Modular flexible digital twin layout optimization method and system

Abstract

The invention relates to the technical field of modular assembly production lines, in particular to a modular flexible digital twin layout optimization method and system. Comprising the following steps: constructing multi-dimensional constraint modeling, constructing a multi-objective optimization model containing route length, equipment cost and human resource quantity, and generating an initial layout scheme based on three-dimensional geometric attributes and logistics paths of production line equipment; dynamic weight distribution is carried out, according to priority parameters input by a user, a route optimization weight # imgabs0 #, a cost optimization weight # imgabs1 # and a human resource weight # imgabs2 # are adjusted in real time, and a dynamic weight combination matrix is generated; digital twinborn simulation is carried out, the initial layout scheme is mapped to a digital twinborn platform, and the actual equipment utilization rate, the material carrying efficiency and the worker operation path are calculated through logistics path simulation; according to the scheme, dynamic balance multiple targets can be provided based on a simulation result, simulation and optimization real-time linkage can be realized, and a layout optimization technology of a multi-dimensional verifiable scheme is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of modular assembly production lines, and particularly to a modular flexible digital twin layout optimization method and system. Background Art

[0002] With the upgrading of the requirements for intelligent manufacturing and flexible production, the island-type assembly production line has become a key layout form in the discrete manufacturing field due to its characteristics such as modularity and reconfigurability. However, the traditional production line layout optimization methods have significant defects in aspects such as dynamics and multi-objective collaboration. Static optimization objectives are rigidified: Existing technologies mostly adopt fixed weight allocation strategies and cannot dynamically adjust the priority of optimization objectives according to scenarios such as production order changes and equipment failure rate fluctuations, resulting in the disconnection between the scheme and the actual requirements. For example, when an emergency order requires cost compression, the traditional method still takes path efficiency as the leading objective and it is difficult to generate the optimal cost scheme. Simulation and optimization are separated: The current digital twin technology focuses on three-dimensional scene visualization, and the simulation data cannot be fed back to the optimization algorithm in real time, forming a one-way process of "simulation - optimization" with low iteration efficiency. Especially in large-scale production lines, due to the lack of a real-time closed-loop tuning mechanism, the adjustment of the layout scheme takes up to several hours, making it difficult to support the rapid production change requirements. Insufficient multi-dimensional decision-making support: Existing methods usually rely on a single optimization scheme and lack the ability to quantitatively compare multiple objectives such as cost, efficiency, and human resources. Decision-makers need to manually analyze the advantages and disadvantages of different schemes, which is prone to selecting sub-optimal schemes due to subjective judgment biases and cannot predict the safety of the scheme.

[0003] Based on the above problems, there is an urgent need for a layout optimization technology that can dynamically balance multiple objectives, achieve real-time linkage between simulation and optimization, and provide multi-dimensional verifiable schemes to meet the production line design requirements of high flexibility, high precision, and high safety in the intelligent manufacturing scenario. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and a modular flexible digital twin layout optimization method is proposed, including:

[0005] S1: Construct a multi-dimensional constraint model, construct a multi-objective optimization model including route length, equipment cost, and the number of human resources, and generate an initial layout scheme based on the three-dimensional geometric attributes of the production line equipment and the logistics path;

[0006] S2: Perform dynamic weight allocation, and adjust the route optimization weight , cost optimization weight , human resource weight in real time according to the priority parameters input by the user, and generate a dynamic weight combination matrix;

[0007] S3: Conduct digital twin simulation, map the initial layout plan to the digital twin platform, and calculate the actual equipment utilization rate, material handling efficiency, and worker operation path through logistics path simulation;

[0008] S4: Real-time linkage optimization, based on the simulation results, use the gradient descent algorithm to iteratively optimize the equipment coordinates and logistics path, and update the dynamic weight combination matrix until the convergence threshold is reached;

[0009] S5: Generate and verify multiple plans, generate at least three optimized plans by adjusting the dynamic weight combination matrix, and verify the layout feasibility based on the collision detection module of the digital twin platform.

[0010] Adopt the above technical solution: This solution can realize adjusting the combined ratio of the three parameters of α (route optimization weight), β (cost optimization weight), and γ (human resource weight) in real time according to the user's cost priority or efficiency priority requirements, forming a dynamic weight matrix; by integrating the three-dimensional geometric data of equipment, logistics path, and human resources, construct a multi-objective optimization model covering route length, equipment cost, and human resource quantity, and generate an initial layout plan; it can solve the problem that the existing technology only supports fixed weight parameters, cannot dynamically adjust the optimization target priority according to actual needs, and can only generate a single optimized plan, unable to meet the multi-dimensional decision-making needs in complex scenarios.

[0011] Preferably, the total cost function of the multi-objective optimization model in S1 is:

[0012] ,

[0013] Among them, is the total cost; is the purchase cost of the th equipment; is the unit distance cost of the th section of the logistics path; is the actual length of the th section of the logistics path; is the average annual maintenance cost of the th type of equipment; is the total number of equipment; is the number of equipment types.

[0014] Adopt the above technical solution: The traditional model cannot update the logistics path length in real time , resulting in cost calculation deviation; this solution can realize that when the logistics path changes due to layout adjustment, update in real time and feedback to the cost function to ensure the accuracy of the calculation results.

[0015] Further preferably, the path optimization function of the dynamic weight combination matrix in S2 is:

[0016] ,

[0017] Among them, is the path comprehensive optimization index; is the actual length of the k-th path; is the theoretical shortest length of the k-th path; is the current total cost; is the benchmark cost; is the number of workers required, is the number of available workers; is the total number of logistics paths.

[0018] Adopt the above technical solution: This solution quantifies the path length deviation, cost overrun rate, and human resource gap rate into a unified index , which is convenient for comparing solutions; through the dynamic adjustment of α, β, and γ, it supports the collaborative optimization of path efficiency improvement, cost compression, and human resource saving.

[0019] Further preferably, the iterative formula of the gradient descent algorithm described in S4 is:

[0020] ,

[0021] Among them, is the device coordinate vector of the -th iteration; is the learning rate, and its value range is from 0.01 to 0.1; is the gradient of the objective function at ; , and are the partial derivatives of the weight parameters with respect to the objective function, respectively.

[0022] Adopt the above technical solution: For traditional optimization algorithms (such as genetic algorithms), it is easy to fall into local optima, resulting in sub-optimal layout solutions. Synchronously update the device coordinates and the weight parameters (α, β, γ). Compared with traditional step-by-step optimization methods, it can reduce the number of convergence iterations.

[0023] Further preferably, the initial layout solution described in S1 is generated in the following way:

[0024] S101: Device coordinate allocation: Based on the three-dimensional size of the device and the safety distance constraint, use the greedy algorithm to sequentially allocate coordinates in the production line space according to the device priority, and preferentially place high-frequency interaction devices in the central area;

[0025] S102: Logistics path planning: Use the Dijkstra algorithm to calculate the shortest feasible path between devices, and dynamically adjust the path width and turning radius in combination with the AGV operation parameters;

[0026] S103: Permutation and combination optimization: Generate an initial device permutation and combination through a genetic algorithm, and screen the optimal solution with the device interaction frequency as the fitness function;

[0027] S104: Three-dimensional space verification: Pre-verify the space occupancy rate and channel accessibility of the initial layout through a digital twin platform.

[0028] Adopt the above technical solutions: For the traditional method of randomly generating the initial layout, it leads to an increase in the number of subsequent optimization iterations; and using a fixed path template, it cannot adapt to the AGV dynamic parameters. This solution generates the initial layout through the cooperation of the greedy algorithm and the genetic algorithm. Compared with the random generation method, it can reduce the number of subsequent optimization iterations. Dynamically adjust the path in combination with the AGV parameters, and the AGV channel qualification rate of the initial plan is increased from 70% of the traditional method to 95%.

[0029] Further preferably, the digital twin platform described in S3 realizes simulation through the following modules:

[0030] Device behavior simulation module: Based on the device PLC signal and processing cycle data, simulate the start-stop state, failure rate, and energy consumption curve of the device;

[0031] Logistics path rendering module: Use the A* algorithm to update the material handling path in real time and superimpose the dynamic obstacle avoidance strategy;

[0032] Human resource scheduling module: Allocate operation tasks according to the worker skill level library, and simulate the worker movement trajectory and collaborative operation conflicts;

[0033] Efficiency evaluation module: Real-time statistics of equipment utilization rate, material retention time, and worker idle rate, and generate a simulation report.

[0034] Adopt the above technical solutions: This solution can optimize the algorithm adjustment accuracy by outputting KPI data in real time through the efficiency evaluation module; integrate the simulation of equipment, logistics, and human resources, and reduce the human conflict rate of the layout plan from 20% to 5%.

[0035] Further preferably, the implementation method of the multi-scheme generation described in S5 includes:

[0036] Cost-dominated scheme: Fix The weight is 0.2, and The weight is increased to 0.6, The weight is set to 0.2 to reduce the equipment procurement and logistics costs;

[0037] Human resource saving plan: fixed The weight is 0.5, and the weight is set to 0.3, the weight is set to 0.2, giving priority to reducing the number of workers and skill dependence;

[0038] Comprehensive balanced plan: Set , , the weights to 0.4, 0.3, and 0.3 respectively to balance the path length, cost, and human resource requirements;

[0039] Custom weight plan: Allow users to manually adjust the weight ratio through the interactive interface to generate a personalized layout in real time.

[0040] Adopting the above technical solutions: This application can generate at least three typical solutions, and the decision-making time for decision-makers is shortened from 2 hours to 20 minutes. It solves the problem that traditional methods can only generate a single optimized solution and cannot meet the decision-making needs of multiple scenarios. The custom weight function improves the configurability of the solution by 80%, especially suitable for small and medium batch customized production lines. It solves the problem that existing technologies only support preset weight combinations and lack user-defined functions.

[0041] Further preferably, the determination condition of the convergence threshold is:

[0042] ,

[0043] where is a preset error threshold, and its value range is from 0.01 to 0.05; is the path comprehensive optimization index of the t-th iteration.

[0044] Further preferably, the collision detection module in S5 verifies the feasibility of the layout in the following ways:

[0045] Equipment collision detection: Detect the minimum distance between equipment based on the axial bounding box of the equipment three-dimensional model to ensure compliance with safety operation standards;

[0046] Personnel safety area verification: Calculate the safety distance between the worker operation area and the moving parts of the equipment. If the overlap rate exceeds 5%, it is determined as a conflict;

[0047] Logistics channel verification: Detect whether the straight-line segment width of the AGV running channel is greater than 1.2 times the AGV body width, and whether the turning radius is greater than the minimum turning radius;

[0048] Emergency escape channel inspection: Verify that the main production line channel is connected and the width is not less than 2 meters, meeting the requirements of fire protection regulations.

[0049] Adopting the above technical solution: This solution reduces the misjudgment rate to less than 5% based on the three-dimensional detection of AABB; the verification of the escape route ensures that the layout complies with the fire protection regulations by 100%, reducing the risk of production accidents.

[0050] A system, applied to a modular flexible digital twin layout optimization method as described in any one of the above, characterized by including:

[0051] ‌Multi-objective modeling unit‌: used to analyze the equipment parameter library, process route library, and human resource library, construct a multi-dimensional constraint model, and output an initial layout plan;

[0052] ‌Dynamic weight allocation unit‌: Based on the optimization objectives input by the user, adjust 、 、 weight parameters in real time through a slider or a numerical input interface;

[0053] ‌Digital twin simulation engine‌: integrated with Unity3D or an equivalent three-dimensional engine, supporting equipment behavior simulation, logistics path rendering, and human resource scheduling;

[0054] ‌Linkage tuning controller‌: embedded with gradient descent algorithm and genetic algorithm, iteratively optimize equipment coordinates and path parameters according to simulation feedback data;

[0055] ‌Multi-scheme decision-making module‌: provides a view for comparing schemes, supporting multi-dimensional screening based on cost, efficiency, and safety, and visualizing the collision detection results.

[0056] Adopting the above technical solution: The present invention can achieve real-time interaction among the modeling, simulation, and optimization modules, and reduce the data delay from 10 minutes to within 5 seconds. The view for comparing schemes improves the decision-making efficiency by 70%, and is especially suitable for the rapid reconstruction of multi-batch production lines. Brief Description of the Drawings

[0057] Figure 1 It is a flowchart of the modular flexible digital twin layout optimization method of this application;

[0058] Figure 2 It is a block diagram of the modular flexible digital twin layout optimization system of this application. Detailed Embodiments

[0059] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0060] Please refer to Figure 1, for the existing technology that only supports fixed weight parameters and cannot dynamically adjust the optimization target priority according to actual requirements. There is a disconnect between simulation and optimization. The existing digital twin technology focuses on visual display and lacks the real-time interaction ability between simulation data and optimization algorithms. The solutions are relatively single. Traditional methods generate a single optimization solution and cannot meet the multi-dimensional decision-making requirements in complex scenarios. Based on the above problems, a modular flexible digital twin layout optimization method proposed in this application includes:

[0061] S1: Construct a multi-dimensional constraint model. Build a multi-objective optimization model that includes route length, equipment cost, and the number of human resources, and generate an initial layout plan based on the three-dimensional geometric attributes of production line equipment and the logistics path.

[0062] S2: Perform dynamic weight allocation. According to the priority parameters input by the user, adjust the route optimization weight , cost optimization weight , human resources weight in real time, and generate a dynamic weight combination matrix.

[0063] S3: Perform digital twin simulation. Map the initial layout plan to the digital twin platform, and calculate the actual equipment utilization rate, material handling efficiency, and worker operation path through logistics path simulation.

[0064] S4: Real-time linkage optimization. Based on the simulation results, use the gradient descent algorithm to iteratively optimize the equipment coordinates and logistics path, and update the dynamic weight combination matrix until the convergence threshold is reached.

[0065] S5: Generate and verify multiple solutions. Generate at least three optimization solutions by adjusting the dynamic weight combination matrix, and verify the layout feasibility based on the collision detection module of the digital twin platform.

[0066] The embodiments of this application include five core steps. Multi-dimensional constraint modeling: By integrating three-dimensional geometric data of equipment (such as size, installation space), logistics path (such as AGV running trajectory), and human resources (such as worker skill level), build a multi-objective optimization model covering route length, equipment cost, and the number of human resources, and generate an initial layout plan.

[0067] Dynamic weight allocation: According to user requirements (such as cost priority or efficiency priority), adjust the combined ratio of the three parameters α (route optimization weight), β (cost optimization weight), and γ (human resources weight) in real time to form a dynamic weight matrix.

[0068] Digital twin simulation: Map the initial layout to a three-dimensional digital twin platform, simulate the material handling path, equipment processing cycle, and worker movement trajectory, and calculate key indicators such as equipment utilization rate and material residence time.

[0069] Real-time linkage optimization: Based on simulation results, a gradient descent algorithm is used to iteratively adjust equipment coordinates and logistics paths, and the weight matrix is simultaneously updated until convergence conditions are met (e.g., the rate of change of the path comprehensive optimization index is ≤0.05).

[0070] Multi-solution generation and verification: Generate at least three optimization solutions using preset weight combinations (e.g., α=0.6, β=0.2, γ=0.2). Use the collision detection module to verify safety indicators such as equipment spacing and logistics channel width.

[0071] It is worth mentioning that this application supports various optimization modes such as cost-driven and efficiency-first through real-time adjustment of α, β, and γ weights to adapt to changing process route scenarios.

[0072] Digital twin simulation and gradient descent algorithm form a real-time feedback loop, which improves iteration efficiency by more than 60% compared to traditional offline optimization methods.

[0073] Generate at least three verifiable solutions (such as the lowest cost solution, the shortest path solution, and the human resource balance solution) to reduce decision-making risks.

[0074] Traditional technical solutions have the following technical problems: one-sided cost accounting: existing technology only considers equipment procurement costs, ignoring logistics and maintenance costs, resulting in an underestimation of total costs by 15%-30%.

[0075] ‌Dynamic Data Missing‌: Traditional models cannot update the logistics path length in real time, resulting in cost calculation errors. Based on this, the total cost function of the multi-objective optimization model described in S1 is:

[0076] ,

[0077] in, is the total cost; For the The purchase cost of the equipment; For the Unit distance cost of the logistics route; For the The actual length of the logistics path; For the Average annual maintenance cost of this type of equipment; is the total number of devices; The number of device types.

[0078] In the above formula, the equipment purchase cost is Automatically match price data according to the equipment model library (such as domestic equipment library and imported equipment library). For example, the unit price of a domestic welding robot is 500,000 yuan, while the imported model is 800,000 yuan.

[0079] The specific calculation formula is as follows:

[0080] Where is the motor power of the kth AGV (Automated Guided Vehicle); is the running time; is the unit price of energy consumption; is the sensor calibration cost coefficient; is the calibration frequency.

[0081] Logistics cost Combined with the energy consumption per unit distance of the AGV (such as 0.5 kWh / m) and the unit price of electricity (such as 0.8 yuan / kWh), the logistics cost is dynamically calculated. For example, if the length of a certain path is 30 meters, the logistics cost is 0.5×30×0.8 = 12 yuan.

[0082] Maintenance cost : Based on the historical failure rate of the equipment (such as the average annual failure times of CNC machine tools ≥ 3 times) and the maintenance unit price (such as the single maintenance cost of 5000 yuan), the maintenance cost is statistically calculated according to the equipment type.

[0083] It is worth mentioning that this embodiment can achieve full - life - cycle cost coverage, integrating acquisition, logistics, and maintenance costs, and reducing the cost calculation error from ±20% of the traditional method to ±5%.

[0084] And it can perform dynamic parameter adaptation. When the logistics path changes due to layout adjustment, it is updated in real - time and fed back to the cost function to ensure the accuracy of the calculation result.

[0085] For the limitations of single - dimension optimization in the existing technology, the existing technology only optimizes the path length or cost and cannot balance the multi - objective conflict. Moreover, there is a defect of fixed weights in the traditional method, which uses fixed weight ratios (such as α = 0.5, β = 0.3, γ = 0.2) and cannot adapt to dynamic production requirements.

[0086] Based on this, the path optimization function of the dynamic weight combination matrix described in S2 is:

[0087] ,

[0088] Where is the path comprehensive optimization index; is the actual length of the kth path; is the theoretical shortest length of the kth path; is the current total cost; is the benchmark cost; is the required number of workers, is the number of available workers; is the total number of logistics paths.

[0089] In the above formula, the route optimization weight has a value range of: ; the cost optimization weight has a value range of: , and the human resources weight has a value range of: .

[0090] When the equipment failure rate exceeds 5% or the order priority changes, the value of

[0091] In the above formula is calculated as:

[0092] where is the skill level weight (e.g., junior = 1.0, senior = 1.5), is the number of people with the corresponding skill, is the equipment preparation time, is the production beat.

[0093] can quantify the path efficiency, such as the ratio of the actual path length to the Dijkstra theoretical shortest length;

[0094] can reflect the cost control level, such as the ratio of the current cost to the industry benchmark cost;

[0095] can measure the utilization rate of human resources, such as the ratio of the required number of workers to the available number of workers;

[0096] It is worth mentioning that: this solution quantifies the path length deviation, cost overrun rate, and man - power gap rate into a unified index , which is convenient for comparing solutions; through the dynamic adjustment of α, β, and γ, it supports the collaborative optimization of path efficiency improvement, cost compression, and human resources saving.

[0097] For traditional optimization algorithms (such as genetic algorithms), it is easy to fall into local optimality, resulting in sub - optimal layout solutions. And the existing technologies independently adjust the weights and equipment coordinates, with slow convergence speed and low accuracy. The iterative formula of the gradient descent algorithm described in S(4) is:

[0098] ,

[0099] where is the equipment coordinate vector at the -th iteration; is the learning rate, which is dynamically adjusted according to the convergence speed (the value range is from 0.01 to 0.1); is the gradient of the objective function at ; , and are the partial derivatives of the weight parameters with respect to the objective function respectively; used to correct the weight matrix.

[0100] In the above formula, the learning rate ‌: The initial value is set to 0.1. When the change rate of the objective function is < 1% for three consecutive iterations, η automatically drops to 0.05 to refine the search step size and avoid oscillations.

[0101] The gradient term ‌: Reflects the direction of equipment coordinate adjustment. For example, if the coordinate of a certain equipment needs to move 2 meters eastward to shorten the logistics path, the gradient vector points to the east.

[0102] The weight partial derivative term‌: Calculates the influence intensity of the weight parameters on the objective function. For example, if , it indicates that when the α weight increases by 1%, can be increased by 0.8%.

[0103] It is worth mentioning that through the establishment of a joint optimization mechanism, this solution can synchronously update the equipment coordinates and the weight parameters (α, β, γ). Compared with the traditional step-by-step optimization method, the number of convergence iterations is reduced by 40%.

[0104] And by dynamically adjusting the value of η, oscillations are avoided and convergence is accelerated, and the optimization time is shortened by more than 50%.

[0105] For the traditional technical solution, the quality of the initial solution is low. Generally, a random generation method is used for the initial layout, resulting in an increase of more than 50% in the number of subsequent optimization iterations. The path planning adopts a fixed path template and cannot adapt to the dynamic parameters of the AGV. Based on this, the initial layout solution described in S1 is generated in the following way:

[0106] Equipment coordinate allocation‌: Based on the three-dimensional size of the equipment (length × width × height) and safety distance constraints (such as the equipment distance ≥ 0.8 meters), the greedy algorithm is used to allocate coordinates in the production line space in turn according to the equipment priority (such as the equipment with high interaction frequency has a priority level of 1), and the high-frequency interaction equipment (such as the assembly robot) is preferentially placed in the central area to reduce the crossing of the logistics path.

[0107] ‌Logistics path planning‌: The Dijkstra algorithm is used to calculate the shortest feasible path between equipment, and the path width and turning radius are dynamically adjusted in combination with the AGV operation parameters (such as the minimum turning radius of 1.5 meters and the channel width ≥ 2 meters) to generate the initial logistics network.

[0108] Permutation and combination optimization: Generate the initial equipment permutation and combination through a genetic algorithm, use the equipment interaction frequency (such as the number of material handling times per hour) as the fitness function to screen the optimal solution, and eliminate the permutation schemes with a crossover rate higher than 20%.

[0109] Three-dimensional space verification: Pre-verify the space occupancy rate (target ≤ 75%) and channel accessibility (main channel 100% accessible) of the initial layout through a digital twin platform, and eliminate the schemes that do not meet the space constraints.

[0110] It is worth mentioning that this solution defines a specific method for generating the initial layout plan. The initial layout is generated through the collaboration of the greedy algorithm and the genetic algorithm. Compared with the random generation method, the subsequent optimization iteration times are reduced by 40%. By dynamically adjusting the path in combination with the AGV parameters, the AGV channel qualification rate of the initial plan is increased from 70% of the traditional method to 95%.

[0111] For the traditional technical solutions, there is a problem of disconnection between simulation and optimization. The existing digital twin technologies only display three-dimensional models and cannot feedback key data such as equipment utilization rate to the optimization algorithm; moreover, the traditional simulation tools ignore the influence of worker skills and movement trajectories on the layout. Based on this, the digital twin platform described in S3 realizes simulation through the following modules:

[0112] Equipment behavior simulation module: Based on the equipment PLC signals (such as start-stop signals, fault codes) and processing cycle data (such as single-piece processing time of 30 seconds), simulate the start-stop state, failure rate (such as MTBF ≥ 1000 hours) and energy consumption curve (such as peak power of 5kW) of the equipment.

[0113] Logistics path rendering module: Use the A* algorithm to update the material handling path in real time and overlay the dynamic obstacle avoidance strategy (such as the response time for the AGV to re-plan the path after encountering an obstacle ≤ 0.5 seconds).

[0114] Human resource scheduling module: Assign operation tasks according to the worker skill level library (such as senior workers can operate complex equipment), simulate the movement trajectories of workers (such as walking speed of 1.2m / s) and collaborative operation conflicts (such as the conflict probability of two people operating the same work station at the same time).

[0115] Efficiency evaluation module: Real-time statistics of equipment utilization rate (target ≥ 85%), material retention time (target ≤ 10 minutes) and worker idle rate (target ≤ 15%), and generate a simulation report containing KPI indicators.

[0116] It should be noted that the A* algorithm is a heuristic search algorithm, mainly used for the shortest path planning in static road networks. It combines the global optimality of the Dijkstra algorithm and the efficiency of the greedy algorithm, and plays an important role in the field of path planning.

[0117] It is worth mentioning that, through the efficiency evaluation module, this solution can output KPI data in real time, and the adjustment accuracy of the optimization algorithm is improved by 30%. In addition, by integrating equipment, logistics, and human resource simulations, the human conflict rate of the layout plan is reduced from 20% to 5%.

[0118] Traditional technologies only generate a single optimization plan, which cannot meet the needs of multi-scenario decision-making; moreover, the weight adjustment is rigid, only supporting preset weight combinations and lacking user-defined functions.

[0119] Based on this, the implementation methods of generating multiple plans described in S5 include:

[0120] Cost-dominated plan: Fix the α weight at 0.2, increase the β weight to 0.6, and set the γ weight to 0.2. Prioritize reducing equipment procurement costs (such as selecting domestic alternative equipment) and logistics costs (such as shortening the path of high-unit-price AGVs).

[0121] Human resource-saving plan: Fix the γ weight at 0.5, set the α weight to 0.3, and set the β weight to 0.2. Reduce labor costs by reducing the dependence on highly skilled workers (such as using automated equipment to replace) and optimizing the movement paths of workers (such as reducing the path length by 30%).

[0122] Comprehensive balanced plan: Set the α, β, and γ weights to 0.4, 0.3, and 0.3 respectively, and balance the path length (allowing a 10% extension to avoid high-cost areas), cost (allowing a 5% overspend to improve equipment utilization), and human resource requirements (allowing a 10% labor gap to reduce training costs).

[0123] User-defined weight plan: Through an interactive interface (such as a slider), allow users to manually adjust the weight ratio (such as α = 0.5, β = 0.25, γ = 0.25), and generate personalized layouts in real time and visually compare the results.

[0124] It is worth mentioning that: This application can generate at least three typical plans, and the decision-making comparison time is shortened from 2 hours to 20 minutes. It solves the problem that traditional methods only generate a single optimization plan and cannot meet the needs of multi-scenario decision-making. The user-defined weight function improves the configurability of the plan by 80%, especially suitable for small and medium batch customized production lines. It solves the problem that existing technologies only support preset weight combinations and lack user-defined functions.

[0125] For traditional technical solutions, excessive iteration is likely to cause waste of resources. Traditional methods use a fixed number of iterations, but the actual number of iterations does not need to reach the fixed number of iterations, so it will lead to waste of computing resources; moreover, existing technologies only check the changes in equipment coordinates and ignore the stability of weight parameters, which may terminate the optimization prematurely.

[0126] Based on this, the determination conditions for the convergence threshold are:

[0127] ,

[0128] Among them, is a preset error threshold, and its value range is from 0.01 to 0.05; is the path comprehensive optimization index for the t-th iteration.

[0129] In the above formula, the preset error threshold is adjusted accordingly based on different manufacturing sites. For example, for semiconductor production lines, due to high process precision requirements, is set, while for automobile assembly lines, due to a higher error tolerance, it can be relaxed to .

[0130] In the above formula, the normalization method of the comprehensive optimization index is to use linear scaling or logarithmic transformation to eliminate the dimension difference. For example, divide the path length, cost, and human resource items by the reference value (such as the historical optimal value) respectively, so that the weight adjustment of each sub-item is not affected by the magnitude.

[0131] It is worth mentioning that this solution can achieve dynamic convergence determination. It determines convergence in real time according to the change rate. Compared with the fixed number of iterations, the computing resource consumption is reduced by 30% - 50%. And it can achieve multi-parameter joint monitoring.

[0132] For the existing technology that uses two-dimensional coordinate detection and ignores the three-dimensional geometric characteristics of the equipment, the misjudgment rate is as high as 25%. And the traditional method does not integrate the verification of the fire escape route, resulting in risks of layout compliance.

[0133] Based on this, the collision detection module described in S5 verifies the layout feasibility through the following methods:

[0134] Equipment collision detection: Calculate the minimum distance between equipment based on the axis-aligned bounding box (AABB) of the equipment three-dimensional model. If it is less than the safety standard (such as ≥0.8 meters), it is determined that there is a collision and the coordinates need to be adjusted.

[0135] It should be noted that AABB (Axis-Aligned Bounding Box), translated into Chinese as axis-aligned bounding box, is a method of a simplified geometric model widely used in computer graphics, collision detection, and spatial division. Its core purpose is to quickly approximate the spatial range of complex objects through a simple geometric shape (cuboid), thereby improving the calculation efficiency.

[0136] Personnel Safety Area Verification: Calculate the overlap rate between the worker's operation area (radius of 1 meter) and the moving parts of the equipment (such as the working range of the robotic arm). If it exceeds 5%, an alarm will be triggered and the conflict area will be marked.

[0137] Logistics Channel Verification: Detect whether the width of the straight section of the AGV running channel is greater than 1.2 times the width of the AGV body (for example, if the AGV is 1 meter wide, the channel needs to be ≥1.2 meters), and whether the turning radius is greater than the minimum turning radius (such as ≥1.5 meters).

[0138] Emergency Escape Route Inspection: Verify whether the main production line channel is connected and the width is not less than 2 meters. If there are breakpoints or insufficient width, a detour path will be forced to be generated.

[0139] It is worth mentioning that: The three-dimensional detection based on AABB in this solution reduces the misjudgment rate to less than 5%; the verification of the escape route ensures that the layout complies with the fire protection regulations 100%, reducing the risk of production accidents.

[0140] Please refer to Figure 2 , for the problem of scattered system modules in the existing technical solutions. For example, the existing system independently runs modeling, simulation, and optimization modules, and the data interaction delay is up to more than 10 minutes. And the decision-making support is weak. The traditional system lacks a multi-dimensional solution comparison view, and decision-making depends on manual report analysis.

[0141] Based on this, this application provides a system, which is applied to a modular flexible digital twin layout optimization method as described in any one of the above. It is characterized by including:

[0142] Multi-objective Modeling Unit: Used to parse the equipment parameter library, process route library, and human resource library, construct a multi-dimensional constraint model, and output an initial layout plan;

[0143] Dynamic Weight Allocation Unit: Based on the optimization objectives input by the user, adjust 、 、 weight parameters in real time through a slider or a numerical input interface;

[0144] Digital Twin Simulation Engine: Integrate Unity3D or an equivalent 3D engine, support equipment behavior simulation, logistics path rendering, and human resource scheduling;

[0145] Linkage Tuning Controller: Embed the gradient descent algorithm and the genetic algorithm, and iteratively optimize the equipment coordinates and path parameters according to the simulation feedback data;

[0146] Multi-scheme Decision-making Module: Provide a scheme comparison view, support multi-dimensional screening based on cost, efficiency, and safety, and visualize the collision detection results.

[0147] Adopting the above technical solution: The present invention can achieve real-time interaction among the modeling, simulation, and optimization modules, and reduce the data delay from 10 minutes to within 5 seconds. The solution comparison view improves the decision-making efficiency by 70%, and is especially suitable for the rapid reconstruction of production lines with multiple batches.

[0148] It is worth mentioning that this solution can perform full-process integration, with real-time interaction among the modeling, simulation, and optimization modules, and reduce the data delay from 10 minutes to within 5 seconds. Moreover, it has an intelligent decision-making function. The solution comparison view improves the decision-making efficiency by 70%, and is especially suitable for the rapid reconstruction of production lines with multiple batches.

[0149] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A modular flexible digital twin layout optimization method, characterized in that: Including: S1: Build a multi-dimensional constraint model, build a multi-objective optimization model including route length, equipment cost, and the number of human resources, and generate an initial layout plan based on the three-dimensional geometric attributes of production line equipment and the logistics path; S2: Perform dynamic weight assignment, and adjust the route optimization weight, cost optimization weight, and human resource weight in real time according to the priority parameters input by the user to generate a dynamic weight combination matrix. and cost optimization weight and human resource weight to generate a dynamic weight combination matrix. S3: Conduct digital twin simulation, map the initial layout plan to the digital twin platform, and calculate the actual equipment utilization rate, material handling efficiency, and worker operation path through logistics path simulation; S4: Real-time linkage optimization, based on the simulation results, use the gradient descent algorithm to iteratively optimize the equipment coordinates and logistics path, and update the dynamic weight combination matrix until the convergence threshold is reached; S5: Generate and verify multiple plans, generate at least three optimization plans by adjusting the dynamic weight combination matrix, and verify the layout feasibility based on the collision detection module of the digital twin platform.

2. A modular flexible digital twin layout optimization method according to claim 1, characterized in that: The total cost function of the multi-objective optimization model described in S1 is: ， in, is the total cost; For the The purchase cost of the equipment; For the Unit distance cost of the logistics route; For the The actual length of the logistics path; For the Average annual maintenance cost of this type of equipment; is the total number of devices; The number of device types.

3. The modular flexible digital twin layout optimization method according to claim 1, wherein, The path optimization function of the dynamic weight combination matrix described in S2 is: ， Among them, is the path comprehensive optimization index; is the actual length of the k-th path; is the theoretical shortest length of the k-th path; is the current total cost; is the benchmark cost; is the required number of workers, is the available number of workers; is the total number of logistics paths.

4. A modular flexible digital twin layout optimization method according to claim 1, characterized in that, The iterative formula of the gradient descent algorithm described in S4 is: ， Among them, is the device coordinate vector for the th iteration; is the learning rate, with a value range of 0.01 to 0.1; is the gradient of the objective function at ; , and are the partial derivatives of the weight parameters with respect to the objective function, respectively.

5. A modular flexible digital twin layout optimization method according to claim 1, characterized in that: The initial layout plan described in S1 is generated by the following method: S101: Equipment coordinate allocation: Based on the three-dimensional size of the equipment and the safety distance constraint, use the greedy algorithm to sequentially allocate coordinates in the production line space according to the equipment priority, and give priority to placing high-frequency interaction equipment in the central area; S102: Logistics path planning: Use the Dijkstra algorithm to calculate the shortest feasible path between equipment, and dynamically adjust the path width and turning radius in combination with the AGV operation parameters; S103: Permutation and combination optimization: Generate an initial equipment permutation and combination through the genetic algorithm, and select the optimal solution with the equipment interaction frequency as the fitness function; S104: Three-dimensional space verification: Pre-verify the space occupancy rate and channel accessibility of the initial layout through the digital twin platform.

6. The modular flexible digital twin layout optimization method according to claim 1, wherein The digital twin platform described in S3 realizes simulation through the following modules: Equipment behavior simulation module: Based on the equipment PLC signal and processing cycle data, simulate the start-stop state, failure rate, and energy consumption curve of the equipment; Logistics path rendering module: Use the A* algorithm to update the material handling path in real time and overlay the dynamic obstacle avoidance strategy; Human resource scheduling module: Allocate operation tasks according to the worker skill level library, and simulate the worker movement trajectory and collaborative operation conflicts; Efficiency evaluation module: Real-time statistics of equipment utilization rate, material retention time, and worker idle rate, and generate a simulation report.

7. A modular flexible digital twin layout optimization method according to claim 1, characterized in that The implementation method of generating multiple plans described in S5 includes: Cost-driven solution: Fixed The weight is 0.2, The weight is increased to 0.6, The weight is set to 0.2 to reduce equipment procurement and logistics costs; Human resource saving plan: fixed The weight is 0.5, and the weight is set to 0.3, the weight is set to 0.2, giving priority to reducing the number of workers and skill dependencies; Comprehensive and balanced solution: 、 、 The weights are set to 0.4, 0.3, and 0.3, respectively, to balance path length, cost, and human resource requirements; Custom weight plan: Allow users to manually adjust the weight ratio through the interactive interface and generate personalized layouts in real time.

8. A modular flexible digital twin layout optimization method according to claim 1, characterized in that, The determination condition of the convergence threshold is: ， in, is the preset error threshold, ranging from 0.01 to 0.05; is the comprehensive optimization index of the path at the tth iteration.

9. A modular flexible digital twin layout optimization method according to claim 1, characterized in that The collision detection module described in S5 verifies the layout feasibility through the following methods: Equipment collision detection: Based on the axial bounding box of the equipment three-dimensional model, detect the minimum distance between equipment to ensure compliance with safety operation standards; Personnel safety area verification: Calculate the safety distance between the worker operation area and the moving parts of the equipment, and if the overlap rate exceeds 5%, it is determined as a conflict; Logistics channel verification: Detect whether the straight section width of the AGV running channel is greater than 1.2 times the AGV vehicle body width, and whether the turning radius is greater than the minimum turning radius; Emergency escape route inspection: Verify that the main production line passage is connected and the width is not less than 2 meters, meeting the requirements of fire protection codes.

10. A modular flexible digital twin layout optimization system, which is applied to a modular flexible digital twin layout optimization method according to any one of claims 1-9, characterized in that, Including: Multi-objective modeling unit: Used to analyze the equipment parameter library, process route library, and human resource library, construct a multi-dimensional constraint model, and output an initial layout plan; Dynamic weight allocation unit: Based on the optimization goal input by the user, adjust the , , weight parameters in real time through a slider or a numerical input interface; Digital twin simulation engine: Integrates Unity3D or an equivalent 3D engine, supporting equipment behavior simulation, logistics path rendering, and human resource scheduling; Linkage optimization controller: Embeds gradient descent algorithm and genetic algorithm, and iteratively optimizes equipment coordinates and path parameters according to simulation feedback data; Multi-scheme decision-making module: Provides a view for comparing schemes, supporting multi-dimensional screening based on cost, efficiency, and safety, and visualizing the collision detection results.

Images