies:

[0246] ;

[0247] In the formula, Let be the extended-dimensional state vector at time t;

[0248] The temperature estimation vector for all discrete nodes inside the forging is derived from the output of step three.

[0249] The average grain size vector of all discrete nodes inside the forging is known at its initial value, and subsequent values ​​are calculated online through an evolution model coupled with temperature.

[0250] Let α be the volume fraction vector of the α phase at all discrete nodes inside the forging.

[0251] The above expression defines a new and more comprehensive description of the system state, and ultimately obtains a real-time updated extended-dimensional state vector. .

[0252] The definition of a multi-temporal framework includes:

[0253] Tactical prediction time domain A relatively short time window (e.g., tens of seconds) is used for fine and rapid temperature trajectory tracking and disturbance suppression;

[0254] Strategic Forecasting Time Domain A relatively long time window spanning the entire forging process, used to assess the far-reaching impact of current decisions on the final product quality (such as final grain size), typically... .

[0255] For sub-step 4.2:

[0256] Using the hybrid model and metallurgical model constructed in step one, and based on a hypothetical future control sequence, the overall picture of temperature and microstructure evolution in two different time domains is simultaneously extrapolated. Specifically, this is achieved through a time series from t to t+... In the loop, for each time step Perform the following steps:

[0257] Calculate the heat source: Based on the current temperature, strain, etc., call the RDNN model to calculate the internal deformation heat source;

[0258] Solving for the temperature field: combining the heat source term and the external heating power Substituting into the baseline physical model (partial differential equations), the temperature field at the next time step is numerically solved. ;

[0259] Solving for the organization field: updating the temperature field As input, the microstructure evolution model (such as the grain growth equation) is substituted to solve for the grain size at the next time step. Sum of phases .

[0260] Update the expanded dimension state to prepare for the calculation in the next time step.

[0261] Understandably, the above process will continue into the strategic time domain. At the end of the process, a complete extended-dimensional state prediction trajectory covering the entire remaining process flow is obtained, which is the future extended-dimensional state prediction trajectory. ;

[0262] It is understandable that the input to step 4.2 in the above process is the current expanded-dimensional state vector output from sub-step 4.1. ;

[0263] The hybrid physical-information model library constructed in step one;

[0264] The microorganism evolution model introduced in step 4.1;

[0265] A candidate control sequence given by the optimization algorithm:

[0266] ,in It controls the time domain.

[0267] For sub-step 4.3:

[0268] By obtaining the predicted trajectory of the future extended-dimensional state output from sub-step 4.2, the optimal dual-objective temperature trajectory dynamically generated in step 2, the micro-organization parameters of the final desired target, and the weighting coefficients of various costs (as defined in step 2), a composite cost function consisting of tactical and strategic costs is constructed, the expression of which satisfies:

[0269] ;

[0270] In the formula, The tactical cost function is a standard quadratic tracking error function;

[0271] This is the predicted future temperature vector;

[0272] The dual-target trajectory vector obtained from step two;

[0273] Let be the weighted norm square, and let represent the weighted sum of tracking errors, where It is a tactical tracking error weight matrix used to adjust the relative importance of core and surface temperature tracking;

[0274] Weighting of tactical control efforts is used to penalize excessive short-term control actions.

[0275] Assess the impact on the final outcome using a strategic cost function;

[0276] The strategic weight coefficient is a key adjustment parameter used to balance short-term tracking accuracy and long-term strategic objectives.

[0277] The final average grain size obtained from the prediction is derived from the endpoint of the dimensional expansion state prediction trajectory;

[0278] The final target grain size required by the process;

[0279] The calculation method for predicting the total economic cost from the current moment to the end of the process is the same as in step two. Similar, but only the future portion is calculated.

[0280] In the above expression, a weighted sum is used to simultaneously evaluate the short-term performance (whether it closely tracks the recent temperature target) and long-term impact (whether it can guarantee the final optimal grain size and economy) of a control decision; thus solving the short-sightedness problem of traditional MPC.

[0281] For sub-step 4.4:

[0282] At each control time, the composite cost function is minimized by solving a solution. For the constrained optimization problem with the objective, an optimal control sequence is obtained, and the first element of the sequence is used as the feedforward control setpoint.

[0283] In this embodiment, at each control time t, the following constrained optimization problem is solved:

[0284] ;

[0285] Specifically, nonlinear programming (NLP) solvers or metaheuristic optimization algorithms are used to find the optimal control sequence. According to the rolling time-domain principle of MPC, although we have calculated an entire sequence However, at the current time t, we only take the first element of the sequence. This serves as the final feedforward control variable; and at the next time t+1, a new state measurement and estimate are obtained, then the entire process of step four is repeated to perform a complete optimization calculation again to obtain the optimal feedforward control setpoint for the current time. .

[0286] Example 6:

[0287] In this embodiment, the construction process of the coupled prediction model includes:

[0288] Obtain external heating power;

[0289] Calculate the deformation heat source of the refining model based on the current temperature and strain.

[0290] Substitute the deformable heat source and external heating power into the baseline physical model to obtain the temperature field at the next time step;

[0291] The temperature field of the next time step is used as input and substituted into the microstructure evolution model to obtain the grain size and phase fraction of the next time step;

[0292] Then update the dimensionality expansion state until the end of the multi-temporal framework is reached.

[0293] It should be noted that during each time step iteration, the external heating power is collected in real time by the actual power sensor of the heating device, and the deformation heat source is obtained through quadratic invariant calculations using the current temperature field and strain rate tensor. Both are simultaneously input into the baseline physical model, and the temperature field distribution at the next time step is obtained by solving the three-dimensional unsteady-state heat conduction equation. This temperature field is then transferred to the microstructure evolution model, where the grain size can be calculated by relating it to the Zener-Hollomon parameter and the dynamic recrystallization model, and the phase fraction can be predicted using the Avrami equation combined with the temperature history. The extended-dimensional state update includes four-dimensional state variables: temperature field, strain field, grain size, and phase fraction, forming a closed-loop iteration until the prediction range covers multiple time domains. This process achieves synergistic prediction of the temperature field and microstructure by simultaneously integrating the dual effects of external heat input and internal heat generation.

[0294] It should also be noted that, based on the above process, this embodiment achieves accurate prediction of the temperature field and microstructure evolution of the forging during the forging process. This allows for better control of the core temperature of the forging, ensuring that it remains within the critical temperature window required for the formation of the target microstructure. Furthermore, this method considers the influence of deformation heat, improving the accuracy of temperature control and contributing to obtaining an ideal microstructure, thereby improving the final mechanical properties and fatigue life of the forging.

[0295] In some preferred embodiments, the above steps can also be expressed as:

[0296] The deviation between the measured surface temperature of the forging and the model prediction is decomposed into bias error components to characterize the mismatch of the slowly varying model by passing a low-pass filter. and dynamic error components used to characterize transient dynamic mismatch ;

[0297] Using the bias error component The physical boundary parameters in the hybrid physical-information model are updated online adaptively using a slow-loop integral update rule.

[0298] Using the dynamic error components The heat source model correction factor described in step four is updated in real time through a fast-loop correction circuit. ;

[0299] The residual error is processed using a low-gain controller to generate the final feedback correction control quantity. .

[0300] It should be noted that in conventional feedback control, the prediction error is input into a simple controller (such as a PID controller) to generate a compensation value for the control quantity. However, in this embodiment, by constructing a hierarchical model abductive correction framework, it no longer only corrects the final control output, but treats the prediction error as a deviation. By analyzing the dynamic characteristics of the error, it analyzes the most likely deviation amount, and then directly corrects the most uncertain parameters or modules within the model in real time.

[0301] It should also be noted that this embodiment establishes an intelligent online feedback mechanism to analyze the deviation between model predictions and physical reality in real time, and traces the deviation back to the most likely source of error within the model, whether it is an inaccurate estimation of slowly changing physical boundary parameters or a failure of model prediction due to rapidly changing internal deformation heat sources. The model itself is then dynamically corrected. Finally, only a low-gain controller is used to handle residual random errors that have not been absorbed by the model. Specifically, this includes the following four logically progressive sub-steps:

[0302] For sub-step 5.1:

[0303] In this embodiment, by assuming a total error It is composed of the superposition of two different types of errors:

[0304] One type is the low-frequency / bias error caused by static or slowly varying deviations of the model from steady-state physical parameters (such as heat exchange coefficients);

[0305] Another type is high-frequency / dynamic errors caused by the model's inaccurate prediction of transient dynamics (mainly deformation heat).

[0306] To separate these two types of errors, a digital low-pass filter is designed to extract the bias error component, whose expression satisfies:

[0307] ;

[0308] In the formula, The bias error component estimated at time t;

[0309] This is the estimated bias error value from the previous time step;

[0310] This represents the original total error at the current moment, derived from sub-step 3.2;

[0311] These are the filter coefficients.

[0312] Finally, the bias component is subtracted from the total error to obtain the dynamic error component, and thus the final bias error component is obtained. With dynamic error components .

[0313] For sub-step 5.2:

[0314] In this embodiment, by utilizing the gradually varying errors separated in the previous step, the parameters in the baseline physical model from step one that are most likely to have static deviations, such as the heat transfer coefficient, are continuously fine-tuned. Specifically, an integral parameter update rule is established to correct the boundary parameters online. If the surface temperature predicted by the model is consistently higher than expected, it means that the model may have underestimated heat dissipation, and the equivalent heat transfer coefficient of the entire system needs to be increased. Its expression satisfies:

[0315] ;

[0316] In the formula, This is the updated estimate of the convective heat transfer coefficient at time t. This updated value will be immediately used in the model calculations of the next steps, Step 3 (State Observation) and Step 4 (MPC Prediction).

[0317] This is the estimated value of the convective heat transfer coefficient at the previous moment, and its initial value comes from the offline identification in step one;

[0318] The adaptive gain is a small, normal number that determines the rate of parameter correction. This value needs to be tuned based on experience or simulation to ensure the stability of parameter adjustment.

[0319] For sub-step 5.3:

[0320] It should be noted that, compared to the integrator in step three, the ID controller designed in this embodiment can respond to dynamic errors more quickly and stably. The change in surface temperature directly translates into rapid adjustments to the prediction of internal heat sources. This can be understood as an online learning mechanism: when an impact forging occurs, if the measured surface temperature rises faster than predicted by the RDNN, the heat source model correction factor in the above expression will immediately increase. This ensures that the model generates more deformation heat in the next instant than originally anticipated, ultimately yielding a heat source model correction factor. Its expression satisfies:

[0321] ;

[0322] In the formula, This is the updated heat source model correction factor at time t. This updated factor will be used in the model predictions of steps three and four at the next time point. ;

[0323] , These are the integral gain and differential gain of the heat source correction loop, respectively. These two parameters need to be tuned through simulation to match the typical generation and dissipation rates of deformation heat.

[0324] For sub-step 5.4:

[0325] It should be noted that after the previous two steps of compensating and correcting the internal parameters of the model, theoretically most of the systematic errors have been absorbed by the model itself. This step is responsible for handling the most random and unpredictable residual errors and generating a traditional control compensation signal.

[0326] In some specific implementations, it is preferable to handle the remaining error using an existing proportional (P) controller, i.e., satisfying:

[0327] ;

[0328] In the formula, The final feedback correction control quantity is expressed in power (W) or equivalent units.

[0329] The residual error proportional gain is a small positive number that needs to be tuned.

[0330] It should also be noted that, through the above four sub-steps, an innovative feedback correction system is constructed. It not only simply corrects the controller output, but also achieves hierarchical, online adaptive correction of physical parameters (slow loop) and data-driven modules (fast loop) through error decomposition and attribution. Finally, it uses only a traditional low-gain controller as tail processing, thereby greatly improving the online prediction accuracy of the entire hybrid physical-information model while ensuring system robustness and stability.

[0331] Example 7:

[0332] In this embodiment, obtaining the measured surface temperature of the forging and the predicted value of the coupled prediction model, and obtaining the feedback correction control quantity based on the measured surface temperature of the forging, the optimal control sequence, and the predicted value of the coupled prediction model, includes the following sub-steps:

[0333] The deviation between the measured surface temperature of the forging and the predicted value of the coupled prediction model is decomposed into bias error component and dynamic error component by a filter.

[0334] Update the physical boundary parameters in the coupled prediction model based on the bias error components;

[0335] Update the correction factor based on the dynamic error components;

[0336] The residual error is then processed by the controller to obtain the feedback correction control quantity.

[0337] It should be noted that the original deviation between the measured surface temperature and the predicted value is first separated by filters: a Butterworth low-pass filter extracts the bias error component below the set cutoff frequency, and a Kalman filter extracts the high-frequency dynamic error component. The bias error component is used to correct the mold contact thermal resistance parameters, for example, by updating the contact thermal resistance value using the least squares method, eliminating steady-state deviations caused by boundary condition modeling errors. The dynamic error component is input to the state estimation module to adjust the weighting coefficients in the temperature state correction gain matrix, enhancing the tracking capability for transient disturbances. The residual error is processed by a proportional-integral controller to generate a feedback correction control quantity, which is superimposed on the feedforward control setpoint to form a closed-loop correction. This process, by processing the error components in frequency bands, avoids high-frequency noise interference with model parameter identification and improves the system's response speed to dynamic deviations.

[0338] This method enables precise estimation and control of the temperature field of forgings. By decomposing temperature deviations into bias and dynamic components and updating model parameters and correction factors accordingly, the accuracy of temperature field estimation can be effectively improved. Simultaneously, the introduction of a feedback correction mechanism can promptly compensate for model prediction errors, ensuring the real-time performance and robustness of temperature control. This approach allows for more precise control of the core temperature of forgings, keeping it within the optimal temperature window, thus facilitating the acquisition of an ideal microstructure and improving the mechanical properties and service life of the forgings.

[0339] In some preferred embodiments, the above steps can also be expressed as:

[0340] The step of fusing control and generating the final actual driving signal adopts an adaptive control authority fusion and direct driving framework based on uncertainty assessment, specifically including:

[0341] According to the heat source model correction factor The degree of deviation from its ideal value is used to calculate a feedforward control confidence score in real time to quantify the reliability of the feedforward control setpoint. ;

[0342] Based on the feedforward control confidence score, a nonlinear permission allocation function is used. The authority weights of the feedforward control setpoint and the feedback correction control quantity are dynamically calculated and weighted to obtain an ideal comprehensive control command.

[0343] The ideal integrated control command is subjected to amplitude and rate of change constraints to ensure that it conforms to the physical constraints of the actuator, thereby generating the final actual drive signal.

[0344] Understandably, in this embodiment, an adaptive control authority fusion and direct drive framework based on uncertainty assessment is constructed, merging fusion and execution into one step. It no longer blindly superimposes control signals, but instead evaluates the credibility of the two decision sources—feedforward (from MPC) and feedback (from HMCF)—in real time, dynamically allocates control weights, and directly synthesizes the drive signal ultimately applied to the physical actuator (such as a power supply or valve). Specifically, it includes the following four sub-steps:

[0345] For sub-step 6.1:

[0346] According to the heat source model correction factor The degree of deviation from its ideal value is used to calculate a feedforward control confidence score in real time to quantify the reliability of the feedforward control setpoint. The expression satisfies:

[0347] ;

[0348] In the formula, The feedforward control confidence score at time t is dimensionless and ranges from [0,1].

[0349] , is the confidence decay coefficient, a positive adjustment constant used to adjust the sensitivity of the confidence level to model bias. This value is tuned through offline simulation.

[0350] This is the correction factor for the real-time heat source model from step three.

[0351] It is understandable that, in the above expression, using an exponential decay function allows the model correction factor to be adjusted. When the value is close to the ideal value of 1, the confidence level of the model (i.e., the confidence level of the feedforward control) is very high (close to 1); however, once the value is close to the ideal value of 1, the confidence level of the model is very high (close to 1); A significant deviation from 1 indicates that the model is undergoing drastic adaptive correction, and its confidence level will drop rapidly. This non-linear relationship is more indicative of the level of confidence when the model is stable than a linear mapping, and ultimately outputs a feedforward control confidence score.

[0352] For sub-step 6.2:

[0353] In this embodiment, a nonlinear permission allocation function is used, based on the feedforward control confidence score. The system dynamically calculates the respective authority weights of the feedforward control setpoint and the feedback correction control quantity, and performs weighted fusion to obtain an ideal comprehensive control command. More specifically, this function is preferably a Sigmoid type authority allocation function, whose expression satisfies:

[0354] ;

[0355] In the formula, The dynamic permission weights for feedforward control range from [0,1].

[0356] Toggle gain for different permissions;

[0357] The confidence threshold is a constant between 0 and 1 (e.g., 0.8), representing the critical point at which we begin to trust the feedforward model.

[0358] Finally, the final ideal control command is synthesized using the following expression. :

[0359] ;

[0360] In the above expression, the feedforward and feedback control quantities are weighted and averaged according to the dynamically calculated authority weights to obtain the most reasonable comprehensive control intention at the current moment. This weighting method ensures that when the model is highly reliable, the system behavior is mainly dominated by step four to achieve optimal performance; while when the model is inaccurate, the system can smoothly switch to a more conservative and robust feedback control mode dominated by step five to ensure safety.

[0361] For sub-step 6.3:

[0362] After processing the ideal command in two steps (amplitude constraint and rate of change constraint), the final actual driving signal is output. .

[0363] For the amplitude constraint, the following is satisfied:

[0364] ;

[0365] The rate of change constraint satisfies:

[0366] ;

[0367] In the formula, This represents the maximum permissible output amplitude of the actuator.

[0368] This represents the maximum power change rate of the actuator.

[0369] This refers to the actual driving signal that was issued at the previous time t-1.

[0370] The above expression is mainly used to ensure that the difference between the instruction issued at the current moment and the instruction at the previous moment does not exceed the maximum amount that the actuator can safely change within a single control cycle.

[0371] For sub-step 6.4:

[0372] It should be noted that this step is used to form a closed loop at the lowest level to verify whether the control commands are executed accurately, and to feed back the execution deviation information to the upper level, thus constituting the monitoring of the health of the entire control system. Specifically, it includes the following steps:

[0373] Obtain the actual state value read back from the actuator ;

[0374] Calculate the execution deviation, the expression satisfies:

[0375] ;

[0376] In the formula, This is an execution error.

[0377] In other embodiments, the following steps are also included:

[0378] execution error The system monitors the actuator and triggers an actuator fault alarm if the absolute value of the error continues to exceed a preset threshold.

[0379] In other embodiments, the following steps are also included:

[0380] Calculated execution error As a new information flow, it is fed back to the HMCF framework in step five. In HMCF, this execution error can be regarded as another form of uncertainty bias and used to further correct the physical model parameters or the output of the RDNN. For example, if the instruction is consistently greater than the actual output, it may mean that the heating efficiency has decreased. HMCF can then fine-tune the heat exchange coefficient in the model accordingly.

[0381] Example 8:

[0382] In this embodiment, obtaining the actual control signal based on the feedforward control setpoint and the feedback correction control quantity includes the following sub-steps:

[0383] The confidence score of feedforward control is obtained based on the degree to which the correction factor deviates from the ideal value;

[0384] Using a nonlinear permission allocation function, the permission weights of the feedforward control setpoint and the feedback correction control quantity are obtained based on the feedforward control confidence score, and then weighted and fused to obtain the ideal integrated control command after fusion.

[0385] After obtaining the ideal integrated control command after fusion, the following steps are also included:

[0386] The ideal integrated control command is subjected to amplitude and rate of change constraints to obtain the actual control signal.

[0387] In conjunction with the aforementioned implementation methods, it can be understood that the confidence score calculation module monitors the deviation between the correction factor output by the state estimator and the theoretical value in real time. When the deviation exceeds a preset threshold, the feedforward control weight is automatically reduced. The nonlinear permission allocation function generates dynamic fusion coefficients based on the confidence score, prioritizing the use of the feedforward setpoint when the model prediction accuracy is high, and enhancing the feedback correction effect when measurement noise is high. The dynamically fused control commands enter the constraint processing stage. The amplitude constraint module clamps command values ​​exceeding the heater's rated power to a safe range, and the rate of change constraint module prevents sudden temperature changes by setting a maximum heating rate.

[0388] Example 9:

[0389] In the process of globally optimizing the parameters of the residual dynamic neural network using the differential evolution algorithm, the global optimization process includes:

[0390] A population containing multiple candidate control trajectories is randomly generated for one generation;

[0391] For each candidate trajectory within the population, the refining model is invoked for simulation to obtain the full temperature evolution of the core and surface of the forging under the candidate trajectory.

[0392] Check whether the temperature evolution process obtained from the simulation satisfies the metallurgical constraints throughout. If it does not, apply a penalty value to the candidate trajectory or eliminate it directly.

[0393] For a trajectory that satisfies all constraints, calculate the total cost using cost constraints and use the total cost as the fitness.

[0394] Based on fitness, next-generation candidate trajectories are generated through selection, crossover, mutation, or velocity-position update operations;

[0395] Repeat the above process until the algorithm converges or reaches the maximum number of iterations.

[0396] Based on the aforementioned implementation methods, it can be understood that the differential evolution algorithm first encodes the candidate trajectories in each iteration, transforming the temperature control parameters into real-valued vectors. The refined model performs a full-process simulation of the control strategy corresponding to each vector, outputting time-series data of the core and surface temperatures. A constraint checking module simultaneously verifies whether the temperature sequence remains within the target phase transition temperature range throughout. For individuals violating constraints, a penalty term is added during the fitness calculation stage to reduce their survival probability. The top 20% of elite individuals in terms of fitness are directly retained for the next generation, while the remaining individuals generate new solutions through differential mutation. During the evolutionary process, the Cauchy mutation operator is used to enhance the global search capability, while a simulated annealing mechanism is introduced to prevent premature convergence. After a preset number of iterations, the algorithm outputs the individual with the optimal fitness as the final temperature control trajectory, which achieves dual optimization of energy and time costs while satisfying metallurgical constraints.

[0397] Example 10:

[0398] As attached Figure 2 As shown, this embodiment provides a control system, which includes:

[0399] An offline modeling module is used to acquire the thermophysical parameters of forgings and construct a model library based on the thermophysical parameters. The model library is used to describe the thermodynamic behavior of forgings during heating and forging processes.

[0400] The trajectory generation module is used to obtain the alloy constraints and cost constraints of the forging, and to obtain a dual-objective temperature control trajectory based on the alloy constraints and cost constraints.

[0401] The state estimation module is used to obtain the temperature state correction gain matrix and obtain the distribution estimate of the temperature field of the forging based on the temperature state correction gain matrix and the model library.

[0402] The feedforward control module is used to construct a coupled prediction model based on the distribution estimation of the temperature field of the forging and the target temperature control trajectory, and to obtain the feedforward control setpoint based on the coupled prediction model.

[0403] The feedback control module is used to obtain the measured temperature of the forging surface and the predicted value of the coupled prediction model, obtain the feedback correction control quantity based on the measured temperature of the forging surface, the optimal control sequence and the predicted value of the coupled prediction model, and correct the coupled prediction model through the feedback correction control quantity.

[0404] A temperature control module is used to obtain the actual control signal based on the feedforward control setpoint and the feedback correction control quantity.

[0405] It should be noted that, by deeply integrating physical laws and data intelligence, a residual dynamics neural network is innovatively employed to specifically learn and quantify the difficult-to-analyze internal heat source term dominated by complex deformation heat. By constructing mathematical metallurgical constraints and multi-objective cost functions, and utilizing a global optimization algorithm, a unique dual-objective temperature trajectory is dynamically generated for each forging task, thereby ensuring that the control objective itself possesses adaptability and optimality for specific working conditions.

[0406] Example 11:

[0407] Based on the same inventive concept as the foregoing embodiments, this embodiment provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0408] Example 12:

[0409] Based on the same inventive concept as the foregoing embodiments, this embodiment provides a computer-readable storage medium storing a computer program, and a processor executes the computer program to implement the above-described method.

[0410] 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 system 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 system. 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 system that includes that element.

[0411] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0412] 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 read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a multimedia terminal device (which may be a mobile phone, computer, television receiver, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0413] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A temperature control method for titanium alloy forging, characterized by, The control method comprises the following steps: acquiring thermal physical parameters of the forging, and constructing a model library according to the thermal physical parameters, the model library being used to describe the thermodynamic behavior of the forging in the heating and forging processes; acquiring alloy constraints and cost constraints of the forging, and acquiring a double-target temperature control trajectory according to the alloy constraints and the cost constraints; acquiring a temperature state correction gain matrix, and acquiring distribution estimation of a temperature field of the forging according to the temperature state correction gain matrix and the model library; constructing a coupled prediction model according to the distribution estimation of the temperature field of the forging and the double-target temperature control trajectory, and acquiring a feedforward control set value according to the coupled prediction model; acquiring a measured temperature of a surface of the forging and a predicted value of the coupled prediction model, acquiring a feedback correction control amount according to the measured temperature of the surface of the forging, an optimal control sequence and the predicted value of the coupled prediction model, and modifying the coupled prediction model through the feedback correction control amount; acquiring an actual control signal according to the feedforward control set value and the feedback correction control amount; the acquiring of the alloy constraints and the cost constraints of the forging and the acquiring of the double-target temperature control trajectory according to the alloy constraints and the cost constraints comprises the following sub-steps: acquiring alloy constraints and constructing the alloy constraints into a constraint condition set; acquiring cost constraints and constructing the cost constraints into a multi-target optimization function; searching and determining an optimal discrete trajectory point that can minimize the alloy constraints and the cost constraints by adopting a global optimization algorithm and performing simulation according to the model library; generating the double-target temperature control trajectory from the optimal discrete trajectory point by using a curve fitting technology; the double-target temperature control trajectory comprises a core temperature target trajectory and a surface temperature target trajectory; the alloy constraints comprise temperature window constraints, thermal stress gradient constraints and soaking time constraints; the constructing of the coupled prediction model according to the distribution estimation of the temperature field of the forging and the double-target temperature control trajectory and the acquiring of the feedforward control set value according to the coupled prediction model comprise the following sub-steps: defining a multi-time-domain framework; acquiring a microstructure evolution model of the forging, and constructing the coupled prediction model in the multi-time-domain framework after dimensionality increasing of the microstructure evolution model, the distribution estimation of the temperature field of the forging and the double-target temperature control trajectory; acquiring a tactical cost and a strategic cost, and constructing a composite cost function according to the tactical cost and the strategic cost; at each control time, obtaining an optimal control sequence by minimizing the composite cost function in the multi-time-domain framework, and taking a first element of the optimal control sequence as the feedforward control set value; the tactical cost is used to evaluate tracking errors of the double-target temperature control trajectory in a short-term tactical prediction time domain, and the strategic cost is used to evaluate influences on a final microstructure and comprehensive economic benefits in a long-term strategic prediction time domain extending to the end of the process.

2. A method of temperature control for forging of a titanium alloy as set forth in claim 1, characterized in that, the acquiring of the thermal physical parameters of the forging and the constructing of the model library according to the thermal physical parameters comprise the following sub-steps: acquiring thermal physical parameters of the forging, and establishing a benchmark physical model based on a Fourier heat conduction law according to the thermal physical parameters, the benchmark physical model comprising an internal heat source term to be identified; According to the thermophysical parameters in the historical data and the benchmark physical model, a residual dynamic neural network is constructed, the residual dynamic neural network is used to learn and model the residual between the benchmark physical model and the real data, and the residual corresponds to the internal heat source term; Global optimization is performed on the parameters of the residual dynamic neural network by using a differential evolution algorithm to obtain a refined model; The refined model is embedded into the benchmark physical model as the internal heat source term to form a model library; The thermophysical parameters include thermal conductivity, specific heat capacity and density of the forging.

3. The titanium alloy forging temperature control method of claim 1, wherein A temperature state correction gain matrix is obtained, and distribution estimation of the forging temperature field is obtained according to the temperature state correction gain matrix and the model library, including the following sub-steps: According to the state estimation at the previous time on the model library, the prior temperature field state at the current time is obtained; The surface temperature measurement value of the forging is obtained, and the temperature measurement innovation is obtained according to the surface temperature measurement value of the forging and the prior temperature field state; The temperature state correction gain matrix is obtained; A first re-correction is performed: the prior temperature field state is corrected using the temperature measurement innovation and the temperature state correction gain matrix; A second re-correction is performed: a correction factor is obtained using the temperature measurement innovation, and an intermediate temperature state estimation is obtained according to the correction factor; The intermediate temperature state estimation is taken as the final posterior state estimation at the current time, and the best estimation value of the core temperature and the distribution estimation of the forging temperature field are obtained.

4. The temperature control method for titanium alloy forging as described in claim 1, characterized in that, The construction process of the coupling prediction model includes: An external heating power is obtained; According to the temperature and strain in the current state, a deformation heat source of the refined model is calculated; The deformation heat source and the external heating power are substituted into the benchmark physical model to obtain the temperature field at the next time step; The temperature field at the next time step is taken as input and substituted into the microstructure evolution model to obtain the grain size and phase fraction at the next time step; The extended state is updated again until the end of the multi-time domain framework.

5. The method of claim 3, wherein the temperature is controlled by heating the titanium alloy to a temperature of 900°C to 1,000°C. The surface measured temperature of the forging and the predicted value of the coupling prediction model are obtained, and a feedback correction control amount is obtained according to the surface measured temperature of the forging, the optimal control sequence and the predicted value of the coupling prediction model, including the following sub-steps: ​ The deviation between the surface measured temperature of the forging and the predicted value of the coupling prediction model is decomposed into a bias error component and a dynamic error component through a filter; According to the bias error component, the physical boundary parameters in the coupling prediction model are updated; According to the dynamic error component, a correction factor is updated; Residual errors are processed through a controller to obtain a feedback correction control amount.

6. A method of temperature control of a titanium alloy forging as recited in claim 1 wherein, The actual control signal is obtained according to the feedforward control set value and the feedback correction control amount, including the following sub-steps: According to the degree of deviation of the correction factor from the ideal value, a feedforward control confidence score is obtained; Through a nonlinear authority allocation function, the authority weight of the feedforward control set value and the feedback correction control amount is obtained according to the feedforward control confidence score, and weighted fusion is performed to obtain a fused ideal comprehensive control instruction.

7. A method of temperature control for forging of a titanium alloy as set forth in claim 6, characterized by After obtaining the fused ideal comprehensive control instruction, the following steps are further included: The ideal comprehensive control instruction is subjected to amplitude constraint and rate of change constraint processing to obtain an actual control signal.

8. The method of claim 2, wherein the temperature control method is used for a titanium alloy forging. In the global optimization of the parameters of the residual dynamic neural network by the differential evolution algorithm, the global optimization process includes: Randomly generating a generation of a population including a plurality of candidate control trajectories; For each candidate trajectory in the population, calling the refined model for simulation to obtain the temperature evolution of the core and surface of the forging throughout the process under the candidate trajectory; Checking whether the temperature evolution obtained by simulation satisfies the alloy constraints throughout the process, and if not, applying a penalty value to the candidate trajectory or directly eliminating it; For the trajectories that satisfy all the constraints, calculating the total cost using the cost constraint, and taking the total cost as the fitness; According to the fitness, generating the next generation of candidate trajectories through selection, crossover, mutation or speed-position update operations; Repeating the above process until the algorithm converges or the maximum number of iterations is reached.

9. A control system for implementing the temperature control method of forging a titanium alloy according to any one of claims 1 to 8, characterized by, The control system comprises: An offline modeling module, which is used to obtain the thermal physical parameters of the forging, and to construct a model library according to the thermal physical parameters, the model library being used to describe the thermodynamic behavior of the forging in the heating and forging process; A trajectory generation module, which is used to obtain the alloy constraints and cost constraints of the forging, and to obtain a double-target temperature control trajectory according to the alloy constraints and cost constraints; A state estimation module, which is used to obtain a temperature state correction gain matrix, and to obtain the distribution estimation of the temperature field of the forging according to the temperature state correction gain matrix and the model library; A feedforward control module, which is used to construct a coupled prediction model according to the distribution estimation of the temperature field of the forging and the double-target temperature control trajectory, and to obtain a feedforward control set value according to the coupled prediction model; A feedback control module, which is used to obtain the measured surface temperature of the forging and the predicted value of the coupled prediction model, to obtain a feedback correction control amount according to the measured surface temperature of the forging, the optimal control sequence and the predicted value of the coupled prediction model, and to modify the coupled prediction model through the feedback correction control amount; A temperature control module, which is used to obtain an actual control signal according to the feedforward control set value and the feedback correction control amount.

10. A computer device, comprising: The computer device comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the method of any one of claims 1-8.

11. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer program, and the processor executes the computer program to realize the method of any one of claims 1-8.

Images