An aluminum profile extrusion speed dynamic optimization method based on rheological stress prediction

Abstract

The application discloses an aluminum profile extrusion speed dynamic optimization method based on rheological stress prediction and belongs to the technical field of metal plastic working. A high-temperature compression test is used to construct an aluminum alloy rheological stress constitutive equation; a thermal-mechanical coupling finite element simulation model is established to quantize the mapping relationship among the extrusion speed, the rheological stress and the temperature rise; and then, based on the plastic work conversion principle, the adiabatic temperature rise is derived, and the optimization constraint condition and the objective function are established. The simulation data are used to train a proxy model to replace the complex real-time calculation, a dynamic optimization algorithm based on a rolling time domain is constructed, and the closed-loop regulation of the extrusion speed is realized in combination with online monitoring data. The application can realize the real-time prediction of the material deformation temperature rise, realize the maximization of the average speed under the premise of isothermal extrusion, effectively solve the control lag problem, and significantly improve the production efficiency and the product quality consistency.

Description

Technical Field

[0001] This invention relates to the field of aluminum profile production optimization technology, and in particular to a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction. Background Technology

[0002] Aluminum profiles are widely used in aerospace, construction, and transportation due to their advantages such as light weight, high strength, and corrosion resistance. Extrusion is a key forming process in aluminum profile production, and extrusion speed, as a core process parameter, directly determines the surface quality, dimensional accuracy, and production efficiency of the product.

[0003] Currently, aluminum profile extrusion production processes mainly employ constant-speed extrusion or isothermal extrusion based on temperature feedback. However, these existing technologies have the following significant drawbacks in practical applications: In traditional extrusion production, to prevent quality defects such as tearing and blistering on the profile surface, a relatively low, constant extrusion speed is typically employed. This is because as extrusion proceeds, deformation work is converted into heat, causing the profile temperature to rise continuously. If the speed is too high, the temperature rise will lead the material into the "brittle zone" or cause "overheating." Traditional isothermal extrusion technology usually relies on a temperature sensor (such as infrared thermography) at the exit for feedback control. This method suffers from severe "hysteresis"—by the time the temperature sensor detects an excessive temperature, internal structural damage to the profile has often already occurred. Furthermore, infrared thermography is susceptible to interference from oxide scale and moisture, resulting in unstable measurement accuracy. While rheological stress constitutive equations and finite element simulation (FEM) can accurately describe the extrusion process, the computation time is extremely long (a single simulation may take several hours), making it impossible to meet the millisecond-level real-time control requirements of production lines. This leads to a disconnect between theory and engineering application. Most existing extrusion control algorithms are based on empirical formulas or simple statistical regression, neglecting the influence of the material's internal microscopic physical mechanisms (such as dislocation movement and dynamic recrystallization) on macroscopic deformation resistance. This leads to poor adaptability to fluctuations in operating conditions (such as slight changes in ingot composition). In actual production, uncertainties such as uneven ingot temperature field distribution and changes in die friction are difficult to avoid. Existing control methods lack the ability to invert and compensate for the material's microscopic deformation mechanisms (rheological stress) in real time, and cannot dynamically adjust process parameters according to changes in the material's instantaneous deformation resistance. This easily leads to machine stalls due to equipment overload or product mechanical performance defects due to process parameter mismatch. In summary, there is an urgent need for a dynamic optimization method for extrusion speed that can accurately predict material rheological behavior, has a fast response speed, and can adapt to changes in operating conditions, in order to solve the technical problems of low efficiency, control lag, and the trade-off between model accuracy and speed in existing technologies. Summary of the Invention

[0004] A method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction, characterized by comprising the following steps; S1. Constructing the constitutive equation for high-temperature rheological stress in aluminum alloys: Uniaxial high-temperature compression tests were conducted using a Gleeble thermal simulation testing machine, covering the typical temperature range of aluminum profile production: 400-550℃ and strain rate range: 0.01-10 s⁻¹. Based on the experimental data, the constitutive equation for rheological stress was fitted: a hyperbolic sine form including deformation activation energy Q and absolute temperature T was used to describe the functional relationship between strain rate, deformation temperature and rheological stress. The predicted values ​​were compared with the experimental values, and the correlation coefficient R and the average relative error AARE were calculated. S2. Establish a thermo-mechanical coupled finite element simulation model of the extrusion process: Through simulation, establish the mapping relationship between process parameters, rheological stress, and temperature field; S3. Establishing the logical relationship between rheological stress and velocity: Based on the simulation data in S2, the relationship between plastic deformation work and temperature rise is analyzed. Most of the plastic deformation work is converted into heat. According to the rheological stress and strain rate of the material during deformation, the plastic deformation work generated per unit volume is obtained by integrating the rheological stress with respect to strain. The formula is expressed as follows: ,in For plastic deformation work, For rheological stress, To account for the degree of strain, most of this work will be converted into heat. Introducing the heat conversion efficiency, the work done in plastic deformation is multiplied by this efficiency to obtain the energy actually converted into heat. This heat is then divided by the product of the material's density and specific heat capacity to calculate the adiabatic temperature rise, expressed by the formula: ,in For adiabatic temperature rise, or The heat conversion efficiency is typically set between 0.9 and 0.95, because the vast majority (over 90%) of plastic work is converted into heat, with only a very small portion stored in the material's crystal lattice defects. Density: The density of aluminum alloys (approximately 2700 kg / m³), used to convert volumetric energy into mass energy. Specific heat capacity: the ability of aluminum alloy to absorb heat (approximately 900 J / (kg·K)). The relationship between extrusion speed, rheological stress and temperature rise is quantified by calculating the adiabatic temperature rise, and the average speed is maximized under the premise of isothermal extrusion by optimizing the objective function. S4. Construct dynamic optimization algorithm and control model: Train agent model based on simulation data. The input is ingot temperature, mold temperature and current speed. The output is outlet temperature and maximum rheological stress. Solve the optimal speed curve based on the model. Speed ​​curve initialization: Set the initial extrusion speed and design a dynamic optimization algorithm based on the rolling time domain, which is executed in each control cycle. S5. Online monitoring and closed-loop control implementation: Deploy the algorithm to the actual production line, collect the main cylinder pressure and extrusion rod displacement / speed signals of the extruder, the PLC controller reads the sensor data, the host computer runs the optimization algorithm module, calculates the optimal speed setpoint for the next moment based on the rheological stress prediction model, the PLC receives the speed command, controls the hydraulic proportional servo valve to act, and adjusts the extrusion speed. S6. Result Verification and Evaluation: Comparative Test: Production comparison was conducted using traditional constant speed extrusion, traditional isothermal extrusion, and dynamic rheological stress optimization. The evaluation indicators were: the fluctuation range of the profile exit temperature (the smaller the fluctuation, the better), the percentage increase in average extrusion speed, and the product microstructure and properties (testing whether the grain size is uniform and whether the surface is smooth).

[0005] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. The specific parameters for the high-temperature compression test in step S1 are set as follows: Sample preparation and dimensions: The sample shape adopts a standard cylindrical sample with specific dimensions: diameter Φ10mm∙height 15mm. This height-to-diameter ratio (1.5) can effectively reduce the end face friction effect, ensure the uniformity of deformation, and at the same time avoid the sample from becoming unstable and bending during compression. Both ends of the sample need to be finely ground and polished to ensure parallelism and smoothness, so as to reduce the frictional resistance on the contact surface with the indenter. Heating process parameters: The heating rate is set to 10℃ / s. Rapid heating can shorten the test cycle and reduce the oxidation time at high temperatures. Holding time: After reaching the set deformation temperature, hold for 180 seconds (3 minutes) to ensure uniform temperature distribution inside and on the surface of the sample, eliminate thermal gradients, and allow the material structure to reach thermodynamic equilibrium. Deformation temperature and strain rate settings: The deformation temperature range is set to 5 temperature points: 400℃, 450℃, 500℃, and 550℃. The strain rate range is set to 4 rate points: 0.01s. -1 0.1s -1 1s -1 10s -1 A total of 5 Four sets of test conditions covered the main process windows of aluminum profile extrusion production; Compression deformation: The maximum true strain is set to 0.7 (corresponding to an engineering strain of approximately 50% compression reduction). This deformation is sufficient to cover most of the deformation range during the extrusion molding process, and can fully capture the interaction between material work hardening and dynamic softening (dynamic recovery / recrystallization). Friction control and data acquisition: 75% graphite + 25% machine oil (or tantalum sheet) is evenly applied to both ends of the sample as a lubricant to approximate a uniaxial compressive stress state. The system automatically records the true stress-true strain curve with a sampling frequency of not less than 100Hz to ensure data integrity under high strain rates.

[0006] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. In step S2, the mapping relationship between process parameters, rheological stress, and temperature field is established, as follows; Geometric modeling: Establish 3D models (or simplify to 2D axisymmetric models) of the mold, extrusion cylinder, and aluminum ingot. Mesh generation and boundary condition setting are performed. Friction boundary conditions (such as shear friction model), heat transfer coefficient, and initial temperatures (ingot temperature, mold temperature, and extrusion cylinder temperature) are set. The constitutive equations obtained in S1 are imported into the finite element software Deform to simulate the forming process under different extrusion speeds. Key data are extracted: the cross-sectional temperature distribution of the profile at the mold exit and the stress distribution inside the extrusion cylinder. The quantitative relationship between extrusion speed and temperature rise caused by dynamic changes in rheological stress is established.

[0007] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. In step S3, the average velocity under isothermal extrusion is maximized through objective function optimization, as detailed below: Constraint settings: Based on the temperature rise calculation results, set constraints for the process window: Upper limit constraint: The outlet temperature of the profile must not exceed the eutectic temperature of the aluminum alloy to prevent surface tearing and bubble defects; Lower limit constraint: Rheological stress is controlled within the rated tonnage range of the extruder to prevent equipment overload and stalling; Objective function: Under the above constraints, find the optimal speed curve that changes with time. The core strategy is to use speed to adjust the adiabatic temperature rise: by dynamically adjusting the speed to compensate for the temperature fluctuation of the ingot, the extrusion process is close to isothermal extrusion, and under this premise, the average extrusion speed is maximized to achieve efficient production.

[0008] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. Step S4 involves training the agent model based on simulation data. The specific steps are as follows: S41. Simulation Data Sample Generation: Using the Latin hypercube sampling method, different combinations of operating conditions are set: Input parameter range setting: Ingot temperature: Covering the allowable fluctuation range in production; Mold temperature: Covers the preheating and steady-state operating range; Current extrusion speed: covers the equipment's capacity range; Data extraction: Run finite element simulation to extract the corresponding output variables: profile outlet temperature and maximum flow stress, and build a data sample library; S42. Data Preprocessing: Due to the large differences in the values ​​of temperature, velocity and stress, the input and output data need to be normalized to a standardized state: mean 0, variance 1; the sample data are randomly shuffled and divided into three parts according to the proportion: training set, validation set and test set, which are used for model training, hyperparameter tuning and final accuracy evaluation, respectively. S43. Proxy Model BP Neural Network Structure Design: The input layer has 3 neurons, which receive the normalized ingot temperature, mold temperature, and current speed respectively. The hidden layer has 1 to 2 hidden layers. The number of neurons is determined by trial and error or grid search (e.g., 10 to 20 neurons). The activation function is ReLU or Sigmoid function to capture the nonlinear mapping relationship between input and output. The output layer has 2 neurons, which output the normalized outlet temperature and maximum rheological stress respectively. The training function is Adam optimizer to speed up training and avoid getting trapped in local minima. S44. Model Training and Parameter Optimization: The mean squared error (MSE) is used as the loss function to measure the deviation between the surrogate model's predicted value and the finite element simulation value. The training set data is input into the network, and the network weights and thresholds are continuously updated through the backpropagation algorithm. The error change of the validation set is monitored. If the validation error does not decrease for several consecutive iterations, training is stopped to prevent the model from overfitting. S45. Model accuracy verification: Using the reserved test set data as input to the trained surrogate model, calculate the correlation coefficient and average relative error between the predicted value and the true value; S46. Model Deployment: Export the trained weight matrix and threshold parameters that meet the accuracy requirements, and encapsulate them into a dynamic link library (DLL) or Python function interface that can be called by the host computer for use by the real-time control algorithm.

[0009] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. In step S4, a dynamic optimization algorithm based on the rolling time domain is designed and executed in each control cycle, as follows: Dynamic optimization algorithm design: Input: Current extruder load, hydraulic pressure, and (possibly) online temperature measurement data; Inverse calculation of rheological stress: Based on the current load and velocity, the actual rheological stress state of the material is inversely calculated using an inverse algorithm; Prediction step: Use a surrogate model to predict whether the outlet temperature will exceed the limit if the current speed is maintained at the next moment; Adjustment steps: When the predicted temperature is lower than the target temperature limit and the equipment load has a margin, accelerate. When the predicted temperature approaches the upper limit, reduce speed. When abnormal fluctuations in rheological stress are detected (such as uneven local temperature in the ingot), the speed is finely adjusted in real time to compensate for the temperature fluctuations.

[0010] The beneficial effects of this invention are as follows: By establishing a quantitative calculation model of rheological stress and temperature rise, the temperature change trend during the extrusion process can be accurately predicted, thereby planning an optimal speed curve that approximates the material's heat resistance limit. This not only completely eliminates defects such as surface tearing and overheating caused by excessive temperature rise, but also maximizes the potential of the equipment, significantly improves the average extrusion speed and production efficiency, and achieves a balance between high-efficiency production and high-quality finished products. The physical model predicts the temperature rise trend in advance and dynamically adjusts the speed. This "predictive" control method transforms post-event remediation into pre-event prevention, effectively avoiding control lag and ensuring the uniformity of the profile's microstructure and properties along its entire length. Using simulation data to train surrogate models such as neural networks retains the high-precision description of complex physical processes by thermo-mechanical coupling simulation while achieving millisecond-level online calculation response, successfully solving the technical challenge of applying high-precision physical models to industrial real-time control. By collecting signals such as master cylinder pressure and combining them with the rheological stress model, the actual state of the material can be inverted. When uncertainties arise in actual production, such as ingot temperature fluctuations and changes in friction conditions, the algorithm can sense and dynamically compensate in real time, avoiding equipment overload and stalling or process instability, and greatly improving the stability of the production process and its adaptability to different working conditions. Attached Figure Description

[0011] Figure 1 This is a flowchart of a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction; Detailed Implementation A dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction, such as Figure 1 As shown, it includes the following steps; S1. Constructing the constitutive equation for high-temperature rheological stress in aluminum alloys: Uniaxial high-temperature compression tests were conducted using a Gleeble thermal simulation testing machine, covering the typical temperature range of aluminum profile production: 400-550℃ and strain rate range: 0.01-10 s⁻¹. Based on the experimental data, the constitutive equation for rheological stress was fitted: a hyperbolic sine form including deformation activation energy Q and absolute temperature T was used to describe the functional relationship between strain rate, deformation temperature and rheological stress. The predicted values ​​were compared with the experimental values, and the correlation coefficient R and the average relative error AARE were calculated. S2. Establish a thermo-mechanical coupled finite element simulation model of the extrusion process: Through simulation, establish the mapping relationship between process parameters, rheological stress, and temperature field; S3. Establishing the logical relationship between rheological stress and velocity: Based on the simulation data in S2, the relationship between plastic deformation work and temperature rise is analyzed. Most of the plastic deformation work is converted into heat. According to the rheological stress and strain rate of the material during deformation, the plastic deformation work generated per unit volume is obtained by integrating the rheological stress with respect to strain. The formula is expressed as follows: ,in For plastic deformation work, For rheological stress, To account for the degree of strain, most of this work will be converted into heat. Introducing the heat conversion efficiency, the work done in plastic deformation is multiplied by this efficiency to obtain the energy actually converted into heat. This heat is then divided by the product of the material's density and specific heat capacity to calculate the adiabatic temperature rise, expressed by the formula: ,in For adiabatic temperature rise, or The heat conversion efficiency is typically set between 0.9 and 0.95, because the vast majority (over 90%) of plastic work is converted into heat, with only a very small portion stored in the material's crystal lattice defects. Density: The density of aluminum alloys (approximately 2700 kg / m³), used to convert volumetric energy into mass energy. Specific heat capacity: the ability of aluminum alloy to absorb heat (approximately 900 J / (kg·K)). Most of this work will be converted into heat. Introducing the heat conversion efficiency, the work of plastic deformation is multiplied by this efficiency to obtain the energy actually converted into heat. This heat is then divided by the product of the material density and specific heat capacity to calculate the adiabatic temperature rise. By calculating the adiabatic temperature rise, the relationship between extrusion speed, rheological stress, and temperature rise is quantified. And through objective function optimization, the average speed under the premise of isothermal extrusion is maximized. S4. Construct dynamic optimization algorithm and control model: Train agent model based on simulation data. The input is ingot temperature, mold temperature and current speed. The output is outlet temperature and maximum rheological stress. Solve the optimal speed curve based on the model. Speed ​​curve initialization: Set the initial extrusion speed and design a dynamic optimization algorithm based on the rolling time domain, which is executed in each control cycle. S5. Online monitoring and closed-loop control implementation: Deploy the algorithm to the actual production line, collect the main cylinder pressure and extrusion rod displacement / speed signals of the extruder, the PLC controller reads the sensor data, the host computer runs the optimization algorithm module, calculates the optimal speed setpoint for the next moment based on the rheological stress prediction model, the PLC receives the speed command, controls the hydraulic proportional servo valve to act, and adjusts the extrusion speed. S6. Result Verification and Evaluation: Comparative Test: Production comparison was conducted using traditional constant speed extrusion, traditional isothermal extrusion, and dynamic rheological stress optimization. The evaluation indicators were: the fluctuation range of the profile exit temperature (the smaller the fluctuation, the better), the percentage increase in average extrusion speed, and the product microstructure and properties (testing whether the grain size is uniform and whether the surface is smooth).

[0012] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. The specific parameters for the high-temperature compression test in step S1 are set as follows: Sample preparation and dimensions: The sample shape adopts a standard cylindrical sample with specific dimensions: diameter Φ10mm∙height 15mm. This height-to-diameter ratio (1.5) can effectively reduce the end face friction effect, ensure the uniformity of deformation, and at the same time avoid the sample from becoming unstable and bending during compression. Both ends of the sample need to be finely ground and polished to ensure parallelism and smoothness, so as to reduce the frictional resistance on the contact surface with the indenter. Heating process parameters: The heating rate is set to 10℃ / s. Rapid heating can shorten the test cycle and reduce the oxidation time at high temperatures. Holding time: After reaching the set deformation temperature, hold for 180 seconds (3 minutes) to ensure uniform temperature distribution inside and on the surface of the sample, eliminate thermal gradients, and allow the material structure to reach thermodynamic equilibrium. Deformation temperature and strain rate settings: The deformation temperature range is set to 5 temperature points: 400℃, 450℃, 500℃, and 550℃. The strain rate range is set to 4 rate points: 0.01s. -1 0.1s -1 1s -1 10s -1 A total of 5 Four sets of test conditions covered the main process windows of aluminum profile extrusion production; Compression deformation: The maximum true strain is set to 0.7 (corresponding to an engineering strain of approximately 50% compression reduction). This deformation is sufficient to cover most of the deformation range during the extrusion molding process, and can fully capture the interaction between material work hardening and dynamic softening (dynamic recovery / recrystallization). Friction control and data acquisition: 75% graphite + 25% machine oil (or tantalum sheet) is evenly applied to both ends of the sample as a lubricant to approximate a uniaxial compressive stress state. The system automatically records the true stress-true strain curve with a sampling frequency of not less than 100Hz to ensure data integrity under high strain rates.

[0013] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. In step S2, the mapping relationship between process parameters, rheological stress, and temperature field is established, as follows; Geometric modeling: Establish 3D models (or simplify to 2D axisymmetric models) of the mold, extrusion cylinder, and aluminum ingot; mesh generation and boundary condition setting; set friction boundary conditions (such as shear friction model), heat transfer coefficient, and initial temperature (ingot temperature, mold temperature, extrusion cylinder temperature); import the constitutive equation obtained in S1 into the finite element software Deform to simulate the forming process under different extrusion speeds; extract key data: cross-sectional temperature distribution of the profile at the mold exit and stress distribution inside the extrusion cylinder; and establish a quantitative relationship between extrusion speed and temperature rise caused by dynamic changes in rheological stress.

[0014] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. In step S3, the average velocity under isothermal extrusion is maximized through objective function optimization, as detailed below: Constraint settings: Based on the temperature rise calculation results, set constraints for the process window: Upper limit constraint: The outlet temperature of the profile must not exceed the eutectic temperature of the aluminum alloy to prevent surface tearing and bubble defects; Lower limit constraint: Rheological stress is controlled within the rated tonnage range of the extruder to prevent equipment overload and stalling; Objective function: Under the above constraints, find the optimal speed curve that changes with time. The core strategy is to use speed to adjust the adiabatic temperature rise: by dynamically adjusting the speed to compensate for the temperature fluctuation of the ingot, the extrusion process is close to isothermal extrusion, and under this premise, the average extrusion speed is maximized to achieve efficient production.

[0015] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. Step S4 involves training the agent model based on simulation data. The specific steps are as follows: S41. Simulation Data Sample Generation: Using the Latin hypercube sampling method, different combinations of operating conditions are set: Input parameter range setting: Ingot temperature: Covering the allowable fluctuation range in production; Mold temperature: Covers the preheating and steady-state operating range; Current extrusion speed: covers the equipment's capacity range; Data extraction: Run finite element simulation to extract the corresponding output variables: profile outlet temperature and maximum flow stress, and build a data sample library; S42. Data Preprocessing: Due to the large differences in the values ​​of temperature, velocity and stress, the input and output data need to be normalized to a standardized state: mean 0, variance 1; the sample data are randomly shuffled and divided into three parts according to the proportion: training set, validation set and test set, which are used for model training, hyperparameter tuning and final accuracy evaluation, respectively. S43. Proxy Model BP Neural Network Structure Design: The input layer has 3 neurons, which receive the normalized ingot temperature, mold temperature, and current speed respectively. The hidden layer has 1 to 2 hidden layers. The number of neurons is determined by trial and error or grid search (e.g., 10 to 20 neurons). The activation function is ReLU or Sigmoid function to capture the nonlinear mapping relationship between input and output. The output layer has 2 neurons, which output the normalized outlet temperature and maximum rheological stress respectively. The training function is Adam optimizer to speed up training and avoid getting trapped in local minima. S44. Model Training and Parameter Optimization: The mean squared error (MSE) is used as the loss function to measure the deviation between the surrogate model's predicted value and the finite element simulation value. The training set data is input into the network, and the network weights and thresholds are continuously updated through the backpropagation algorithm. The error change of the validation set is monitored. If the validation error does not decrease for several consecutive iterations, training is stopped to prevent the model from overfitting. S45. Model accuracy verification: Using the reserved test set data as input to the trained surrogate model, calculate the correlation coefficient and average relative error between the predicted value and the true value; S46. Model Deployment: Export the trained weight matrix and threshold parameters that meet the accuracy requirements, and encapsulate them into a dynamic link library (DLL) or Python function interface that can be called by the host computer for use by the real-time control algorithm.

[0016] Furthermore, a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction is proposed. In step S4, a dynamic optimization algorithm based on the rolling time domain is designed and executed in each control cycle, as follows: Dynamic optimization algorithm design: Input: Current extruder load, hydraulic pressure, and (possibly) online temperature measurement data; Inverse calculation of rheological stress: Based on the current load and velocity, the actual rheological stress state of the material is inversely calculated using an inverse algorithm; Prediction step: Use a surrogate model to predict whether the outlet temperature will exceed the limit if the current speed is maintained at the next moment; Adjustment steps: When the predicted temperature is lower than the target temperature limit and the equipment load has a margin, accelerate. When the predicted temperature approaches the upper limit, reduce speed. When abnormal fluctuations in rheological stress are detected (such as uneven local temperature in the ingot), the speed is finely adjusted in real time to compensate for the temperature fluctuations.

[0017] Example 2 This embodiment takes the extrusion production of 6063 aluminum alloy solid profiles as an example to detail the specific implementation process of a dynamic optimization method for aluminum profile extrusion speed based on rheological stress prediction. The extrusion press used has a tonnage of 2500 tons, and the ingot specifications are as follows.

[0018] The specific implementation steps are as follows: S1. Constructing the constitutive equation for high-temperature rheological stress in aluminum alloy: Uniaxial high-temperature compression tests were conducted on 6063 aluminum alloy using a Gleeble-3500 thermal simulation testing machine. The test setup included deformation temperatures of 400℃, 450℃, 500℃, and 550℃, and strain rates of 0.01 s⁻¹, 0.1 s⁻¹, 1 s⁻¹, and 10 s⁻¹. The specimen size was a cylinder with a diameter of 10mm and a height of 15mm, and the total compression deformation was 60% (true strain approximately 0.9). Data processing: Based on the collected true stress-true strain data, a nonlinear fitting was performed using an Arrhenius hyperbolic sine model. The fitted material parameters for 6063 aluminum alloy were: deformation activation energy... Q =145.2 kJ / mol, stress index n=5.4, α=0.02 MPa -1 Accuracy verification: Calculate the statistical indicators between the model's predicted values ​​and the experimental values, and display the correlation coefficient. R=0.989 The average relative error AARE = ​​4.2%, proving that the established constitutive equation has high accuracy and can be used as the basis for subsequent calculations.

[0019] S2. Establishing a thermo-mechanical coupled finite element simulation model of the extrusion process: An extrusion model was established using HyperXtrude finite element simulation software. Model settings: The die geometry was simplified to a flat die structure. The Lagrange-Eulerian (ALE) method was used for mesh generation, with a total of approximately 120,000 mesh elements. Boundary conditions: The initial ingot temperature was set to 480℃, the die preheating temperature was set to 450℃, and the extrusion cylinder temperature was set to 440℃. The friction boundary condition adopted a shear friction model, with a friction factor of m=0.6. Simulation operation: A series of different extrusion speeds (1 mm / s to 15 mm / s) were set for steady-state simulation. The simulation results database was extracted, and the average temperature of the die exit section and the maximum rheological stress data inside the profile were recorded under different working conditions. A quantitative mapping relationship of "increased speed → increased strain rate → change in rheological stress → increased deformation work → significant temperature rise" was established.

[0020] S3. Establish the logical relationship between rheological stress and velocity: Based on the simulation data in S2, analyze the relationship between plastic deformation work and temperature rise, and calculate the logic: Under a specific working condition (e.g., velocity 5 mm / s), obtain the plastic work W per unit volume by integrating the rheological stress curve. p Approximately 85 J / mm³. Heat conversion efficiency. or=0.9, combined with the density of 6063 aluminum alloy r =2700 kg / m³, specific heat capacity c.p. =900 J / (kg·K), and the calculated adiabatic temperature rise ΔT≈31.5℃; Optimization target setting: The upper limit constraint is set as the profile outlet temperature not exceeding 550℃ (the overheating temperature threshold of 6063 alloy); the lower limit constraint is set as the extruder main cylinder pressure not exceeding 24 MN (90% of the rated pressure); the objective function is set as finding the maximum extrusion speed under the above constraints. S4. Constructing a Dynamic Optimization Algorithm and Control Model: Proxy Model Training: A three-layer BP neural network is constructed as a proxy model. The input layer has 3 nodes (ingot temperature, mold temperature, current speed), the hidden layer has 10 nodes, and the output layer has 2 nodes (outlet temperature, maximum flow stress). The model is trained using 150 sets of simulation samples generated in S2. The predicted temperature error of the trained model is controlled within ±3℃. Optimization Algorithm Design: A particle swarm optimization (PSO) algorithm combined with a rolling time-domain control strategy is adopted. The control cycle is set to 1 second. Within each cycle, the algorithm searches for the optimal speed increment at the current moment, guided by the goal of "the outlet temperature approaching the target value of 540℃". S5. Online Monitoring and Closed-Loop Control Implementation: The above algorithm is deployed on an industrial control host computer and communicates with a Siemens S7-1500 PLC. Production process: At the start of extrusion, the initial speed is 2 mm / s. The PLC collects the pressure and displacement signals of the main cylinder in real time. The host computer corrects the actual "equivalent temperature" of the current ingot through inverse calculation. Dynamic adjustment: At the 10th second, the algorithm predicts that if the current speed is maintained, the outlet temperature will reach 550℃ (approaching the upper limit) after 20 seconds, and then outputs a deceleration command. At the 40th second, as the ingot shortens and the heat capacity decreases, the temperature rise slows down. The algorithm automatically increases the speed from 4 mm / s to 8 mm / s. The hydraulic proportional servo valve responds to the command and smoothly adjusts the hydraulic flow to achieve stepless speed regulation. S6. Result Verification and Evaluation: A batch of profiles produced using this method was compared with those produced using traditional processes. Temperature fluctuation comparison: The outlet temperature fluctuation range of traditional constant-speed extrusion (3mm / s) was 25℃ throughout the entire process; the fluctuation range of traditional isothermal extrusion was 12℃; using the method of this invention, the outlet temperature fluctuation range was controlled within ±4℃. Efficiency comparison: Compared with traditional constant-speed extrusion, the average extrusion speed of this method increased from 3mm / s to 5.5mm / s, and the production efficiency increased by about 83%. Quality results: The profile surface was smooth, with no visible tears or coarse grain rings. Mechanical property tests showed that the difference between longitudinal and transverse tensile strength was less than 2 MPa, verifying the effectiveness of the method.

Claims

1. A method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction, characterized in that, Includes the following steps; S1. Constructing the constitutive equation for high-temperature rheological stress in aluminum alloys: Uniaxial high-temperature compression tests were conducted using a Gleeble thermal simulation testing machine, covering the typical temperature range of aluminum profile production: 400-550℃ and strain rate range: 0.01-10 s⁻¹. Based on the experimental data, the constitutive equation for rheological stress was fitted: a hyperbolic sine form including deformation activation energy Q and absolute temperature T was used to describe the functional relationship between strain rate, deformation temperature and rheological stress. The predicted values ​​were compared with the experimental values, and the correlation coefficient R and the average relative error AARE were calculated. S2. Establish a thermo-mechanical coupled finite element simulation model of the extrusion process: Through simulation, establish the mapping relationship between process parameters, rheological stress, and temperature field; S3. Establish the logical relationship between rheological stress and velocity: Based on the simulation data in S2, analyze the relationship between plastic deformation work and temperature rise. Most of the plastic deformation work is converted into heat. According to the rheological stress and strain rate of the material during the deformation process, the plastic deformation work generated per unit volume is obtained by integrating the rheological stress with respect to strain. Most of this work will be converted into heat. Introduce the heat conversion efficiency, multiply the plastic deformation work by the efficiency to obtain the energy actually converted into heat, and divide the heat by the product of the material density and specific heat capacity to calculate the adiabatic temperature rise. Quantify the relationship between extrusion speed, rheological stress and temperature rise through the adiabatic temperature rise, and maximize the average speed under the premise of isothermal extrusion through objective function optimization. S4. Construct dynamic optimization algorithm and control model: Train agent model based on simulation data. The input is ingot temperature, mold temperature and current speed. The output is outlet temperature and maximum rheological stress. Solve the optimal speed curve based on the model. Speed ​​curve initialization: Set the initial extrusion speed and design a dynamic optimization algorithm based on the rolling time domain, which is executed in each control cycle. S5. Online monitoring and closed-loop control implementation: Deploy the algorithm to the actual production line, collect the main cylinder pressure and extrusion rod displacement / speed signals of the extruder, the PLC controller reads the sensor data, the host computer runs the optimization algorithm module, calculates the optimal speed setpoint for the next moment based on the rheological stress prediction model, the PLC receives the speed command, controls the hydraulic proportional servo valve to act, and adjusts the extrusion speed. S6. Results Verification and Evaluation: Comparative Test: Production comparison was conducted using traditional constant speed extrusion, traditional isothermal extrusion, and dynamic rheological stress optimization. The evaluation index was the fluctuation range of the profile exit temperature: the smaller the fluctuation, the better. The percentage increase in average extrusion speed was calculated.

2. The method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction as described in claim 1, characterized in that, The specific parameters for the high-temperature compression test in step S1 are set as follows: Sample preparation and dimensions: The sample shape adopts a standard cylindrical sample with specific dimensions: diameter Φ10mm∙height 15mm. Both ends of the sample are finely ground and polished to ensure parallelism and smoothness, and to reduce frictional resistance on the contact surface with the indenter. Heating process parameters: The heating rate is set to 10℃ / s, and the holding time is 180s after the set deformation temperature is reached to ensure uniform temperature distribution inside and on the surface of the sample, eliminate thermal gradient, and make the material structure reach thermodynamic equilibrium. Deformation temperature and strain rate settings: The deformation temperature range is set to 5 temperature points: 400℃, 450℃, 500℃, and 550℃. The strain rate range is set to 4 rate points: 0.01s. -1 0.1s -1 1s -1 10s -1 A total of 5 Four sets of test conditions covered the main process windows of aluminum profile extrusion production; Compression deformation: The maximum true strain is set to 0.

7. This deformation is sufficient to cover most of the deformation range during the extrusion molding process and can fully capture the interaction between material work hardening and dynamic softening. Friction control and data acquisition: 75% graphite + 25% machine oil is evenly applied to both ends of the sample as a lubricant to approximate a uniaxial compressive stress state. The system automatically records the true stress-true strain curve with a sampling frequency of not less than 100Hz to ensure data integrity under high strain rates.

3. The method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction as described in claim 1, characterized in that, In step S2, the mapping relationship between process parameters, rheological stress, and temperature field is established, as follows; Geometric modeling: Establish three-dimensional models of the mold, extrusion cylinder, and aluminum ingot; mesh generation and boundary condition setting; set friction boundary conditions, heat transfer coefficient, and initial temperature; import the constitutive equation obtained in step S1 into the finite element software Deform to simulate the forming process under different extrusion speeds; extract key data: cross-sectional temperature distribution of the profile at the mold exit and stress distribution inside the extrusion cylinder; and establish a quantitative relationship between extrusion speed and temperature rise caused by dynamic changes in rheological stress.

4. The method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction as described in claim 1, characterized in that, In step S3, the average velocity under isothermal extrusion is maximized through objective function optimization, as detailed below: Constraint settings: Based on the temperature rise calculation results, set constraints for the process window: Upper limit constraint: The outlet temperature of the profile must not exceed the eutectic temperature of the aluminum alloy to prevent surface tearing and bubble defects; Lower limit constraint: Rheological stress is controlled within the rated tonnage range of the extruder to prevent equipment overload and stalling; Objective function: Under the above constraints, find the optimal speed curve that changes with time. The core strategy is to use speed to adjust the adiabatic temperature rise: by dynamically adjusting the speed to compensate for the temperature fluctuation of the ingot, the extrusion process is close to isothermal extrusion, and under this premise, the average extrusion speed is maximized to achieve efficient production.

5. The method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction as described in claim 1, characterized in that, Step S4 involves training the agent model based on simulation data. The specific steps are as follows: S41. Simulation Data Sample Generation: Using the Latin hypercube sampling method, different combinations of operating conditions are set: Input parameter range setting: Ingot temperature: Covering the allowable fluctuation range in production; Mold temperature: Covers the preheating and steady-state operating range; Current extrusion speed: covers the equipment's capacity range; Data extraction: Run finite element simulation to extract the corresponding output variables: profile outlet temperature and maximum flow stress, and build a data sample library; S42. Data Preprocessing: Due to the large differences in the values ​​of temperature, velocity and stress, the input and output data need to be normalized to a standardized state: mean 0, variance 1; the sample data are randomly shuffled and divided into three parts according to the proportion: training set, validation set and test set, which are used for model training, hyperparameter tuning and final accuracy evaluation, respectively. S43. Proxy Model BP Neural Network Structure Design: The input layer has 3 neurons, which receive the normalized ingot temperature, mold temperature, and current speed respectively. The hidden layer has 1 to 2 hidden layers. The number of neurons is determined by trial and error or grid search. The activation function is the Sigmoid function, which is used to capture the nonlinear mapping relationship between input and output. The output layer has 2 neurons, which output the normalized outlet temperature and maximum rheological stress respectively. The training function is the Adam optimizer, which speeds up training and avoids getting trapped in local minima. S44. Model Training and Parameter Optimization: The mean squared error (MSE) is used as the loss function to measure the deviation between the surrogate model's predicted value and the finite element simulation value. The training set data is input into the network, and the network weights and thresholds are continuously updated through the backpropagation algorithm. The error change of the validation set is monitored. When the validation error no longer decreases after several consecutive iterations, training is stopped to prevent the model from overfitting. S45. Model accuracy verification: Using the reserved test set data as input to the trained surrogate model, calculate the correlation coefficient and average relative error between the predicted value and the true value; S46. Model Deployment: Export the trained weight matrix and threshold parameters that meet the accuracy requirements, and encapsulate them into a dynamic link library interface that can be called by the host computer for use by the real-time control algorithm.

6. The method for dynamic optimization of aluminum profile extrusion speed based on rheological stress prediction as described in claim 1, characterized in that, In step S4, a dynamic optimization algorithm based on the rolling time domain is designed and executed in each control cycle, as follows: Dynamic optimization algorithm design: Input: Current extruder load, hydraulic pressure, and online temperature measurement data; Inverse calculation of rheological stress: Based on the current load and velocity, the actual rheological stress state of the material is inversely calculated using an inverse algorithm; Prediction step: Use a surrogate model to predict whether the outlet temperature will exceed the limit at the next moment if the current speed is maintained; Adjustment step: When the predicted temperature is lower than the upper limit of the target temperature and the equipment load has a margin → accelerate; When the predicted temperature approaches the upper limit, reduce speed. When abnormal fluctuations in rheological stress are detected, the speed is finely adjusted in real time to compensate for temperature fluctuations.

Images