Dynamic deformation distribution optimization method for large-specification bar rolling integrated with digital twinning

Abstract

The invention relates to the technical field of metal plastic processing, and provides an integrated digital twinning large-specification bar rolling dynamic deformation distribution optimization method which comprises the following steps: constructing a digital twinning model in a large-specification bar rolling process; equipment state data, material parameters and process parameters in the rolling process are collected in real time, and the digital twin model is synchronously updated; based on the digital twinborn model, predicting strain field distribution, temperature field distribution, material metallographic structure performance and potential defect risk of the rolled piece in the current pass; with the minimum total rolling energy consumption, the highest size precision and the optimal structure uniformity as multiple objectives, deformation distribution parameters of subsequent passes are dynamically distributed through a hybrid optimization algorithm; and the optimized deformation distribution parameters are issued to a rolling mill control system to be executed, and model parameters are corrected based on online detection data feedback. In this way, the problems that in traditional rolling, due to strong experience dependence and poor dynamic working condition adaptability, efficiency is low, and many defects exist are solved.

Description

Technical Field

[0001] Embodiments of the present invention generally relate to the field of metal plastic processing technology, and more particularly to a large-size bar rolling dynamic deformation distribution optimization method integrating digital twins. Background Art

[0002] Large-size bars (usually ≥100mm in diameter) are key basic materials in the fields of heavy machinery, energy equipment, aerospace, etc. The precision, energy efficiency and microstructure performance control of their rolling process directly affect the reliability of downstream products. However, traditional rolling technology faces the following technical bottlenecks in multi-pass deformation distribution: existing processes mostly rely on manual experience or static mathematical models to distribute the deformation of each pass, which is difficult to adapt to dynamic working conditions such as material rheological stress fluctuations and equipment state changes (such as roll wear and temperature drift) during rolling, which can easily lead to problems such as head and tail size tolerances and local strain overload causing surface cracks. The scrap rate of traditional methods due to unreasonable deformation distribution is as high as 10% to 15%, and the rolling energy consumption is more than 20% higher than the theoretical optimal value.

[0003] Large-sized bars are prone to core loosening and surface overburning due to their large cross-sectional dimensions and difficulty in deformation and penetration. At the same time, difficult-to-deform materials such as high-alloy steel and titanium alloy are highly sensitive to temperature, and traditional rolling processes are difficult to suppress the organizational performance gradient caused by the temperature difference between the head and tail of the rolled piece. The rolling process must take into account energy economy, dimensional accuracy, and microstructural uniformity, but existing methods often use single-objective optimization or weighted summation methods, which are difficult to effectively balance conflicting objectives. For example, excessive pursuit of energy minimization may aggravate grain unevenness and affect the fatigue life of the bar; while simply improving accuracy requires increasing the number of rolling passes, which significantly increases production costs. Summary of the invention

[0004] To solve the above problems, the present invention uses digital twins to map the physical rolling process in real time, responds to equipment fluctuations and changes in material properties, and adopts a multi-objective optimization algorithm to collaboratively reduce energy consumption, improve accuracy and improve microstructure. It solves the problems of low efficiency and many defects in traditional rolling due to strong dependence on experience and poor adaptability to dynamic working conditions. At the same time, it supports edge computing deployment to meet the millisecond-level real-time control requirements of the rolling line.

[0005] According to an embodiment of the present invention, a method for optimizing dynamic deformation distribution in large-gauge bar rolling using integrated digital twin is provided.

[0006] In a first aspect of the present invention, a method for optimizing dynamic deformation distribution of large-size bar rolling with integrated digital twin is provided. The method comprises:

[0007] Step S01: constructing a digital twin model of the large-size bar rolling process;

[0008] Step S02: Collect the equipment status data, material parameters, and process parameters during the rolling process in real time, and synchronously update the digital twin model;

[0009] Step S03: Based on the digital twin model, predict the strain field distribution, temperature field distribution, material metallographic structure performance, and potential defect risks of the rolled piece in the current pass;

[0010] Step S04: With the multi-objectives of minimizing the total rolling energy consumption, the highest dimensional accuracy, and the best tissue uniformity, dynamically allocate the deformation distribution parameters for the subsequent passes through a hybrid optimization algorithm, including: reduction, rotational speed;

[0011] Step S05: Send the optimized deformation distribution parameters to the rolling mill control system for execution, and correct the model parameters based on the online detection data feedback.

[0012] Further, the specific steps of Step S01 are:

[0013] Step S011: Train the material flow stress prediction neural network based on finite element simulation and actual rolling data. The input parameters of the material flow stress prediction neural network are temperature T, strain ε, strain rate ε˙, and the output is flow stress σ;

[0014] Step S012: Integrate the rolling mill stiffness model, roll wear model, and heating furnace temperature field model. The heating furnace temperature field model is established using the ANSYS Fluent / STAR-CCM+ tool and is used to predict the temperature distribution in the furnace;

[0015] Step S013: Establish the coupling relationship between rolling force - temperature - deformation amount through the fusion of the mechanism model and the data-driven model.

[0016] Further, the establishment process of the rolling mill stiffness model described in Step S012 is: Under no-load and loaded conditions, use a laser displacement sensor to measure the elastic deformation amount ΔL of the rolling mill housing, apply a step-by-increasing rolling force F, record the F-ΔL curve, and obtain the static stiffness Considering the temperature influence factor, where α = 0.003℃ -1 , K0 is the stiffness at T0 = 25℃, and T is the working temperature of the rolling mill.

[0017] Further, the establishment process of the roll wear model described in Step S012 is: Based on the Archard wear theory, establish the wear amount where k is the wear coefficient, P is the rolling pressure, L is the contact arc length of the rolled piece, and H is the hardness of the roll material; calculate the dynamic wear amount where W0 is the initial wear amount and N is the cumulative number of rolling passes

[0018] Furthermore, the mechanism model described in step S013 includes a rolling force model P = B·Q P ·K·l d and a heat transfer model where B is the width of the rolled piece, Q P is the stress state coefficient, K is the material deformation resistance, l d is the contact arc length. α is the thermal diffusivity, and T is the temperature gradient.

[0019] Furthermore, the data-driven model described in step S013 is a set of sensor data: rolling force, temperature, rotational speed, experimental data: metallographic structure, mechanical properties, simulation data: finite element simulation results, and derivative features processed by normalization, missing value handling, rolling power extraction, and cumulative deformation amount processing.

[0020] Furthermore, the hybrid optimization algorithm described in step S04 is a combined optimization strategy of genetic algorithm and gradient descent method, specifically including:

[0021] Step S041: Use the genetic algorithm to globally search for feasible solutions for deformation distribution:

[0022] Step S0411: Encoding and population initialization:

[0023] Encoding method: Real number encoding, each individual represents a set of reduction amount distribution sequences, and each pass rolling is an individual;

[0024] Population size: 100 - 200 individuals to ensure coverage of the feasible solution space;

[0025] Initial population generation: 70% of the individuals are randomly generated according to a uniform distribution, and 30% of the individuals are generated based on empirical rules;

[0026] Step S0412: Fitness function design: The fitness function needs to comprehensively evaluate three objectives: energy consumption, accuracy, and tissue performance, and satisfy the process constraints Fitness = ω1·f energy +ω2·f precision +ω3·f homogeneity +Penalty, where is the energy consumption, P i is the pass power, t i is the time; is the accuracy, D i is the round steel size; f homogeneity = the standard deviation of grain size is the tissue uniformity, the standard deviation of grain size is the metallographic structure performance of the material, and is predicted by the digital twin model; ω1, ω2, ω3 are weights and are dynamically adjusted according to the process priority;

[0027] Constraint penalty term: Penalty = λ1·max(0, Pi -P max ) + λ2·max(0, Δh i - 30%), λ1, λ2 are penalty coefficients, P max is the maximum power, Δh i is the reduction;

[0028] Step S0413: Genetic operation: Selection: Use the tournament selection method to retain the top 20% of elite individuals in terms of fitness; Crossover: Two - point crossover, exchange the reduction allocation segments of adjacent passes; Mutation: Gaussian mutation, with a standard deviation of 5% of the reduction rate range;

[0029] Step S0414: Termination condition: Reach the maximum number of iterations and the convergence criterion;

[0030] Step S042: Use the gradient - descent method to perform local refinement on the feasible solution:

[0031] Step S0421: Input preparation:

[0032] Initial solution: From the Pareto front output by the genetic algorithm, select several groups of solutions with the highest comprehensive score as the starting point of gradient descent;

[0033] Objective function: Take the multi - objective weighted sum and / or single - objective as the optimization direction;

[0034] Step S0422: Gradient calculation and update:

[0035] Numerical gradient estimation: For each variable Δh i , calculate the partial derivative through the perturbation method: Call the digital twin model to quickly simulate the rolling results after perturbation;

[0036] Gradient - descent direction: Update the reduction along the negative gradient direction:

[0037] Learning rate adjustment: Given the initial η, use the Armijo line search to ensure convergence;

[0038] Step S0423: Constraint handling:

[0039] Projection method: If the updated Δh i exceeds the process range, truncate it to the boundary value;

[0040] Rolling force constraint: After each iteration, check the rolling force. If it exceeds the limit, roll back and reduce the learning rate η ← 0.5η;

[0041] Step S0424: Termination condition: Reach the maximum number of iterations and the target change rate;

[0042] Step S043: Hybrid algorithm process:

[0043] Step S0431: Initialization: Load the digital twin model and set the optimization weights;

[0044] Step S0432: Global search by genetic algorithm: Generate the initial population, calculate the fitness, select / crossover / mutate, update the population, and judge convergence;

[0045] Step S0433: Local optimization by gradient descent: Select the Pareto solution, calculate the gradient, update the reduction amount, correct the constraints, and judge convergence;

[0046] Step S0434: Online detect the rolling result, correct the parameters of the digital twin model, and trigger a new round of optimization.

[0047] In the second aspect of the present invention, there is provided an apparatus for optimizing the dynamic deformation distribution of large-sized bar rolling integrated with a digital twin. The apparatus includes:

[0048] Model construction module: Used to construct the digital twin model of the large-sized bar rolling process;

[0049] Model update module: Used to collect the equipment status data, material parameters, and process parameters during the rolling process in real time, and synchronously update the digital twin model;

[0050] Model prediction module: Used to predict the strain field distribution, temperature field distribution, material metallographic structure performance, and potential defect risk of the rolled piece in the current pass based on the digital twin model;

[0051] Dynamic allocation module: Used to dynamically allocate the deformation distribution parameters of the subsequent passes with the multi-objectives of minimizing the total rolling energy consumption, maximizing the dimensional accuracy, and optimizing the tissue uniformity through a hybrid optimization algorithm;

[0052] Model correction module: Used to send the optimized deformation distribution parameters to the rolling mill control system for execution, and correct the model parameters based on the online detection data feedback.

[0053] In the third aspect of the present invention, there is provided an electronic device. The electronic device includes: a memory and a processor, and a computer program is stored on the memory. When the processor executes the program, it implements the method according to the first aspect of the present invention.

[0054] In the fourth aspect of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method according to the first aspect of the present invention.

[0055] The present invention uses digital twins to map the physical rolling process in real time, cope with equipment fluctuations and changes in material properties, and adopts a multi-objective optimization algorithm to synergistically reduce energy consumption, improve accuracy, and optimize the microstructure, thus solving the problems of low efficiency and numerous defects in traditional rolling caused by strong dependence on experience and poor adaptability to dynamic working conditions. Meanwhile, it supports edge computing deployment to meet the millisecond-level real-time control requirements of the rolling line.

[0056] It should be understood that the content described in the Summary of the Invention section is not intended to limit the key or important features of the embodiments of the present invention, nor to limit the scope of the present invention. Other features of the present invention will become easily understandable through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In conjunction with the accompanying drawings and with reference to the following detailed description, the above and other features, advantages, and aspects of the embodiments of the present invention will become more apparent. Among them:

[0058] Figure 1 FIG. shows a flowchart of a method for optimizing dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to an embodiment of the present invention;

[0059] Figure 2 FIG. shows a system architecture diagram according to an embodiment of the present invention;

[0060] Figure 3 FIG. shows a flowchart of a hybrid algorithm for optimizing dynamic deformation distribution according to an embodiment of the present invention;

[0061] Figure 4 FIG. shows a block diagram of a device for optimizing dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to an embodiment of the present invention;

[0062] Figure 5 FIG. shows a schematic diagram of equipment for optimizing dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments of the present invention fall within the protection scope of the present invention.

[0064] According to an embodiment of the present invention, a method for optimizing the dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins is proposed. By using digital twins to map the physical rolling process in real time and cope with equipment fluctuations and changes in material properties, a multi-objective optimization algorithm is adopted to synergistically reduce energy consumption, improve accuracy, and optimize the microstructure, solving the problems of low efficiency and many defects in traditional rolling due to strong dependence on experience and poor adaptability to dynamic working conditions. At the same time, it supports edge computing deployment to meet the millisecond-level real-time control requirements of the rolling line.

[0065] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments of the present invention.

[0066] Figure 1 FIG. is a schematic flow chart of a method for optimizing the dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to an embodiment of the present invention. The method includes:

[0067] Step S01: Construct a digital twin model of the rolling process of large-sized bars;

[0068] Step S02: Real-time collect equipment status data, material parameters, and process parameters during the rolling process, and synchronously update the digital twin model;

[0069] Step S03: Based on the digital twin model, predict the strain field distribution, temperature field distribution, material metallographic structure performance, and potential defect risks of the rolled piece in the current pass;

[0070] Step S04: Taking the minimum total rolling energy consumption, the highest dimensional accuracy, and the best tissue uniformity as multi-objectives, dynamically allocate the deformation distribution parameters for the subsequent passes through a hybrid optimization algorithm;

[0071] Step S05: Send the optimized deformation distribution parameters to the mill control system for execution, and correct the model parameters based on the online detection data feedback.

[0072] It should be noted that although the operations of the method of the present invention are described in a specific order in the above embodiments and the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be decomposed into multiple steps for execution.

[0073] In order to more clearly explain the above method for optimizing the dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins, a specific embodiment will be described below. However, it should be noted that this embodiment is only for better explaining the present invention and does not constitute an improper limitation of the present invention.

[0074] The following uses a specific example to further illustrate the method for optimizing the dynamic deformation distribution of large-sized bar rolling with integrated digital twins in more detail:

[0075] As Figure 2 shown, it is the system architecture diagram of this embodiment, and the specific steps are as follows.

[0076] Step S01: Build a digital twin model of the large-sized bar rolling process.

[0077] Step S011: Train a neural network for predicting the rheological stress of materials based on finite element simulation and actual rolling data. The input parameters of the neural network for predicting the rheological stress of materials are temperature T, strain ε, and strain rate ε˙, and the output is rheological stress σ.

[0078] Step S012: Integrate the mill stiffness model, roll wear model, and heating furnace temperature field model;

[0079] The establishment process of the mill stiffness model is as follows: Under no-load and loaded states, use a laser displacement sensor to measure the elastic deformation ΔL of the mill stand, apply a stepwise increasing rolling force F (hydraulic loading device), record the F-ΔL curve, and obtain the static stiffness Unit: kN / mm. After considering the temperature influence factor, where α = 0.003 °C -1 , K0 is the stiffness at T0 = 25 °C, T is the working temperature of the mill, and is monitored by an infrared thermal imager.

[0080] The establishment process of the roll wear model is as follows: Based on the Archard wear theory, establish the wear amount where k is the wear coefficient, material-related, calibrated through experiments, P is the rolling pressure in kN, L is the contact arc length of the rolled piece in mm, H is the hardness HB of the roll material, which can be measured before use; calculate the dynamic wear amount Unit: mm, where W0 is the initial wear amount, N is the cumulative number of rolling passes, dynamically update the roll pass profile size in combination with the rolling tonnage data in MES, measure the roll surface profile with a laser scanner, and update the pass profile size every 100 tons of rolling; dynamically correct the k value in combination with the rolling force sensor data.

[0081] Use the ANSYS Fluent / STAR-CCM+ tool to establish the heating furnace temperature field model described above for predicting the temperature distribution in the furnace.

[0082] Obtain the stress-strain curves of the material at different temperatures (800 - 1200 °C) and strain rates (0.1 - 10 s-1) through a thermal simulation testing machine (Gleeble); define the allowable rolling temperature range Tmin ≤ T ≤ Tmax according to the recrystallization characteristics of the material; main motor power limit Among them, M is the rolling torque, n is the roll speed, and η is the transmission efficiency; roll strength check Among them, is the roll neck diameter, and [τ] is the allowable shear stress; limit the reduction rate per pass Avoid local strain overload leading to cracks, where Δh is the reduction height and h0 is the height before reduction.

[0083] Specifically, the mill stiffness parameters: frame stiffness ≥ 15 GN / m, roll wear coefficient: 0.1 mm wear per thousand tons, and heating furnace temperature control error: ±10 °C.

[0084] Step S013: Establish the coupling relationship between rolling force - temperature - deformation amount through the fusion of the mechanism model and the data - driven model.

[0085] The mechanism model includes the rolling force model P = B·Q P ·K·l d and the heat transfer model Among them, B is the width of the rolled piece, Q P is the stress state coefficient, K is the material deformation resistance, l d is the contact arc length. α is the thermal diffusivity, T is the temperature gradient, measure the material deformation resistance K through a thermal simulation testing machine (Gleeble), and obtain the Q P empirical value from the "Metal Plastic Processing Handbook".

[0086] The data - driven model is to collect sensor data: rolling force, temperature, speed, experimental data: metallographic structure, mechanical properties, simulation data: finite element simulation results, and adopt derivative features after normalization, processing missing values, extracting rolling power, and cumulative deformation amount.

[0087] Step S02: Real - time collect the equipment state data, material parameters, and process parameters during the rolling process, and synchronously update the digital twin model.

[0088] Step S03: Based on the digital twin model, predict the strain - field distribution, temperature - field distribution, material metallographic structure performance, and potential defect risks of the rolled piece in the current pass.

[0089] Step S04: With the goals of minimizing the total rolling energy consumption, the highest dimensional accuracy, and the best tissue uniformity, dynamically allocate the deformation distribution parameters for the subsequent passes through a hybrid optimization algorithm, including: reduction amount, speed.

[0090] As Figure 3 shown, the hybrid optimization algorithm is a combined optimization strategy of the genetic algorithm and the gradient - descent method, specifically including:

[0091] Step S041: Use the genetic algorithm for global search of feasible solutions for deformation distribution:

[0092] Step S0411: Encoding and population initialization:

[0093] Encoding method: Real number encoding. Each individual represents a set of reduction amount allocation sequences. For 12-pass rolling, the individual is represented as [Δh1, Δh2,..., Δh 12 , where Δh i ∈[5%, 30%] (range of single-pass reduction rate);

[0094] Population size: 100 - 200 individuals to ensure coverage of the feasible solution space;

[0095] Initial population generation: 70% of the individuals are randomly generated according to a uniform distribution; 30% of the individuals are generated based on empirical rules (such as the "front large and back small" allocation strategy);

[0096] Step S0412: Fitness function design: The fitness function needs to comprehensively evaluate three objectives of energy consumption, accuracy, and tissue performance, and satisfy the process constraints Fitness = ω1·f energy +ω2·f precision +ω3·f homogeneity +Penalty;

[0097] Energy consumption: P i is the pass power, and t i is the time;

[0098] Accuracy: D i is the round steel size;

[0099] Tissue uniformity: f homogeneity = standard deviation of grain size. The standard deviation of grain size is the metallographic structure performance of the material and is predicted by the digital twin model;

[0100] Set initial weights: ω1 = 0.4, ω2 = 0.4, ω3 = 0.2, and adjust dynamically according to the process priority;

[0101] Constraint penalty term: Penalty = λ1·max(0, P i -P max ) + λ2·max(0, Δh i -30%), where λ1 and λ2 are penalty coefficients, P max is the maximum power, and Δh i is the reduction amount;

[0102] Step S0413: Genetic operations:

[0103] Selection: Use the tournament selection method (Tournament Size = 5) to retain the top 20% of the elite individuals in terms of fitness;

[0104] Crossover: Two-point crossover (Crossover Rate = 80%), exchange the reduction allocation segments of adjacent passes;

[0105] Mutation: Gaussian mutation (Mutation Rate = 10%), with a standard deviation of 5% of the reduction rate range;

[0106] Step S0414: Termination condition: Reach the maximum number of iterations: 50 generations; Convergence criterion: The fitness change rate < 1% for 5 consecutive generations;

[0107] Step S042: Use the gradient descent method to perform local refinement on the feasible solution:

[0108] Step S0421: Input preparation:

[0109] Initial solution: From the Pareto front output by the genetic algorithm, select 3 - 5 groups of solutions with the highest comprehensive score as the starting point for gradient descent;

[0110] Objective function: Take the multi-objective weighted sum / or single objective (such as energy consumption) as the optimization direction, e.g., minF = 0.5·f energy +0.3·f precision +0.2·f homogeneity ;

[0111] Step S0422: Gradient calculation and update:

[0112] Numerical gradient estimation: For each variable Δh i , calculate the partial derivative through the perturbation method: Call the digital twin model to quickly simulate the rolling results after perturbation;

[0113] Gradient descent direction: Update the reduction along the negative gradient direction:

[0114] Learning rate adjustment: Initial η = 0.01, use the Armijo line search to ensure convergence;

[0115] Step S0423: Constraint handling:

[0116] Projection method: If the updated Δh i exceeds the process range by > 30%, truncate it to the boundary value;

[0117] Rolling force constraint: Check the rolling force after each iteration. If it exceeds the limit, roll back and reduce the learning rate η ← 0.5η;

[0118] Step S0424: Termination condition: Reach the maximum number of iterations: 20 steps; Objective change rate |F (k+1) -F (k) | < 0.1%;

[0119] Step S043: Hybrid algorithm process:

[0120] Step S0431: Initialization: Load the digital twin model and set the optimization weights;

[0121] Step S0432: Global search using genetic algorithm: Generate the initial population, calculate fitness, select / crossover / mutate, update the population, and judge convergence;

[0122] Step S0433: Local optimization using gradient descent: Select the Pareto solution, calculate the gradient, update the reduction amount, correct the constraints, and judge convergence;

[0123] Step S0434: Online detection of rolling results: Dimensions, temperature, microstructure, correct the digital twin model parameters, and trigger a new round of optimization.

[0124] Step S05: Send the optimized deformation distribution parameters to the rolling mill control system for execution, and correct the model parameters based on the online detection data feedback.

[0125] Specifically, the objective functions are: Total rolling energy consumption ≤ 150 kWh, diameter deviation ≤ 0.3 mm, and grain size grade ≥ 8.

[0126] Genetic algorithm is adopted: Population size is 100, iterate 50 times, initially allocate the reduction amount for 12 passes, and then optimize to convergence using the gradient descent method; The temperature difference between the head and tail of the rolled piece is collected in real time ≤ 20 °C, and the reduction amount distribution of the 6th - 8th passes is adjusted.

[0127] Effect verification: For the rolling of Φ200 mm 42CrMo alloy steel bars, the comparison of dimensional accuracy is shown in Table 1:

[0128] Table 1

[0129] Dimension Before improvement After improvement Middle of Φ200mm 201.5-198.8 200.5-199.8 Head of Φ200mm 202-198 200.3-199.8 Tail of Φ200mm 201-197 200.3-199.8

[0130] Compared with the traditional empirical distribution method, the energy consumption is reduced by 18%, and the dimensional qualification rate is increased from 82% to 96%; The predicted strain error of the digital twin model ≤ 5%, and the defect warning accuracy rate ≥ 90%.

[0131] Specifically, if during the rolling process, the cooling water of the rolling mill suddenly fails and the local temperature rises by 50 °C, the system responds through the following steps:

[0132] The digital twin model predicts that the surface crack risk in the subsequent passes increases by 30%;

[0133] The optimization algorithm automatically reduces the reduction amount of the 9th - 10th passes from 12% to 8%, and increases the reduction amount compensation of the 11th pass.

[0134] Based on the same inventive concept, the present invention also provides an apparatus for optimizing the dynamic deformation distribution of large-sized bar rolling integrated with digital twins. The implementation of this apparatus can refer to the implementation of the above method, and the repeated parts will not be elaborated here. As Figure 2 shown, the apparatus 100 includes:

[0135] Model construction module 101: used to construct a digital twin model of the large-sized bar rolling process;

[0136] Model update module 102: used to collect the equipment status data, material parameters, and process parameters during the rolling process in real time, and synchronously update the digital twin model;

[0137] Model prediction module 103: used to predict the strain field distribution, temperature field distribution, material metallographic structure performance, and potential defect risks of the rolled piece in the current pass based on the digital twin model;

[0138] Dynamic distribution module 104: used to take the minimum total rolling energy consumption, the highest dimensional accuracy, and the best tissue uniformity as multi-objectives, and dynamically distribute the deformation distribution parameters of the subsequent passes through a hybrid optimization algorithm;

[0139] Model correction module 105: used to send the optimized deformation distribution parameters to the mill control system for execution, and correct the model parameters based on the online detection data feedback.

[0140] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the described modules can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated here.

[0141] As Figure 3 shown, the device includes a central processing unit (CPU), which can execute various appropriate actions and processes according to the computer program instructions stored in the read-only memory (ROM) or the computer program instructions loaded from the storage unit into the random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM, and RAM are connected to each other through a bus. The input / output (I / O) interface is also connected to the bus.

[0142] Multiple components in the device are connected to the I / O interface, including: an input unit, such as a keyboard, a mouse, etc.; an output unit, such as various types of displays, speakers, etc.; a storage unit, such as a disk, an optical disc, etc.; and a communication unit, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit allows the device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0143] The processing unit executes the various methods and processes described above, such as method steps S01 to S05. For example, in some embodiments, method steps S01 to S05 may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed onto the device via the ROM and / or the communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more of the method steps S01 to S05 described above may be executed. Alternatively, in other embodiments, the CPU may be configured to execute method steps S01 to S05 by any other suitable means (e.g., by means of firmware).

[0144] The functions described above herein may be performed, at least in part, by one or more hardware logic components. By way of example, and without limitation, the types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0145] The program code for implementing the methods of the present invention may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, a special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine, or entirely on the remote machine or server.

[0146] In the context of the present invention, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium may be either a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0147] In addition, although the operations are depicted in a particular order, this should be understood as requiring that the operations be performed in the particular order shown or in sequential order, or that all of the illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of the present invention. Certain features that are described in the context of separate embodiments may also be implemented in combination in a single implementation. Conversely, the various features that are described in the context of a single implementation may also be implemented separately or in any suitable subcombination in multiple implementations.

[0148] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

1. Method for optimizing dynamic deformation distribution in rolling of large-sized bars integrated with digital twins, characterized in that, The method includes: Step S01: Construct a digital twin model for the rolling process of large-sized bars; Step S02: Collect device status data, material parameters, and process parameters during the rolling process in real time, and synchronously update the digital twin model; Step S03: Based on the digital twin model, predict the strain field distribution, temperature field distribution, material metallographic structure performance, and potential defect risks of the rolled piece in the current pass; Step S04: With the goals of minimizing the total rolling energy consumption, maximizing the dimensional accuracy, and optimizing the tissue uniformity, dynamically allocate the deformation distribution parameters for the subsequent passes through a hybrid optimization algorithm; Step S05: Send the optimized deformation distribution parameters to the mill control system for execution, and correct the model parameters based on the feedback of on-line detection data.

2. The method for optimizing the dynamic deformation distribution of large-sized bar rolling integrated with digital twin according to claim 1, characterized in that The specific steps of Step S01 are: Step S011: Train a neural network for predicting material flow stress based on finite element simulation and actual rolling data. The input parameters of the neural network for predicting material flow stress are temperature T, strain ε, and strain rate ε˙, and the output is flow stress σ; Step S012: Integrate the mill stiffness model, roll wear model, and heating furnace temperature field model. The heating furnace temperature field model is established using ANSYS Fluent / STAR-CCM+ tools and is used to predict the temperature distribution in the furnace; Step S013: Establish a coupling relationship between rolling force, temperature, and deformation through the fusion of mechanism models and data-driven models.

3. The method for optimizing the dynamic deformation distribution of large-scale bar rolling integrated with digital twin according to claim 2, characterized in that The establishment process of the rolling mill stiffness model described in step S012 is as follows: under no-load and loaded states, use a laser displacement sensor to measure the elastic deformation ΔL of the rolling mill housing, apply a stepwise increasing rolling force F, record the F-ΔL curve, and obtain the static stiffness After considering the temperature influence factor, where α = 0.003 °C -1 , K0 is the stiffness at T0 = 25 °C, and T is the working temperature of the rolling mill.

4. The method for optimizing the dynamic deformation distribution of large-sized bar rolling integrated with digital twins according to claim 2, characterized in that, The establishment process of the roll wear model described in step S012 is as follows: Based on the Archard wear theory, the wear volume is established where k is the wear coefficient, P is the rolling pressure, L is the contact arc length of the rolled piece, and H is the hardness of the roll material; calculate the dynamic wear volume where W0 is the initial wear volume and N is the cumulative number of rolling passes.

5. The method for optimizing dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to claim 2, characterized in that, The mechanism model described in step S013 includes a rolling force model P = B·Q P ·K·l d and a heat transfer model where B is the width of the rolled piece, Q P is the stress state coefficient, K is the material deformation resistance, l d is the contact arc length. α is the thermal diffusivity and T is the temperature gradient.

6. The method for optimizing dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to claim 2, characterized in that The data-driven model in Step S013 is a set of sensor data: rolling force, temperature, rotational speed, experimental data: metallographic structure, mechanical properties, simulation data: finite element simulation results, and derivative features processed by normalization, handling missing values, extracting rolling power, and cumulative deformation.

7. The method for optimizing dynamic deformation distribution in the rolling of large-sized bars integrated with digital twins according to claim 1, characterized in that The hybrid optimization algorithm in Step S04 is a combined optimization strategy of genetic algorithm and gradient descent method, specifically including: Step S041: Use the genetic algorithm to globally search for feasible solutions for deformation distribution: Step S0411: Encoding and population initialization: Encoding method: Real number encoding, each individual represents a set of reduction amount distribution sequences, and the rolling of each pass is an individual; Population size: 100 - 200 individuals to ensure coverage of the feasible solution space; Initial population generation: 70% of the individuals are randomly generated according to a uniform distribution, and 30% of the individuals are generated based on empirical rules; Step S0412: Fitness function design: The fitness function needs to comprehensively evaluate the three objectives of energy consumption, precision, and tissue performance and satisfy the process constraint Fitness = ω1·f energy + ω2·f precision + ω3·f homogeneity + Penalty, where is the energy consumption, P i is the pass power, t i is the time; is the precision, D i is the round steel size; f homogeneity = the standard deviation of grain size is the tissue uniformity, the standard deviation of grain size is the material metallographic tissue performance, predicted by the digital twin model; ω1, ω2, ω3 are weights, dynamically adjusted according to the process priority; Constraint penalty term: Penalty = λ1·max(0, P i -P max ) + λ2·max(0, Δh i - 30%), where λ1 and λ2 are penalty coefficients, P max is the maximum power, and Δh i is the reduction amount; Step S0413: Genetic operations: Selection: Use the tournament selection method to retain the top 20% of elite individuals in terms of fitness; Crossover: Two-point crossover, exchange the reduction amount distribution segments of adjacent passes; Mutation: Gaussian mutation, with a standard deviation of 5% of the reduction rate range; Step S0414: Termination condition: Reach the maximum number of iterations and convergence criterion; Step S042: Use the gradient descent method to locally refine and adjust the feasible solutions: Step S0421: Input preparation: Initial solution: Select several groups of solutions with the highest comprehensive score from the Pareto front output by the genetic algorithm as the starting point of the gradient descent; Objective function: Optimize in the direction of multi-objective weighted sum and / or single objective; Step S0422: Gradient calculation and update: Numerical Gradient Estimation: For each variable Δh i , calculate the partial derivative by the perturbation method: Call the digital twin model to quickly simulate the rolling results after perturbation; Gradient descent direction: Update the pressing amount along the negative gradient direction: Learning rate adjustment: Given an initial η, use the Armijo line search to ensure convergence; Step S0423: Constraint handling: Projection method: If Δh after update i exceeds the process range, truncate it to the boundary value; Rolling force constraint: After each iteration, check the rolling force. If it exceeds the limit, roll back and decrease the learning rate η ← 0.5η; Step S0424: Termination conditions: Reach the maximum number of iterations and the target change rate; Step S043: Hybrid algorithm process: Step S0431: Initialization: Load the digital twin model and set the optimization weights; Step S0432: Global search of genetic algorithm: Generate the initial population, calculate the fitness, select / crossover / mutate, update the population, and judge convergence; Step S0433: Local optimization of gradient descent: Select the Pareto solution, calculate the gradient, update the reduction amount, perform constraint correction, and judge convergence; Step S0434: Online detect the rolling result, correct the parameters of the digital twin model, and trigger a new round of optimization.

8. Device for optimizing dynamic deformation distribution in rolling of large-sized bars integrated with digital twin, characterized in that, The device implements the method described in any one of claims 1 to 7, including: Model construction module: Used to construct a digital twin model for the rolling process of large-sized bars; Model update module: Used to collect the equipment status data, material parameters, and process parameters during the rolling process in real time, and synchronously update the digital twin model; Model prediction module: Used to predict the strain field distribution, temperature field distribution, material metallographic structure performance, and potential defect risk of the rolled piece in the current pass based on the digital twin model; Dynamic allocation module: Used to dynamically allocate the deformation allocation parameters of the subsequent passes through a hybrid optimization algorithm with the multi-objectives of minimizing the total rolling energy consumption, maximizing the dimensional accuracy, and optimizing the tissue uniformity; Model correction module: Used to send the optimized deformation allocation parameters to the mill control system for execution, and correct the model parameters based on the online detection data feedback.

9. An electronic device, comprising a memory and a processor, wherein a computer program is stored on the memory, characterized in that, When the processor executes the program, it implements the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method described in any one of claims 1 to 7.

Images