A Real-Time Scheduling Method and System for Flexible Sheet Metal Production Lines Based on Digital Twins

Abstract

This invention discloses a real-time scheduling method and system for a flexible sheet metal production line based on digital twins, belonging to the field of production scheduling and optimization technology. The method includes calculating the deviation between the predicted and actual values ​​of the digital twin based on production data; dynamically correcting the mapping relationship between equipment processing efficiency and output quality in the digital twin based on the deviation; and performing scheduling deduction based on the corrected digital twin to generate a scheduling scheme, which integrates the corrected equipment processing efficiency and output quality constraints. This invention, based on the corrected digital twin, employs a multi-objective optimization algorithm with on-time delivery rate, overall equipment efficiency, energy consumption, quality risk, and raw material overconsumption as optimization objectives for scheduling deduction. This enables the generated scheduling scheme to take into account multiple dimensions such as production efficiency, resource utilization, cost control, and quality assurance, thereby improving the overall production efficiency of the flexible sheet metal production line.

Description

Technical Field

[0001] This invention relates to the field of production scheduling and optimization technology, and in particular to a real-time scheduling method and system for flexible sheet metal production lines based on digital twins. Background Technology

[0002] With the advancement of industrial development and the iteration of intelligent manufacturing, the sheet metal manufacturing industry is shifting from large-scale standardized production to flexible production of multiple varieties and small batches. Downstream customers have higher requirements for delivery cycles, processing accuracy, and quality stability. Digital twin technology, relying on precise mapping, real-time interaction, and virtual simulation capabilities, provides a new path to solve the dynamic and complex problems of flexible production line scheduling. It urgently needs to be deeply integrated into the real-time scheduling of sheet metal production lines to adapt to complex production needs.

[0003] Current sheet metal flexible production line scheduling technology has gone through three stages: traditional experience-based scheduling relies on manual judgment, which is inefficient, easily affected by human factors, and difficult to adapt to dynamic changes in orders; static algorithm scheduling, although optimized by mathematical models, is based on preset equipment parameters and order information, without considering equipment health fluctuations and dynamic changes in quality, and lacks real-time performance; preliminary intelligent scheduling, although introducing digital twins, mostly only maps equipment status, without building a multi-dimensional dynamic model of efficiency and quality, and lacks multi-objective collaborative optimization, resulting in limited practicality of scheduling solutions.

[0004] Key issues remain with existing technologies: First, there is insufficient coordination between the scheduling model and physical entities. Traditional methods and some digital twins lack dynamic correction mechanisms, leading to increased discrepancies between predicted and actual values ​​due to equipment anomalies, and scheduling schemes becoming disconnected from reality. Second, there is insufficient integration of equipment health data. Existing technologies mostly consider only operating parameters and do not deeply integrate multi-sensor health data with scheduling, making it difficult to accurately predict efficiency and quality risks. Third, multi-objective optimization is incomplete. Most solutions focus only on on-time delivery or equipment utilization, neglecting energy consumption, excessive raw material consumption, and quality risks, making it difficult to meet the demands of efficient, low-carbon, and high-quality production, thus hindering the intelligent development of sheet metal production lines.

[0005] Therefore, it is essential to invent a real-time scheduling method and system for flexible sheet metal production lines based on digital twins to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a real-time scheduling method and system for flexible sheet metal production lines based on digital twins, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a real-time scheduling method for a flexible sheet metal production line based on digital twins, specifically including the following steps:

[0008] S1. Construct a digital twin of the production line, wherein the digital twin at least maps the processing efficiency and output quality of the equipment;

[0009] The digital twin maps the processing efficiency and output quality of the equipment using an efficiency mapping model and a quality mapping model. The efficiency mapping model establishes a mapping relationship between equipment health status data and processing efficiency coefficients, while the quality mapping model establishes a mapping relationship between equipment health status data, order attributes, and predicted pass rate.

[0010] S2. Real-time collection of production data, which includes at least the actual operating parameters of the equipment and product quality data;

[0011] The actual operating parameters include equipment health status data and actual processing time, and the product quality data includes real-time pass rate;

[0012] S3. Based on the production data, calculate the deviation between the predicted value and the actual value of the digital twin;

[0013] The deviation includes efficiency deviation and quality deviation; the efficiency deviation is calculated by the ratio of actual processing time to theoretical processing time, and the quality deviation is calculated by the ratio of real-time pass rate to predicted pass rate.

[0014] S4. Based on the deviation, dynamically correct the mapping relationship between equipment processing efficiency and output quality in the digital twin;

[0015] The dynamic correction includes: updating the current efficiency coefficient in the digital twin using the reciprocal of the efficiency deviation, weighting the predicted pass rate using the quality deviation, and updating the current predicted pass rate in the digital twin.

[0016] S5. Based on the corrected digital twin, perform scheduling simulation to generate a scheduling scheme, which integrates the corrected equipment processing efficiency and output quality constraints.

[0017] The scheduling simulation adopts a multi-objective optimization algorithm, and the optimization objectives include on-time delivery rate, overall equipment efficiency, energy consumption, quality risk and raw material overconsumption;

[0018] The multi-objective optimization algorithm is a reinforcement learning algorithm. Its state space includes order queue, equipment state, efficiency coefficient matrix, pass rate matrix and material state. The reward function is a weighted combination of multiple objectives.

[0019] S6. Execute the scheduling scheme and return the execution data to continuously correct the digital twin.

[0020] Preferably, the device health status data is calculated based on the fusion of data collected by one or more of the vibration sensor, laser power sensor, and servo current sensor.

[0021] A real-time scheduling system for flexible sheet metal production lines based on digital twins includes:

[0022] The perception layer is used to collect production data in real time.

[0023] The transport layer is used for data transmission;

[0024] The digital twin and decision-making layer, including a digital twin engine and a scheduling optimization module, are used to build, run, and modify the digital twin, and to execute the scheduling simulation to generate a scheduling scheme.

[0025] The execution layer includes sheet metal processing equipment for executing the scheduling scheme.

[0026] Preferably, the sensing layer includes a timing sensing unit for collecting processing time, a performance sensing unit for collecting equipment health status data, and a quality sensing unit for collecting product quality data.

[0027] The technical effects and advantages of this invention are as follows:

[0028] 1. This invention constructs a digital twin that maps the processing efficiency and output quality of a device, and calculates the device health status data based on data collected by one or more of vibration sensors, laser power sensors, and servo current sensors. This enables the digital twin to accurately map the physical device status from the dimensions of efficiency and quality, providing a reliable virtual model foundation for subsequent scheduling simulations and avoiding scheduling decision deviations caused by the disconnect between the virtual model and the physical device.

[0029] 2. This invention collects actual equipment operating parameters and product quality data in real time, calculates the efficiency deviation and quality deviation between the predicted and actual values ​​of the digital twin, and uses the deviation to dynamically correct the mapping relationship between equipment processing efficiency and output quality in the digital twin. This allows the digital twin to continuously match the actual operating conditions of the physical production line, ensuring that the scheduling plan is always generated based on the latest equipment status and quality constraints, and avoiding the scheduling plan from deviating from actual production needs.

[0030] 3. This invention uses a multi-objective optimization algorithm based on a modified digital twin to perform scheduling simulation, with on-time delivery rate, overall equipment efficiency, energy consumption, quality risk and raw material overconsumption as optimization objectives. This enables the generated scheduling scheme to take into account multiple dimensions such as production efficiency, resource utilization, cost control and quality assurance, thereby improving the overall production efficiency of the sheet metal flexible production line.

[0031] 4. This invention establishes an efficiency mapping model and a quality mapping model in the digital twin, respectively, to establish the mapping relationship between equipment health status data and processing efficiency coefficient, and between equipment health status data and order attributes and predicted pass rate. This enables accurate prediction of the processing efficiency and output quality of different equipment when handling different orders, allowing the scheduling scheme to avoid quality risks at the task allocation stage, reduce the generation of defective products, and reduce raw material waste and rework costs.

[0032] 5. This invention constructs a layered system architecture that includes a perception layer, a transmission layer, a digital twin and decision-making layer, and an execution layer. This enables a collaborative closed loop of real-time production data acquisition, reliable transmission, virtual decision-making, and physical execution, ensuring efficient connection of the entire process from generation to execution of the scheduling scheme. At the same time, it provides data flow assurance for the continuous correction of the digital twin, improving the stability and feasibility of the scheduling system.

[0033] 6. This invention employs reinforcement learning as a multi-objective optimization algorithm, incorporating order queues, equipment status, efficiency coefficient matrix, pass rate matrix, and material status into the state space. It uses a weighted combination of multiple objectives as the reward function, enabling scheduling simulation to continuously learn the optimal scheduling strategy in a virtual environment. This allows the system to adapt to dynamic situations such as order changes and equipment status fluctuations during the production process, thereby improving the flexibility and adaptability of the scheduling scheme. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0035] Figure 2 This is a schematic diagram of the system architecture of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] This invention provides, for example Figure 1 The real-time scheduling method for a flexible sheet metal production line based on digital twins, as shown, specifically includes the following steps:

[0038] S1. Construct a digital twin of the production line, wherein the digital twin at least maps the processing efficiency and output quality of the equipment;

[0039] Furthermore, in the above technical solution, the digital twin maps the processing efficiency and output quality of the equipment through an efficiency mapping model and a quality mapping model; the efficiency mapping model establishes a mapping relationship between equipment health status data and processing efficiency coefficients, and the quality mapping model establishes a mapping relationship between equipment health status data, order attributes, and predicted pass rate.

[0040] Furthermore, in the above technical solution, the device health status data is calculated based on the fusion of data collected by one or more of the vibration sensor, laser power sensor, and servo current sensor.

[0041] It is important to understand that the efficiency mapping model is used to establish a mapping relationship between equipment health status data and processing efficiency coefficients. Its purpose is to quantify comprehensive equipment health status indicators into a specific efficiency coefficient that can be used for scheduling calculations, thereby providing efficiency-level data constraints for scheduling simulations. The relationship established by the efficiency mapping model can be expressed as: E = f(H). In one embodiment, the relationship is linear: E = α*H + β; where E is the processing efficiency coefficient, a dimensionless scaling factor; H is the equipment health status value; and α and β are coefficients pre-set based on equipment type and historical process data.

[0042] The quality mapping model is used to establish a mapping relationship between equipment health status data, order attributes, and predicted pass rates. Its purpose is to pre-assess the potential quality outcomes when executing different orders on different equipment before scheduling, thereby providing data constraints at the quality risk level for scheduling decisions. The relationship established by the quality mapping model can be expressed as: Q the =g(H, O). In one embodiment, the order attribute O includes the material type T and thickness D of the sheet material to be processed, and the relationship can be further expressed as: Q the =γ*H+λ*C(T,D); among them, Q the The function is used to predict the pass rate; C(T, D) is a benchmark pass rate query function built based on historical processing data, reflecting the inherent processing difficulty of the order; γ and λ are weighting coefficients used to balance the influence of equipment health status data and order attributes on the predicted pass rate. The weighting coefficients γ and λ are determined by optimization based on historical production data, or preset based on domain knowledge.

[0043] The equipment health status value H is a comprehensive evaluation value, calculated based on the fusion of data collected from one or more of the vibration sensor, laser power sensor, and servo current sensor. Its purpose is to provide a unified and reliable health input for the aforementioned mapping model. In one embodiment, the equipment health status value H is calculated as follows: real-time data from the vibration sensor, laser power sensor, and servo current sensor are collected to obtain equipment health status data; the vibration sensor data undergoes frequency domain transformation and the amplitude of the main frequency bands is extracted to calculate the vibration health score Sv; the ratio of the actual power collected by the laser power sensor to the rated power is calculated to obtain the power health score Sp; the fluctuation variance of the current data collected by the servo current sensor is calculated to obtain the current health score Sc; finally, a weighted fusion algorithm is used to synthesize the above scores to calculate the equipment health status value H, for example: H=(ω1*Sv+ω2*Sp+ω3*Sc) / (ω1+ω2+ω3); where ω1, ω2, and ω3 are pre-set weighting factors reflecting the degree of influence of each sensor data on equipment health.

[0044] S2. Real-time collection of production data, which includes at least the actual operating parameters of the equipment and product quality data;

[0045] Furthermore, in the above technical solution, the actual operating parameters include equipment health status data and actual processing time, and the product quality data includes real-time pass rate.

[0046] It is important to know that the actual processing time can be collected through the timing function of the equipment controller, such as a PLC or CNC system, or through a timing module embedded in the host computer system. Specifically, when the equipment controller receives a processing start signal, the system records a start timestamp; when it receives a processing completion signal, the system records an end timestamp; the difference between the two is the actual processing time of the workpiece.

[0047] The product quality data refers to the measured or statistical values ​​obtained after inspecting the finished products, which determine whether the products meet quality standards. These data are compared with the predicted values ​​in the digital twin to calculate the quality deviation, serving as a crucial basis for subsequent adjustments to the quality mapping model. The product quality data includes the real-time pass rate, a parameter that can be collected through online inspection systems such as machine vision inspection units or laser measuring instruments, or through offline quality inspection station data entry terminals such as MES system terminals. The system records the inspection results of each workpiece or batch in real-time or near real-time, and calculates and outputs the current real-time pass rate within a set statistical period, such as every half hour or after each order is completed. The calculation formula is the number of qualified workpieces divided by the total number of inspected workpieces.

[0048] S3. Based on the production data, calculate the deviation between the predicted value and the actual value of the digital twin;

[0049] Furthermore, in the above technical solution, the deviation specifically includes the efficiency deviation η. eff and quality deviation η qua The efficiency deviation η eff Based on the actual processing time T of the equipment act Theoretical processing time T for digital twin prediction the The ratio is calculated, and its formula is: η eff =T act / T the The quality deviation η qua Real-time pass rate Q collected from the production line act The prediction pass rate Q of digital twin prediction the The ratio is calculated, and its formula is: η qua =Q act / Q the .

[0050] It is important to know that the theoretical processing time T the The efficiency mapping model derived from the digital twin maps an efficiency coefficient representing processing capacity based on the current health status of the equipment. The theoretical processing time is then calculated from this efficiency coefficient and the baseline processing time of the equipment under ideal health conditions. The calculation formula is: Theoretical processing time = Baseline processing time / Efficiency coefficient. The baseline processing time is based on standard parameters provided by the equipment manufacturer or initialized through learning from historical process data.

[0051] The real-time pass rate Q act The ratio of qualified products in the current production batch to the total output is obtained in real time by quality sensing units deployed in the quality inspection process, such as machine vision inspection systems and laser measuring instruments.

[0052] By calculating efficiency deviation and quality deviation, the degree of difference between the digital twin prediction model and the actual operating state of the physical entity can be accurately quantified. Efficiency deviation directly reflects the deviation between the actual processing efficiency and the expected efficiency of the equipment; a ratio greater than 1 indicates that the actual processing efficiency is lower than the predicted level, while a ratio less than 1 indicates that the actual efficiency is better than the prediction. Quality deviation quantifies the deviation between the production output quality and the expected quality level, and is a key indicator for evaluating the stability of the production process and the accuracy of the prediction model. These two deviations provide a clear and quantitative data foundation for subsequent dynamic correction of the digital twin, ensuring that the model on which scheduling decisions are based always remains synchronized with production reality, thereby improving the accuracy and reliability of the entire scheduling method.

[0053] S4. Based on the deviation, dynamically correct the mapping relationship between equipment processing efficiency and output quality in the digital twin;

[0054] Furthermore, in the above technical solution, the dynamic correction includes: updating the current efficiency coefficient in the digital twin using the reciprocal of the efficiency deviation, weighting the predicted pass rate using the quality deviation, and updating the current predicted pass rate in the digital twin.

[0055] It is important to understand that the dynamic correction is the core element to ensure that the digital twin can continuously and accurately reflect the real state of the physical entity. Its purpose is to use the deviation between the real-time collected production data and the predicted value of the digital twin to reverse-calibrate and update the internal mapping model of the digital twin, namely the efficiency mapping model and the quality mapping model, thereby reducing the difference between the digital twin and the physical entity and providing a more accurate model basis for subsequent scheduling decisions.

[0056] Specifically, the dynamic correction includes the following two parallel update processes: first, utilizing the efficiency deviation η eff The reciprocal updates the current efficiency coefficient E in the digital twin. cur Its calculation formula is , of which E updated This is the updated efficiency coefficient; when the actual processing time is greater than the theoretical time (η) eff When the value is greater than 1, it indicates that the actual processing efficiency of the equipment is lower than the model's predicted value. Therefore, it is necessary to multiply the efficiency by a factor less than 1 (i.e., 1 / η). eff The efficiency coefficient E is lowered to make the model's predicted efficiency more consistent with the actual slowdown; conversely, when the actual processing time is less than the theoretical time (η), the efficiency coefficient E is lowered. eff If <1), then the efficiency coefficient is increased to reflect the fact that the actual performance of the equipment is better than expected; secondly, the quality deviation η is used. qua The predicted pass rate Q the A weighted correction is performed to update the current prediction pass rate in the digital twin, and the calculation formula is as follows: , where a is a preset learning rate factor with a value between 0 and 1, used to control the magnitude and speed of model updates;

[0057] S5. Based on the corrected digital twin, perform scheduling simulation to generate a scheduling scheme, which integrates the corrected equipment processing efficiency and output quality constraints.

[0058] Furthermore, in the above technical solution, the scheduling simulation adopts a multi-objective optimization algorithm, and the optimization objectives include on-time delivery rate, overall equipment efficiency, energy consumption, quality risk, and raw material overconsumption.

[0059] It should be noted that the aforementioned scheduling simulation refers to the simulation and optimization calculation of current and future production tasks in a digital twin environment using a digital twin model that has been corrected by real-time data. Its purpose is to generate a feasible and efficient set of scheduling instructions that can simultaneously take into account multiple production objectives.

[0060] The scheduling scheme specifies which equipment, at what time, and in what order each production order will be processed, and fully incorporates the equipment processing efficiency coefficient and predicted pass rate constraints after dynamic correction in step S4, thereby ensuring the accuracy and reliability of scheduling decisions.

[0061] The on-time delivery rate is used to evaluate the ability of the scheduling scheme to meet the order delivery time, and its calculation formula is the ratio of the number of orders completed on time to the total number of orders.

[0062] The overall equipment efficiency is used to evaluate the combined performance of equipment time utilization, performance utilization, and pass rate. The calculation formula is: Overall Equipment Efficiency = Time Utilization × Performance Utilization × Pass Rate.

[0063] The energy consumption is used to evaluate the total energy consumption in the production process. It is calculated based on equipment power, processing time and standby energy consumption, and the optimization objective is to minimize the total energy consumption.

[0064] The quality risk is used to assess the product quality uncertainty that the scheduling scheme may bring. It is calculated based on the modified predicted pass rate in the digital twin, for example, it can be defined as a weighted sum of (1-predicted pass rate), and the optimization objective is to minimize the total quality risk.

[0065] The raw material overconsumption is used to assess the raw material consumption exceeding the standard quota due to fluctuations in processing quality (such as scrap and rework). It is negatively correlated with the predicted pass rate, and the optimization objective is to minimize the raw material overconsumption.

[0066] S6. Execute the scheduling scheme and return the execution data to continuously correct the digital twin.

[0067] Furthermore, in the above technical solution, the multi-objective optimization algorithm is a reinforcement learning algorithm, whose state space includes order queue, equipment state, efficiency coefficient matrix, pass rate matrix and material state, and the reward function is a weighted combination of multiple objectives.

[0068] It is important to understand that the core of using reinforcement learning algorithms for scheduling simulation lies in learning the optimal scheduling strategy through continuous interaction with the digital twin environment.

[0069] The state space is used to describe the complete state of the production system at a certain moment and serves as the basis for the reinforcement learning agent to make decisions.

[0070] The order queue contains attributes of all pending orders, such as order ID, delivery date, process route, standard processing time, etc.

[0071] The device status indicates the real-time status of each device, such as idle, busy, faulty, under maintenance, etc.

[0072] The efficiency coefficient matrix is ​​an m×n matrix, with elements E ij This represents the current efficiency coefficient of device i in processing order type j, where m is the number of devices and n is the order type;

[0073] The pass rate matrix is ​​an m×n matrix, with element Q. ij This represents the current predicted pass rate for equipment i processing order type j;

[0074] The material status refers to the inventory status and location information of raw materials, work-in-process, and finished products;

[0075] The reward function is the direct objective of reinforcement learning algorithm optimization. It quantifies the immediate reward obtained by performing a scheduling action, such as assigning an order to a device, under a specific state. The calculation formula is: Reward = w1 × R OTD +w2×R OEE -w3×R E -w4×R QR -w5×R MO , where R OTD To ensure timely delivery of the reward portion, R OEE R is a bonus component for overall equipment efficiency. E As the energy consumption penalty component, R QR As a quality risk penalty component, R MO The excess raw material consumption penalty is represented by w1, w2, w3, w4, and w5, which are pre-set positive weighting coefficients for each objective, used to adjust the importance of different optimization objectives in the overall reward.

[0076] It is important to know that the on-time delivery reward component R OTD This incentive scheduling scheme prioritizes on-time order completion, and its calculation is based on the difference between the order's completion time and delivery date. For a set of orders completed within a scheduling cycle, its value can be calculated as follows:

[0077] ,

[0078] Where K is the total number of orders completed within this period, and C k Let D be the completion time of order k. k This represents the delivery date for order k; the larger this value, the better the on-time delivery performance.

[0079] The equipment's overall efficiency bonus component R OEE This incentivizes scheduling schemes to improve equipment utilization and efficiency. Its calculation is based on the classic equipment comprehensive efficiency framework, which is the product of time uptime, performance uptime, and product qualification rate. For multiple devices within a scheduling cycle, its value can be calculated as follows:

[0080] ,

[0081] Where M is the total number of devices, A m The operating rate is calculated using the formula A. m = Actual operating time of equipment m / Planned load time, P m The performance utilization rate is calculated using the formula P. m = (Ideal cycle time × Number of processes) / Actual operation time, Q m The pass rate is calculated using the formula Q. m = Quantity of qualified products / Total number of products processed; the larger this value, the higher the overall utilization efficiency of the equipment.

[0082] The energy consumption penalty component R E This is used to penalize the energy consumption caused by the scheduling scheme. The optimization objective is to minimize energy consumption, and its calculation is based on the integral of the power and operating status of each device. For a scheduling cycle, its value can be calculated as:

[0083] ,

[0084] in, and These are the processing power and standby power of equipment m, respectively. and Let be the cumulative processing time and standby time of device m during this period, respectively. This component value directly represents the total energy consumption value and serves as a penalty term in the reward function.

[0085] The quality risk penalty component R QR This is used to penalize product quality uncertainties that may be caused by the scheduling scheme. The optimization objective is to minimize quality risk, and its calculation is directly related to the modified predicted yield rate provided by the digital twin. For all processes scheduled within a scheduling cycle, its value can be calculated as follows:

[0086] ,

[0087] Where J represents the total number of processes planned for processing within this cycle. V is the predicted pass rate for this process using a digital twin. j This represents the value or priority weight of the order corresponding to this process. The larger this value, the higher the overall quality risk, and it serves as a penalty in the reward function.

[0088] The raw material overconsumption penalty component R MO Used to penalize additional raw material consumption caused by defective products and rework, its calculation is negatively correlated with the predicted pass rate. For all processing steps scheduled within a scheduling cycle, its value can be calculated as follows:

[0089] ,

[0090] Where J is the total number of processes planned for processing within this cycle, and M... j This represents the standard rated material consumption for this process. The larger this value, the more severe the estimated material overconsumption, which is used as a penalty in the reward function.

[0091] A real-time scheduling system for flexible sheet metal production lines based on digital twins, such as Figure 2 As shown, it includes:

[0092] The perception layer is used to collect production data in real time.

[0093] The transport layer is used for data transmission;

[0094] The digital twin and decision-making layer, including a digital twin engine and a scheduling optimization module, are used to build, run, and modify the digital twin, and to execute the scheduling simulation to generate a scheduling scheme.

[0095] The execution layer includes sheet metal processing equipment for executing the scheduling scheme.

[0096] It is important to know that the system adopts a layered architecture design, the purpose of which is to clarify the responsibilities of each layer, achieve decoupling and coordination of data acquisition, transmission, processing and execution, and thus build a closed-loop real-time scheduling and optimization system.

[0097] Furthermore, in the above technical solution, the sensing layer includes a timing sensing unit for collecting processing time, a performance sensing unit for collecting equipment health status data, and a quality sensing unit for collecting product quality data.

[0098] It is important to understand that the perception layer is the data source for the system's interaction with the physical production line. Its core responsibility is to comprehensively and in real-time acquire the status information of physical entities through sensors and data acquisition devices deployed at various stages of the production line. The perception layer specifically includes:

[0099] Time-series sensing unit: Used to collect production process data related to time series. This unit accurately records the start and end timestamps of each processing task through a timing module integrated in the equipment controller (such as PLC or CNC system) or an external high-precision timer, thereby calculating the actual processing time. In addition, this unit can also collect time-series information such as equipment start / stop status and order queue waiting time, providing a data basis for efficiency evaluation.

[0100] Performance Sensing Unit: This unit collects equipment health status data. It utilizes various sensors, such as vibration sensors (installed on critical moving parts like the spindle to monitor vibration amplitude and frequency), laser power sensors (integrated into the optical path of the laser cutting equipment to monitor the stability of laser output power in real time), and servo current sensors (installed in the servo drive unit to monitor motor operating current and fluctuations). The data from these sensors is fused and calculated to assess the equipment's health status, providing input for the efficiency and quality mapping models in the digital twin.

[0101] Quality sensing unit: Used to collect data related to product quality after or during product processing. This unit may include online inspection systems (such as machine vision inspection units to measure hole positions and contour dimensions of processed sheet metal parts in real time, or laser measuring instruments to scan surface flatness) and offline quality inspection station data entry terminals (such as MES system terminals, where quality inspectors enter inspection results). The data collected by this unit is used to calculate product quality indicators such as real-time pass rate, providing a basis for calculating quality deviation.

[0102] The transmission layer serves as a bridge connecting the perception layer and the upper decision-making layer. Its core responsibility is to ensure that the collected production data can be reliably, efficiently, and securely transmitted to the digital twin and decision-making layers. The transmission layer can employ high-speed communication networks such as industrial Ethernet, 5G, and Wi-Fi 6, and adhere to industrial communication protocols such as OPC UA and MQTT to achieve unified access and transmission of heterogeneous data.

[0103] The digital twin and decision-making layer are the core of the system, and their functions include:

[0104] The digital twin engine is responsible for constructing and dynamically updating a virtual mapping (i.e., a digital twin) of the production line based on data uploaded from the perception layer. It has built-in efficiency and quality mapping models, enabling it to predict the processing efficiency coefficient and output quality (predicted pass rate) of equipment based on equipment health status data. Simultaneously, it receives actual production data (actual processing time, real-time pass rate), calculates efficiency and quality deviations, and dynamically adjusts the model parameters in the digital twin accordingly, ensuring that the virtual model and the physical entity remain synchronized.

[0105] The scheduling optimization module is responsible for performing scheduling simulations based on the latest revised digital twin. This module runs multi-objective optimization algorithms (such as reinforcement learning algorithms) to optimize various objectives, including on-time delivery rate, overall equipment efficiency, energy consumption, quality risk, and raw material overconsumption. It simulates multiple scheduling strategies in a virtual environment and ultimately generates the optimal scheduling scheme. This scheme fully integrates the revised equipment processing efficiency and output quality constraints.

[0106] The execution layer is the final executor of the scheduling instructions. Its core includes various sheet metal processing equipment on the production line (such as laser cutting machines, CNC punching machines, bending machines, etc.) and their equipment controllers. This layer receives scheduling schemes from the digital twin and decision-making layer and converts them into instructions and work order sequences that the equipment can recognize, driving the physical equipment to perform specific processing tasks, thereby completing the closed loop from virtual decision-making to physical execution.

[0107] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time scheduling of a sheet metal flexible production line based on digital twinning, characterized in that, Specifically comprising the following steps: S1, constructing a digital twin of the production line, the digital twin at least mapping the processing efficiency and output quality of the equipment; The digital twin maps the processing efficiency and output quality of the equipment through an efficiency mapping model and a quality mapping model; the efficiency mapping model establishes a mapping relationship between the equipment health state data and the processing efficiency coefficient, and the quality mapping model establishes a mapping relationship between the equipment health state data, order attributes and predicted pass rate; S2, collecting production data in real time, the production data at least including actual running parameters of the equipment and product quality data; The actual running parameters include equipment health state data and actual processing time, and the product quality data includes real-time pass rate; S3, calculating the deviation between the predicted value and the actual value of the digital twin according to the production data; The deviation includes efficiency deviation degree and quality deviation degree; the efficiency deviation degree is calculated by the ratio of actual processing time to theoretical processing time, and the quality deviation degree is calculated by the ratio of real-time pass rate to predicted pass rate; S4, dynamically correcting the mapping relationship between the processing efficiency and the output quality of the equipment in the digital twin based on the deviation; The dynamic correction includes: updating the current efficiency coefficient in the digital twin by using the reciprocal of the efficiency deviation degree, and weighting correcting the predicted pass rate by using the quality deviation degree to update the current predicted pass rate in the digital twin; S5, generating a scheduling scheme based on the corrected digital twin, the scheduling scheme integrating the corrected processing efficiency and output quality constraints of the equipment; The scheduling deduction adopts a multi-objective optimization algorithm, and the optimization objectives include on-time delivery rate, comprehensive efficiency of the equipment, energy consumption, quality risk and raw material overconsumption; The multi-objective optimization algorithm is a reinforcement learning algorithm, and its state space includes order queue, equipment state, efficiency coefficient matrix, pass rate matrix and material state, and the reward function is a weighted combination of multiple objectives; S6, executing the scheduling scheme and returning execution data to continuously correct the digital twin.

2. The digital-twin-based real-time scheduling method for sheet metal flexible production lines according to claim 1, characterized in that, The equipment health state data is calculated based on data fusion collected by one or more of a vibration sensor, a laser power sensor and a servo current sensor.

3. The sheet metal flexible production line real-time scheduling system based on digital twinning, applied to the sheet metal flexible production line real-time scheduling method based on digital twinning of any one of claims 1-2, characterized in that, It comprises: a perception layer for collecting production data in real time; a transmission layer for data transmission; a digital twin and decision-making layer including a digital twin engine and a scheduling optimization module, for constructing, running and correcting the digital twin, and executing the scheduling deduction to generate a scheduling scheme; an execution layer including sheet metal processing equipment for executing the scheduling scheme.

4. The digital-twin-based sheet metal flexible production line real-time scheduling system according to claim 3, characterized in that, The perception layer includes a timing perception unit for collecting processing time, a performance perception unit for collecting equipment health state data, and a quality perception unit for collecting product quality data.

Images