An aluminum profile extrusion process energy consumption optimization method based on digital twinning

Abstract

The present application relates to a kind of aluminium profile extrusion process energy consumption optimization method based on digital twinning, by constructing the digital twinning model of aluminium profile extrusion process, the range of extrusion process parameters is set, the preset process parameters are imported into model and simulated, quality index and energy consumption index are extracted, and process parameters are optimized by multi-objective optimization algorithm;The optimized parameters are imported into model simulation again, and the results before and after optimization are compared to determine whether optimal is reached.If product quality is qualified and energy consumption is optimal, then the optimized parameters are applied to actual production;Otherwise, continue to optimize until the predetermined requirement is reached.Through digital twinning means, the strength and workload of repeatedly finding the process parameters of aluminium profile extrusion process energy consumption optimization by physical experiment are reduced, the energy saving and emission reduction of aluminium profile extrusion process are realized, and the production efficiency of enterprise is improved.

Description

Technical Field

[0001] This invention relates to the field of energy consumption optimization technology in aluminum profile processing, specifically to a method for optimizing energy consumption in the aluminum profile extrusion process based on digital twins. Background Technology

[0002] With the increasing severity of global energy shortages and environmental pollution, the manufacturing industry faces enormous challenges in energy conservation and emission reduction. Aluminum extrusion, as a crucial processing technology widely used in construction, transportation, and electronics, has drawn significant attention due to its energy consumption. However, the aluminum extrusion process is highly energy-intensive and emission-prone, influenced by process parameters, leading to increased production costs for manufacturers. Traditional aluminum extrusion processes often rely on experience and repeated manual experiments to determine optimal process parameters, resulting in high energy consumption and low production efficiency.

[0003] In recent years, with the rapid development of digital technology, the application of digital twin technology in the industrial manufacturing field has gradually emerged. Digital twin technology achieves accurate simulation and real-time monitoring of actual production processes through the deep integration of virtual models and physical equipment. It can not only simulate various process parameter changes during production, but also obtain key indicators of product quality and energy consumption through virtual experiments, providing data support for optimizing process parameters.

[0004] In the field of aluminum extrusion, the application of digital twin technology can effectively reduce the number of physical experiments, lower experimental costs, and improve production efficiency. However, most current research based on digital twin technology focuses on product quality optimization, with relatively little research and optimization on energy consumption. In particular, how to minimize energy consumption through digital twin methods during aluminum extrusion remains a pressing technical challenge. Summary of the Invention

[0005] The purpose of this invention is to provide an energy consumption optimization method for aluminum profile extrusion process based on digital twins. A digital twin model of the aluminum profile extrusion process is constructed, and the process parameters are optimized by combining the simulation results of product quality and energy consumption and using a multi-objective optimization algorithm to improve production efficiency.

[0006] Based on a key aspect of the present invention, a method for optimizing energy consumption in an aluminum profile extrusion process based on digital twins is provided, comprising one or a combination of the following steps:

[0007] Step 1: Construct a digital twin model of the aluminum profile extrusion process;

[0008] Step 2: Set the range of various process parameters for aluminum profile extrusion. The process parameters include extrusion temperature T, extrusion speed V, hydraulic system pressure Ph, electrical system power Pe, die geometry G, and raw material characteristic parameters M.

[0009] Step 3: Import the preset aluminum profile extrusion process parameters into the digital twin model, simulate the aluminum profile extrusion forming process, and extract the quality index value Q (Pi) and energy consumption index value E (Pi) from the simulation results of the digital twin model.

[0010] Step 4: Using the quality index and energy consumption index values ​​obtained in Step 3) as constraints, optimize the aluminum profile extrusion process parameters using a multi-objective optimization algorithm.

[0011] Step 5: Import the optimized aluminum profile extrusion process parameters into the digital twin model, simulate the aluminum profile extrusion forming process again, and extract the quality index and energy consumption index values.

[0012] Step 6: Compare the aluminum profile extrusion process parameters and product quality indicators before and after optimization to determine whether the product quality is qualified and whether the energy consumption is optimal. If the product quality is qualified and the energy consumption is optimal, the optimized aluminum profile extrusion process parameters are applied to actual production. Otherwise, the quality indicator values ​​and energy consumption indicator values ​​obtained in Step 5 are used as constraints, and the multi-objective optimization algorithm described in Step 4 is used to optimize the aluminum profile extrusion process parameters. Steps 5 and 6 are repeated until the product quality is qualified and the energy consumption is optimal.

[0013] In the above scheme, by constructing a digital twin-based model of the aluminum profile extrusion process, digital simulation and optimization of the extrusion process can be achieved, accurately simulating process parameters such as temperature, speed, and hydraulic system pressure in actual production. This allows the production process to be rapidly iterated in a virtual environment, which can significantly reduce the cost and time consumption of traditional experimental methods, while avoiding potential material waste in production.

[0014] Setting reasonable ranges for process parameters is fundamental to ensuring optimization effectiveness. By limiting these parameters to a certain range, we can ensure the feasibility of optimization and prevent parameters from exceeding the actual operating limits of the equipment, thus guaranteeing safety and product quality.

[0015] By importing process parameters into a digital twin model for simulation, the actual extrusion process can be modeled, and the corresponding quality and energy consumption indicators can be quickly obtained. The simulation results provide a data foundation for subsequent optimization, enabling the quantification of each parameter and its impact on quality and energy consumption.

[0016] Furthermore, multi-objective optimization algorithms can simultaneously consider the balance between quality and energy consumption, minimizing energy consumption while maintaining product quality requirements. This optimization approach effectively addresses the limitations of traditional optimization methods that focus solely on a single objective, such as minimizing energy consumption or maximizing quality, thus ensuring comprehensive production efficiency. Using a multi-strategy improved multi-objective particle swarm optimization algorithm based on non-dominated sorting, crowding distance calculation, and survival-of-the-fittest selection strategies, the optimal solution satisfying multiple constraints can be quickly found.

[0017] In some embodiments, as a further preferred embodiment, the digital twin model includes:

[0018] A hydraulic system model is used to simulate the working process of the hydraulic components of an aluminum extrusion press. The hydraulic system model accurately describes the behavior of the hydraulic circuit, including pumps, valves, pipelines, and actuators.

[0019] Mechanical system model used to simulate the motion mechanism of extrusion equipment;

[0020] The electrical system model is used to analyze the power consumption and operating efficiency of the motors and control circuits in the extrusion equipment, and to further optimize power distribution.

[0021] The extrusion simulation model is used to model and simulate the plastic deformation process of aluminum profiles and predict the material flow behavior under different temperature and pressure conditions.

[0022] The product quality inspection model combines data from multiple sensors to monitor the changes in the product's shape during the extrusion process in real time, ensuring product quality indicators.

[0023] As a further preferred option, the optimized process parameters are updated to the production equipment in real time through a cloud-based digital twin platform to achieve automated adjustment and energy consumption monitoring.

[0024] As a further preferred embodiment, the quality indicators include strength, hardness, plasticity, and toughness, and the energy consumption indicator is the power consumption of the hydraulic pump in the hydraulic system.

[0025] In some embodiments, as a further preferred embodiment, the multi-objective optimization algorithm is a multi-strategy improved multi-objective particle swarm optimization algorithm based on non-dominated sorting, crowding distance calculation strategy, and survival of the fittest selection strategy, including one or a combination of the following steps:

[0026] Initialize the particle swarm, where each particle represents a set of aluminum profile extrusion process parameter combinations Pi = [Ti, Vi, Phi, Pei, Gi, Mi], where Ti represents the extrusion temperature in the parameter combination represented by the i-th particle, Vi represents the extrusion speed used in the i-th particle, Phi represents the working pressure of the hydraulic system in the i-th particle, Pei represents the power of the electrical system in the i-th particle, Gi represents the die geometry parameters in the i-th particle, and Mi represents the raw material characteristic parameters used in the i-th particle.

[0027] The initial position Xi and velocity Vi of each particle are randomly generated and satisfy the upper and lower limits of the process parameters.

[0028] In each iteration, the energy consumption E(Pi) and quality index Q(Pi) of each particle Pi are simulated and calculated using a digital twin model. The fitness of each particle is determined by its objective function value.

[0029] The position and velocity of the particles are updated using the following formula:

[0030]

[0031] In the formula, Let be the velocity of particle i in the t-th iteration.

[0032] Let i be the position of particle i in the t-th iteration.

[0033] ω: Inertia weight, ranging from [0.4, 0.9], controls the balance between global search and local optimization. A larger ω value (close to 0.9) encourages global search and can help the algorithm escape local optima; a smaller ω value (close to 0.4) encourages local search and helps converge to the optimal solution.

[0034] c1 = c2 = 2.0: Learning factors. c1 controls how much an individual follows its own historical best position, and c2 controls how much an individual follows the group's best position. c1 = c2 = 2.0 is a relatively conservative and effective choice because it balances the influence of the individual and the group, which helps to achieve stable convergence.

[0035] r1, r2: Random numbers, ranging from [0,1].

[0036] P best The historical best position of particle i.

[0037] G best The optimal position among all particles;

[0038] For each particle swarm, a non-dominated ranking is performed based on the bi-objectives of energy consumption and mass. The particle dominance relationship satisfies the condition: if particle Pi is no worse than Pj on all objectives and better than Pj on at least one objective, then Pi dominates Pj, i.e.:

[0039]

[0040] For each particle Pi, calculate the number ni that it is dominated by other particles:

[0041]

[0042] Where II is an indicator function, II = 1 when Pj dominates Pi, otherwise II = 0;

[0043] Particles with a dominance number of 0 form the first layer of non-dominated solutions. These solutions are removed from the particle set. The dominance number of the remaining particles is calculated to form the second layer of non-dominated solutions. This process is repeated until all particles are classified.

[0044] Calculate the crowding distance of each particle in the non-dominated layer to ensure diversity;

[0045] Particles are selected to enter the next generation based on non-dominated sorting and crowding distance, with priority given to solutions that have low energy consumption and meet mass constraints.

[0046] For each non-dominated layer, first select the solution with the larger crowding distance to increase the diversity of solutions. If the number of solutions in the solution set exceeds the requirement, select the solution with lower energy consumption by comparing energy consumption values.

[0047] The optimization process is determined based on the convergence condition, and the optimal aluminum profile extrusion process parameters are output.

[0048] As a further preferred embodiment, the convergence and termination conditions include:

[0049] The energy consumption improvement rate during the optimization process is lower than the preset threshold ∈;

[0050] The maximum number of iterations Tmax is reached;

[0051] The Pareto front remained stable with no significant changes.

[0052] As a further preferred option, the multi-objective optimization algorithm introduces global search and local search mechanisms to accelerate convergence, combines fuzzy logic control and neural networks to assist decision-making in the optimization process, and dynamically adjusts ω, c1, and c2 according to the convergence state of the particle swarm.

[0053] Global and local search mechanisms are introduced into the algorithm to improve the convergence speed and solution quality. During the search process, fuzzy logic control and neural networks are combined to assist in decision-making regarding the search behavior of the particle swarm. Based on the current convergence state of the particle swarm, the inertia weight, learning factor, and search range are dynamically adjusted to ensure that the particle swarm explores a large range of solution space with high global search capability in the early stage of optimization, while gradually transitioning to local search in the later stage of optimization to accurately optimize the details of the solution, thereby improving optimization efficiency and global convergence.

[0054] As a further preferred option, the multi-objective optimization algorithm takes minimizing the energy consumption of the aluminum profile extrusion process as its main objective.

[0055] Based on another key aspect of the present invention, a system is provided for implementing an energy consumption optimization method for an aluminum profile extrusion process based on digital twins, comprising:

[0056] The process parameter setting module is used to set the range of aluminum profile extrusion process parameters;

[0057] The digital twin model module will import the set aluminum profile extrusion process parameters, simulate the extrusion process, and generate simulation results, including energy consumption indicators and product quality indicators.

[0058] The optimization module uses a multi-objective optimization algorithm to optimize the extrusion process parameters of aluminum profiles;

[0059] The feedback module is used to import the optimized process parameters into the digital twin model for re-simulation, generate new energy consumption and quality indicators, and compare them with the results before optimization.

[0060] The control module is used to determine whether the product quality is up to standard and whether the energy consumption has reached the optimization target based on the results of the feedback module. If the conditions are met, the optimized process parameters are output and applied to actual production; otherwise, the process parameters are optimized until the predetermined energy consumption and quality requirements are met.

[0061] By integrating multiple functional modules, intelligent energy consumption optimization is achieved in the aluminum profile extrusion process. This ensures minimal energy consumption while meeting quality requirements, thereby significantly improving energy efficiency and reducing production costs. This process not only reduces energy consumption but also lightens the workload on equipment, extends equipment lifespan, and contributes to achieving green manufacturing and sustainable development goals.

[0062] Based on another key aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed, implements the aforementioned energy consumption optimization method for aluminum profile extrusion process based on digital twins.

[0063] Based on another key aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the aforementioned energy consumption optimization method for aluminum profile extrusion process based on digital twin.

[0064] Advantages and beneficial effects of the present invention:

[0065] This invention achieves digital simulation of the aluminum profile extrusion process by constructing a digital twin model of the aluminum profile extrusion equipment, extrusion process, and product quality inspection. It proposes a multi-strategy improved multi-objective particle swarm optimization algorithm based on non-dominated sorting, crowding distance calculation strategy, and survival of the fittest selection strategy to optimize the aluminum profile extrusion process parameters based on the quality index and energy consumption index values ​​of the digital twin model simulation results, so as to achieve optimal energy consumption in the aluminum profile extrusion process.

[0066] By utilizing digital twin technology, this invention reduces the repetitive experiments and manual interventions required in traditional physical experiments to find optimal energy consumption parameters. This not only significantly reduces experimental costs and intensity but also drastically shortens the R&D cycle, allowing the energy-saving effects of the aluminum profile extrusion process to be fully realized. Furthermore, through the combination of digital twin technology and optimization algorithms, companies can find the optimal process parameters more quickly and accurately, ultimately achieving energy conservation and emission reduction, reducing resource waste, and further improving production efficiency and product consistency. This digital and intelligent technological innovation not only brings significant economic benefits to enterprises but also provides technical support for the sustainable development of the aluminum profile industry. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, obtaining other drawings based on these drawings without creative effort still falls within the scope of the present invention.

[0068] Figure 1 This is a flowchart illustrating the present invention;

[0069] Figure 2 This is a schematic diagram of the multi-objective optimization algorithm of the present invention. Detailed Implementation

[0070] The preferred embodiments of the present invention will be described in detail below to provide a clearer understanding of the purpose, features, and advantages of the invention. It should be understood that the following embodiments are not intended to limit the scope of the invention, but are merely illustrative of the essential spirit of the technical solution of the invention.

[0071] In the following description, certain specific details are set forth for the purpose of illustrating various disclosed embodiments in order to provide a thorough understanding of the various disclosed embodiments. However, those skilled in the art will recognize that embodiments may be practiced without one or more of these specific details. In other instances, well-known apparatuses, structures, and techniques associated with this application may not have been shown or described in detail to avoid unnecessarily obscuring the description of the embodiments.

[0072] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0073] Example

[0074] like Figure 1 The diagram shown is a flowchart of an energy consumption optimization method for aluminum profile extrusion process based on digital twin, provided by an embodiment of the present invention.

[0075] like Figure 2 The diagram shown is a flowchart of a multi-strategy improved multi-objective particle swarm optimization algorithm based on non-dominated sorting, crowding distance calculation strategy, and survival of the fittest selection strategy provided in an embodiment of the present invention.

[0076] 1. Construction of Digital Twin Model

[0077] In this invention, the digital twin model is the core component. It accurately simulates the aluminum profile extrusion process by integrating multiple subsystems, including a hydraulic system model, a mechanical system model, an electrical system model, an extrusion simulation model, and a product quality inspection model. The model is constructed based on advanced physical modeling and simulation technologies, ensuring the accuracy and reliability of the simulation results.

[0078] 1.1 A hydraulic system model is used to simulate the working process of the hydraulic components of an aluminum extrusion press. This system model accurately describes the behavior of the hydraulic circuit, including pumps, valves, pipelines, and actuators. The hydraulic system is a key drive system in the aluminum profile extrusion process, responsible for providing the pressure and power required for the extrusion process. To accurately model the hydraulic system, its behavior is determined by fluid dynamics equations, which describe parameters such as pressure drop and flow velocity as the liquid flows through the hydraulic lines.

[0079] 1.2 Mechanical system model, used to simulate the motion mechanism of the extrusion equipment, simulating the mechanical transmission and motion characteristics of the aluminum profile extrusion equipment. The following aspects are mainly considered:

[0080] The motion and mechanical properties of the equipment are described using the Newton-Euler equations.

[0081] The Coulomb friction model or the viscous friction model is used to simulate the friction loss of the extrusion equipment during the movement process;

[0082] Using the theory of plastic deformation, we analyze the forces and deformation characteristics of aluminum profiles during extrusion and predict the working state of the equipment under different loads.

[0083] 1.3 Electrical system model, used to analyze the power consumption and operating efficiency of motors and control circuits in the extrusion equipment, and further optimize power distribution;

[0084] The motor in the extrusion equipment is modeled using the torque balance equation and power equation of the motor to simulate the relationship between its output power, torque and speed.

[0085] 1.4 The extrusion simulation model is used to model and simulate the plastic deformation process of aluminum profiles and predict the material flow behavior under different temperature and pressure conditions. The extrusion simulation model is the core part of the entire digital twin model and mainly involves the following:

[0086] Based on the stress-strain relationship in plasticity mechanics, the deformation behavior of aluminum under different temperatures and pressures is simulated, and the fluidity of the material is described by plastic flow theory and Navier-Stokes equations.

[0087] The flow behavior of aluminum profiles was numerically solved using the finite element method (FEM) to predict material flow, temperature field, stress field and deformation under different extrusion conditions.

[0088] The effect of temperature changes on material flow during the extrusion process is simulated using coupled temperature and stress field equations.

[0089] 1.5 Product quality inspection model: Combines data from multiple sensors to monitor the shape changes of the product in real time during the extrusion process. Through a virtual test bench in the simulation environment, virtual tests are conducted on strength, hardness, toughness, etc., to predict the final physical properties of the product and ensure product quality indicators.

[0090] 1.6 Digital Simulation: After modeling is completed, numerical simulation tools (such as Ansys, MATLAB Simulink, COMSOL, etc.) are used to perform digital simulations of the hydraulic system, mechanical system, and electrical system, and the simulations are compared and verified with actual production data. Quality index values ​​Q (Pi) and energy consumption index values ​​E (Pi) are extracted from the simulation results of the digital twin model.

[0091] 2. Multi-objective optimization algorithm

[0092] After the simulation is completed, the multi-strategy improved multi-objective particle swarm optimization algorithm based on non-dominated sorting, crowding distance calculation strategy and survival of the fittest selection strategy is used to optimize the aluminum profile extrusion process parameters according to the quality index and energy consumption index values ​​of the digital twin model simulation results, so as to achieve the optimal energy consumption of the aluminum profile extrusion process.

[0093] 2.1 Improvement Strategies for Multi-Objective Particle Swarm Optimization (MOPSO) Algorithm

[0094] MOPSO is a classic optimization algorithm that finds the optimal solution to a problem by simulating the swarm behavior of particles. It is particularly effective in multi-objective optimization, capable of simultaneously optimizing multiple conflicting objectives. For this invention, the two objectives are:

[0095] Minimize energy consumption: This means reducing the energy consumption required during the aluminum profile extrusion process by optimizing parameters;

[0096] Maximizing quality means ensuring key quality indicators such as product strength and toughness while reducing energy consumption.

[0097] The primary objective is to minimize energy consumption during the aluminum profile extrusion process.

[0098] In multi-objective optimization, the Pareto front is an effective way to describe the trade-offs between multiple optimization objectives. This optimization process is improved through three core strategies: non-dominated sorting, crowding distance calculation, and a survival-of-the-fittest selection strategy.

[0099] 2.2 Non-dominated sorting

[0100] Non-dominated sorting is the process of classifying a swarm of particles according to whether they are dominated. This algorithm divides all particles into several layers: the first layer consists of non-dominated solutions, i.e., solutions on the Pareto front; the second layer consists of solutions dominated only by solutions in the first layer; and so on.

[0101] For each particle swarm, a non-dominated ranking is performed based on the bi-objectives of energy consumption and mass. The particle dominance relationship satisfies the condition: if particle Pi is no worse than Pj on all objectives and better than Pj on at least one objective, then Pi dominates Pj, i.e.:

[0102]

[0103] Let ni be the number of times the particle is dominated by other particles, initialized to 0.

[0104] Let Si be the set of all particles dominated by Pi.

[0105] For each particle Pi, calculate the number ni that it is dominated by other particles:

[0106]

[0107] Where II is an indicator function, II = 1 when Pj dominates Pi, otherwise II = 0;

[0108] Particles with a dominance number of ni of 0 form the first layer of non-dominated solutions. These solutions are removed from the particle set and added to the Pareto front.

[0109] For each particle Pi belonging to the first-level non-dominated solution, find all particles in the set Si that it dominates, and perform the following operation on these particles:

[0110] For each particle Pj∈Si dominated by Pi, its domination count is decremented by one, i.e., nj=nj-1.

[0111] If nj = 0, then add Pj to the second layer of non-dominated solutions.

[0112] Repeat this process until all particles have been assigned to a certain level.

[0113] Through this process, all particles are stratified, with the first layer representing the optimal Pareto front and subsequent layers representing successively better solutions.

[0114] 2.3 Crowding Distance Calculation

[0115] Crowding distance is used to evaluate the distribution of solutions on the Pareto front. Its purpose is to maintain the diversity of the Pareto solution set and prevent particles from concentrating in a local region of the front.

[0116] The crowding distance D(Pi) calculates the spatial sparsity around each particle.

[0117] For each particle Pi on the Pareto front, its crowding distance is initialized to 0. For each objective function, the particles are sorted, and the minimum and maximum values ​​are found. The crowding distances of the solutions with the minimum and maximum objective function values ​​on the Pareto front are set to infinity, as they are located on the boundary and are often solutions that need to be retained.

[0118] For the middle particles, the crowding distance is calculated using the following formula:

[0119]

[0120] Among them, f m (P i+1 ) and f m (P i-1 () are particles P on target m. i+1 and P i-1 The objective function value, and The objective function f m The maximum and minimum values.

[0121] Particles with larger crowding distances are located in sparse regions at the forefront and are given priority for retention; particles with smaller crowding distances are more likely to be eliminated.

[0122] 2.4 Survival of the fittest selection strategy

[0123] At the end of each iteration, the solutions in the current population are compared with the solutions in the historical population, and the best solution is selected to enter the next generation.

[0124] Combining non-dominated sorting and crowding distance, the survival-of-the-fittest selection strategy ensures that the solution with the lowest energy consumption and that meets the quality standard is retained, inferior solutions are gradually eliminated, and the optimal solution at the Pareto front is retained.

[0125] For each non-dominated layer, first select the solution with the larger crowding distance to increase the diversity of solutions. If the number of solutions in the solution set exceeds the requirement, select the solution with lower energy consumption by comparing energy consumption values.

[0126] 2.5 Strategies to Accelerate Convergence

[0127] 2.51 Global and Local Search Mechanisms

[0128] Global search and local search are two search mechanisms in particle swarm optimization algorithms:

[0129] Global search: In the early stages of optimization, the algorithm needs a larger exploration range to avoid getting trapped in local optima. Global search increases the search range of particles in the solution space by using a larger inertia weight ω, thereby improving exploration capability.

[0130] Local search: As optimization progresses, the inertia weight ω is gradually reduced, causing particles to focus more on the vicinity of the current optimal solution, thereby improving the ability to refine the optimal solution and enhancing the algorithm's development capabilities.

[0131] The dynamic adjustment strategy for the inertia weight ω is based on the following formula:

[0132]

[0133] In the formula, ω(t): the inertial weight of the t-th generation.

[0134] ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively: 0.9 and 0.4.

[0135] t: Current iteration number

[0136] T max Maximum number of iterations.

[0137] In this way, the inertia weight gradually decreases with the number of iterations, thus achieving a smooth transition from global search to local search. This dynamic adjustment effectively improves the convergence speed and avoids getting trapped in local optima.

[0138] 2.52 Introducing fuzzy logic control and neural network-assisted decision-making

[0139] To further optimize the convergence speed, fuzzy logic control and neural networks are combined to dynamically adjust the parameters in the algorithm.

[0140] Fuzzy rules are established based on key parameters in the aluminum profile extrusion process. The fuzzy controller adjusts the weighting of global and local searches in each iteration according to these rules to ensure appropriate search intensity at each stage. Fuzzy logic control adjusts the input variables and reasoning based on the rule base, considering the current solution distribution and convergence trend. When convergence is slow, ω is increased or the learning factor is adjusted to enhance the global search; when convergence is fast but diversity is low, ω is decreased to strengthen the local search.

[0141] Neural networks can judge different states during the optimization process based on training data and output appropriate parameter adjustment suggestions. During the optimization process, the neural network outputs adjustment suggestions for ω, c1, and c2 in real time based on the current particle distribution and convergence state, thereby accelerating convergence.

[0142] During the optimization process, the fuzzy controller and the neural network work together. The fuzzy controller evaluates the current optimization state in real time and adjusts the optimization direction according to rules; the neural network fine-tunes the optimization parameters based on historical data it has learned.

[0143] 2.6 Multi-objective optimization process and calculation

[0144] Initialize the particle swarm, where each particle represents a set of aluminum profile extrusion process parameter combinations Pi = [Ti, Vi, Phi, Pei, Gi, Mi], where Ti represents the extrusion temperature in the parameter combination represented by the i-th particle, Vi represents the extrusion speed used in the i-th particle, Phi represents the working pressure of the hydraulic system in the i-th particle, Pei represents the power of the electrical system in the i-th particle, Gi represents the die geometry parameters in the i-th particle, and Mi represents the raw material characteristic parameters used in the i-th particle.

[0145] The initial position Xi and velocity Vi of each particle are randomly generated and satisfy the upper and lower limits of the process parameters.

[0146] In each iteration, the energy consumption E(Pi) and quality index Q(Pi) of each particle Pi are simulated and calculated using a digital twin model. The fitness of each particle is determined by its objective function value.

[0147] The position and velocity of the particles are updated using the following formula:

[0148]

[0149] In the formula, Let be the velocity of particle i in the t-th iteration.

[0150] Let i be the position of particle i in the t-th iteration.

[0151] ω: Inertia weight, ranging from [0.4, 0.9], controls the balance between global search and local optimization. A larger ω value (close to 0.9) encourages global search and can help the algorithm escape local optima; a smaller ω value (close to 0.4) encourages local search and helps converge to the optimal solution.

[0152] c1 = c2 = 2.0: Learning factors. c1 controls how much an individual follows its own historical best position, and c2 controls how much an individual follows the group's best position. Choosing c1 = c2 = 2.0 is a relatively conservative and effective choice because it balances the influence of individuals and the group, which helps to achieve stable convergence.

[0153] r1, r2: Random numbers, ranging from [0,1].

[0154] P best The historical best position of particle i.

[0155] G best The optimal position among all particles;

[0156] For each particle swarm, a non-dominated ranking is performed based on the bi-objectives of energy consumption and mass. The particle dominance relationship satisfies the condition: if particle Pi is no worse than Pj on all objectives and better than Pj on at least one objective, then Pi dominates Pj, i.e.:

[0157]

[0158] For each particle Pi, calculate the number ni that it is dominated by other particles:

[0159]

[0160] Where II is an indicator function, II = 1 when Pj dominates Pi, otherwise II = 0;

[0161] Particles with a dominance number of 0 form the first layer of non-dominated solutions. These solutions are removed from the particle set. The dominance number of the remaining particles is calculated to form the second layer of non-dominated solutions. This process is repeated until all particles are classified.

[0162] Calculate the crowding distance of each particle in the non-dominated layer to ensure diversity;

[0163] Particles are selected to enter the next generation based on non-dominated sorting and crowding distance, with priority given to solutions that have low energy consumption and meet mass constraints.

[0164] For each non-dominated layer, first select the solution with the larger crowding distance to increase the diversity of solutions. If the number of solutions in the solution set exceeds the requirement, select the solution with lower energy consumption by comparing energy consumption values.

[0165] The optimization process is determined based on the convergence condition, and the optimal aluminum profile extrusion process parameters are output.

[0166] As a further preferred embodiment, the convergence and termination conditions include:

[0167] The energy consumption improvement rate during the optimization process is lower than the preset threshold ∈;

[0168] If the rate of change of the optimal energy consumption value over G consecutive generations... If the value is less than the set threshold ∈, then it is considered converged.

[0169] The optimization process terminates when the number of algorithm iterations reaches the maximum number of iterations Tmax.

[0170] The Pareto front is stable and shows no significant change. Calculate the stability distance of the Pareto front solution set.

[0171]

[0172] In the formula, N is the number of Pareto front solutions.

[0173] and and are the Pareto front solutions for generation t and generation (t+1), respectively.

[0174] If D is consecutively within G generations Pareto If the distance is below the set stability distance threshold δ of the Pareto front, it is considered convergent.

[0175] Any aspects of this invention not described in detail are well-known to those skilled in the art.

[0176] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing energy consumption in aluminum profile extrusion processes based on digital twins, characterized in that, Includes the following steps: Step 1: Construct a digital twin model of the aluminum profile extrusion process; Step 2: Set the range of various process parameters for aluminum profile extrusion. The process parameters include extrusion temperature T, extrusion speed V, hydraulic system pressure Ph, electrical system power Pe, die geometry G, and raw material characteristic parameters M. Step 3: Import the preset aluminum profile extrusion process parameters into the digital twin model, simulate the aluminum profile extrusion forming process, and extract the quality index value Q (Pi) and energy consumption index value E (Pi) from the simulation results of the digital twin model. Step 4: Using the quality index and energy consumption index values ​​obtained in Step 3 as constraints, optimize the aluminum profile extrusion process parameters using a multi-objective optimization algorithm. Step 5: Import the optimized aluminum profile extrusion process parameters into the digital twin model, simulate the aluminum profile extrusion forming process again, and extract the quality index and energy consumption index values. Step 6: Compare the aluminum profile extrusion process parameters and product quality indicators before and after optimization to determine whether the product quality is qualified and whether the energy consumption has reached the optimal level. If the product quality is qualified and the energy consumption is optimal, the optimized aluminum profile extrusion process parameters will be applied to actual production. Otherwise, the quality index and energy consumption index values ​​obtained in step 5 will be used as constraints, and the multi-objective optimization algorithm described in step 4 will be used to optimize the aluminum profile extrusion process parameters. Steps 5 and 6 will be repeated until the product quality is qualified and the energy consumption is optimal. The multi-objective optimization algorithm is a multi-strategy improved multi-objective particle swarm optimization algorithm based on non-dominated sorting, crowding distance calculation strategy, and survival of the fittest selection strategy, including: Initialize the particle swarm, where each particle represents a set of aluminum profile extrusion process parameter combinations Pi = [Ti, Vi, Phi, Pei, Gi, Mi], where Ti represents the extrusion temperature in the parameter combination represented by the i-th particle, Vi represents the extrusion speed used in the i-th particle, Phi represents the working pressure of the hydraulic system in the i-th particle, Pei represents the power of the electrical system in the i-th particle, Gi represents the die geometry parameters in the i-th particle, and Mi represents the raw material characteristic parameters used in the i-th particle. The initial position Xi and velocity Vi of each particle are randomly generated and satisfy the upper and lower limits of the process parameters. In each iteration, the energy consumption E(Pi) and quality index Q(Pi) of each particle Pi are simulated and calculated using a digital twin model. The fitness of each particle is determined by its objective function value. The position and velocity of the particles are updated using the following formula: In the formula, Let be the velocity of particle i in the t-th iteration. Let i be the position of particle i in the t-th iteration. ω: Inertia weight, ranging from [0.4, 0.9], used to control the balance between large-scale search and local optimization. c1 = c2 = 2.0: Learning factors, where c1 controls how much an individual follows its own historical best position, and c2 controls how much an individual follows the group's best position. r1, r2: Random numbers, ranging from [0,1]. P best The historical best position of particle i. G best The optimal position among all particles; For each particle swarm, a non-dominated ranking is performed based on the bi-objectives of energy consumption and mass. The particle dominance relationship satisfies the condition: if particle Pi is no worse than Pj on all objectives and better than Pj on at least one objective, then Pi dominates Pj, i.e.: P i Dominate And Q(P) i )≥Q min ) or (E(P) i ) <E(P j And Q(P) i )>Q(P j )) For each particle Pi, calculate the number ni that it is dominated by other particles: Where II is an indicator function, II = 1 when Pj dominates Pi, otherwise II = 0; Particles with a dominance number of 0 form the first layer of non-dominated solutions. These solutions are removed from the particle set. The dominance number of the remaining particles is calculated to form the second layer of non-dominated solutions. This process is repeated until all particles are classified. Calculate the crowding distance of each particle in the non-dominated layer to ensure diversity; Particles are selected to enter the next generation based on non-dominated sorting and crowding distance, with priority given to solutions that have low energy consumption and meet mass constraints. For each non-dominated layer, first select the solution with the larger crowding distance to increase the diversity of solutions. If the number of solutions in the solution set exceeds the requirement, select the solution with lower energy consumption by comparing energy consumption values. The optimization process is determined based on the convergence condition, and the optimal aluminum profile extrusion process parameters are output.

2. The energy consumption optimization method for aluminum profile extrusion process based on digital twin as described in claim 1, characterized in that, The digital twin model includes: A hydraulic system model is used to simulate the working process of the hydraulic components of an aluminum extrusion press. The hydraulic system model accurately describes the behavior of the hydraulic circuit, including pumps, valves, pipelines, and actuators. Mechanical system model used to simulate the motion mechanism of extrusion equipment; The electrical system model is used to analyze the power consumption and operating efficiency of the motors and control circuits in the extrusion equipment, and to further optimize power distribution. Extrusion simulation model is used to model and simulate the plastic deformation process of aluminum profiles; The product quality inspection model combines data from multiple sensors to monitor the changes in the product's shape during the extrusion process in real time.

3. The energy consumption optimization method for aluminum profile extrusion process based on digital twin as described in claim 1, characterized in that, The quality indicators include strength, hardness, plasticity, and toughness, and the energy consumption indicator is the power consumption of the hydraulic pump in the hydraulic system.

4. The energy consumption optimization method for aluminum profile extrusion process based on digital twin as described in claim 1, characterized in that, The convergence conditions include: The energy consumption improvement rate during the optimization process is lower than the preset threshold ∈; The maximum number of iterations Tmax is reached; The Pareto front remained stable with no significant changes.

5. The energy consumption optimization method for aluminum profile extrusion process based on digital twin as described in claim 1, characterized in that, The multi-objective optimization algorithm introduces global and local search mechanisms, combines fuzzy logic control and neural networks to assist in the optimization process, and dynamically adjusts ω, c1, and c2 according to the convergence state of the particle swarm.

6. The energy consumption optimization method for aluminum profile extrusion process based on digital twin as described in claim 1, characterized in that, The multi-objective optimization algorithm described above takes minimizing the energy consumption of the aluminum profile extrusion process as its main objective.

7. A system for implementing the energy consumption optimization method for aluminum profile extrusion process based on digital twins as described in any one of claims 1-6, characterized in that, include: The process parameter setting module is used to set the range of aluminum profile extrusion process parameters; The digital twin model module imports the set aluminum profile extrusion process parameters, simulates the extrusion process, and generates simulation results, including energy consumption indicators and product quality indicators. The optimization module uses a multi-objective optimization algorithm to optimize the extrusion process parameters of aluminum profiles; The feedback module is used to import the optimized process parameters into the digital twin model for re-simulation, generate new energy consumption and quality indicators, and compare them with the results before optimization. The control module is used to determine whether the product quality is up to standard and whether the energy consumption has reached the optimization target based on the results of the feedback module. If the conditions are met, the optimized process parameters are output and applied to actual production; otherwise, the process parameters are optimized until the predetermined energy consumption and quality requirements are met.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed, it implements the energy consumption optimization method for aluminum profile extrusion process based on digital twins as described in any one of claims 1-6.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the energy consumption optimization method for aluminum profile extrusion process based on digital twin as described in any one of claims 1-6.

Images