Digital Twin and Dynamic Intelligent Scheduling Method and System for Rebar Processing

By constructing equipment-level digital twin models and multi-objective intelligent scheduling, the problems of neglecting equipment changeover time constraints and insufficient model self-calibration in steel bar processing have been solved, achieving efficient production management and resource utilization, and improving the robustness of production planning and the accuracy of quality traceability.

CN123113718APending Publication Date: 2026-11-17XINTIAN COUNTY XIONGCHUAN TRADING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611173688.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-04
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing steel bar processing scheduling technologies suffer from problems such as neglecting equipment changeover time constraints, lack of self-calibration in digital twin models, lack of robust assessment and refined quality traceability, resulting in insufficient closed-loop production management and resource utilization.

Method used

By constructing equipment-level digital twin models, calibrating equipment wear coefficients in real time, combining differentiated equipment changeover time constraints, performing multi-objective intelligent scheduling, and conducting multi-scenario simulations, we can achieve refined quality traceability and surplus material management.

Benefits of technology

It significantly improves the accuracy and reliability of production planning forecasts, increases equipment utilization, enhances the resilience of production plans, and achieves accurate quality traceability and surplus material utilization from batch level to single-piece level.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113718A_ABST
    Figure CN123113718A_ABST
Patent Text Reader

Abstract

This invention discloses a digital twin and dynamic intelligent scheduling method and system for steel bar processing. The method includes acquiring equipment operation and order data; constructing a digital twin model; dynamically calibrating the wear coefficient using recursive least squares or Kalman filtering based on the deviation between actual and predicted processing forces; constructing a multi-objective scheduling model by combining a differentiated changeover time matrix to generate an initial plan; performing multi-scenario simulation in the digital twin environment, calculating robustness scores, and optimizing to obtain the optimal plan; and triggering dynamic rescheduling when anomalies occur during execution. This invention solves the problems of ignoring changeover constraints in scheduling, lack of model self-calibration, and poor robustness of the plan, achieving efficient and accurate intelligent scheduling and closed-loop management.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology for steel reinforcement processing, specifically to a digital twin and dynamic intelligent scheduling method and system for steel reinforcement processing. Background Technology

[0002] With the acceleration of industrialization in the construction industry, steel bar processing is shifting from decentralized on-site operations to centralized, intelligent factory manufacturing. This involves multiple complex processes such as straightening, cutting, bending, and threading, placing extremely high demands on production efficiency, processing precision, and resource utilization. The combination of digital twin technology and intelligent scheduling algorithms aims to optimize the allocation of production resources and achieve precise control over the entire process by constructing a virtual model that maps the physical production process. This has become a key means to enhance the core competitiveness of steel bar processing enterprises.

[0003] However, existing technologies still have significant shortcomings when applied to steel bar processing scenarios: First, the production scheduling process often ignores the equipment changeover time constraints when switching between different specifications of steel bars, or fails to establish independent models for the differentiated changeover logic of different equipment such as straightening and bending, resulting in a large deviation between the production scheduling plan and the actual production capacity; Second, digital twin models are mostly static mappings, lacking a dynamic self-calibration mechanism for equipment wear coefficients based on real-time data (such as torque, vibration, etc.), causing the model fidelity to decrease over time; Third, the production scheduling plan lacks robust quantitative evaluation based on multi-scenario simulation, making it difficult to cope with uncertainties such as equipment failure or emergency orders; In addition, quality traceability is usually limited to the batch level and is disconnected from production scheduling decisions, and surplus material management also lacks refined matching methods based on three-dimensional shape features, resulting in insufficient closed-loop management and resource utilization in the overall production. Summary of the Invention

[0004] The present invention aims to overcome the above-mentioned defects in existing steel bar processing scheduling technology, and provides a digital twin and dynamic intelligent scheduling method and system for steel bar processing, so as to realize adaptive calibration of digital twin model, robust optimization of scheduling scheme and closed-loop control of the entire production process.

[0005] To address the aforementioned technical problems, this invention provides a digital twin and dynamic intelligent scheduling method for steel bar processing, comprising:

[0006] S1: Obtain the operating status parameters of steel bar processing equipment in the physical space, work-in-process processing progress information, material inventory data, and order task information;

[0007] S2: Construct an equipment-level digital twin model of the rebar processing equipment based on the operating status parameters, and dynamically calibrate the equipment wear coefficient in the equipment-level digital twin model based on the deviation between the actual processing force collected in real time and the processing force predicted by the digital twin model.

[0008] S3: Based on the calibrated equipment-level digital twin model, combined with the differentiated equipment changeover time constraints corresponding to the processing of steel bars of different specifications, a multi-objective intelligent scheduling model is constructed to generate an initial scheduling plan;

[0009] S4: Perform multi-scenario simulation and deduction of the initial production scheduling scheme in a digital twin environment, calculate the robustness score of the production scheduling scheme, and optimize the initial production scheduling scheme based on the robustness score to obtain the optimal production scheduling scheme.

[0010] S5: If a serious equipment malfunction or an urgent order is detected during the execution of the optimal production scheduling plan, dynamic rescheduling is triggered to regenerate the production scheduling plan.

[0011] The dynamic calibration of the equipment wear coefficient in the equipment-level digital twin model includes:

[0012] Calculate the predicted deviation between the actual processing force and the predicted processing force. ;

[0013] When the absolute value of the prediction deviation does not exceed 3% of the theoretical value, the recursive least squares method is used to smoothly update the equipment wear coefficient.

[0014] When the absolute value of the prediction deviation exceeds 3% but does not exceed 20% of the theoretical value, Kalman filtering is initiated for state estimation; wherein, when the absolute value of the prediction deviation exceeds 8% of the theoretical value, an early warning signal is triggered.

[0015] When the absolute value of the prediction deviation exceeds 20% of the theoretical value, it is determined to be a serious equipment malfunction, and dynamic production rearrangement is directly triggered.

[0016] Furthermore, the equipment-level digital twin model establishes a multi-axis linkage kinematic model for the CNC bending machine, including three degrees of freedom: bending axis, feed axis, and rotation axis; the predicted processing force... The calculation formula is:

[0017]

[0018] in, For a moment Predicted processing force, For a moment The equipment wear coefficient, This is based on the theoretical value of bending force under ideal conditions. For the bending angle, The diameter of the reinforcing bar. For the material's yield strength, This represents the model residuals.

[0019] Furthermore, the step of smoothly updating the equipment wear coefficient using the recursive least squares method includes:

[0020] The wear coefficient of the equipment is updated using the recursive least squares method with a forgetting factor. The update formula is as follows:

[0021]

[0022] in, This is the gain matrix of the recursive least squares algorithm. For a moment Prediction bias;

[0023] For bending machines with large load fluctuations, the forgetting factor is set to 0.95; for straightening machines with relatively stable operation, the forgetting factor is set to 0.98.

[0024] Furthermore, the initiation of Kalman filtering for state estimation includes:

[0025] Establish a state-space model, state vectors ,in The rate of change of the wear coefficient;

[0026] The state equation is:

[0027]

[0028] The observation equation is:

[0029]

[0030] Wherein, the state transition matrix Observation matrix , For process noise, To measure noise, when the wear coefficient value estimated by the Kalman filter shows a monotonically decreasing trend for three consecutive sampling periods, an early warning signal is triggered to prompt maintenance personnel to check.

[0031] Furthermore, the construction of the multi-objective intelligent scheduling model includes:

[0032] Establish differentiated changeover time matrices for straightening and cutting machines, CNC bending machines, and threading machines respectively. ,in and Representing different steel bar specifications, matrix elements indicate specifications. Switch to Specifications Required changeover time;

[0033] The optimization objectives are to minimize the total completion time, minimize the total changeover time, and maximize the utilization rate of surplus materials.

[0034] The constraints of equipment changeover time, process sequence, equipment capacity, and material availability are treated as hard constraints.

[0035] The multi-objective intelligent scheduling model is solved using a non-dominated sorting genetic algorithm to generate a Pareto optimal solution set, which serves as the initial scheduling solution.

[0036] Furthermore, the multi-scenario simulation and deduction of the initial production scheduling plan in the digital twin environment includes:

[0037] In the digital twin model, multiple abnormal scenarios are set, including at least two of the following: equipment failure and downtime, emergency order insertion, material delays, quality abnormality rework, and staff absence.

[0038] Simulations were run for each abnormal scenario, and the actual completion time, number of delayed orders, and capacity loss were recorded under each scenario.

[0039] Calculate the overall cost under each scenario :

[0040]

[0041] in, The planned completion time under normal circumstances. For the context The actual completion time is as follows. For the context The number of delayed orders Total number of orders For the context Reduced production capacity and lost working hours The total planned production capacity working hours, , , These are the weighting coefficients. = 0.4, = 0.3, = 0.3;

[0042] Calculate production scheduling plan robustness score :

[0043]

[0044] in, This represents the total number of abnormal scenarios;

[0045] The production scheduling scheme with the lowest robustness score is selected as the optimal production scheduling scheme.

[0046] Furthermore, it also includes a closed-loop traceability process for quality production scheduling:

[0047] Assign a globally unique identifier to each steel bar;

[0048] During the processing, the processing parameters and quality inspection data of each steel bar are collected in real time, and the processing parameters and quality inspection data are bound to the globally unique identifier to form a single-bar quality traceability digital thread.

[0049] When a quality anomaly occurs, the production batch and processing equipment parameters that caused the anomaly are located in reverse based on the quality traceability digital thread.

[0050] When the product defect rate of a certain piece of equipment exceeds a preset threshold, the task allocation weight of the equipment in subsequent production scheduling is reduced. For every percentage point the defect rate exceeds the threshold, the weight is reduced by 5%, down to a minimum of 50% of the original weight.

[0051] After the equipment has been repaired and has continuously processed 50 defect-free products, the task allocation weight will automatically be restored to its original value.

[0052] Furthermore, it also includes a smart matching step for surplus materials:

[0053] Three-dimensional point cloud data of steel bar scraps are obtained through 3D scanning equipment, and a 3D digital model of the scraps is constructed.

[0054] Extract the geometric features of the three-dimensional digital model of the scrap material, wherein the geometric features include at least the length of the straight segment, the bending angle, and the flatness of the end;

[0055] The scrap material is classified and graded based on the aforementioned geometric features. The classification and grading include:

[0056] The first level is a hard match between material and specifications, requiring that the material grade be consistent and the diameter be within the allowable deviation range of the required specifications, wherein the allowable deviation is ±0.5 mm;

[0057] The second level is shape classification. Leftover material with a bending angle of no more than 5 degrees and an end flatness of no less than 0.9 is classified as straight strip type, with a straightening loss rate of zero; leftover material with a bending angle of more than 5 degrees but no more than 30 degrees is classified as adjustable type, with a straightening loss rate of 2% to 4%; leftover material with a bending angle of more than 30 degrees is classified as curved type.

[0058] For adjustable straightening scrap, the effective length equal to inventory length Multiply by (1 minus the straightening loss rate) );

[0059] The third level is the overall fit score, and the overall fit score... The calculation formula is:

[0060]

[0061] in, = 0.6, = 0.25, = 0.15;

[0062] Length fit The calculation rule is: when hour, = 1; when hour, ;when or hour, = 0;

[0063] Shape fit Warehouse location convenience The path distance of the automated guided vehicle is calculated and then normalized to the interval between 0 and 1.

[0064] Based on the classification results, the surplus material with the highest overall suitability score is selected and included in the production material list, and the material inventory data is updated.

[0065] Another aspect of the present invention provides a digital twin and dynamic intelligent scheduling system for steel bar processing, comprising:

[0066] The physical space perception module is deployed on-site in the steel bar processing workshop to collect multi-dimensional data of the physical workshop in real time. The multi-dimensional data includes the operating status parameters of each processing equipment, the processing progress information of work-in-process, material inventory data, and order task information.

[0067] The equipment-level digital twin modeling and dynamic calibration module, running on the central controller, is used to construct a refined digital twin model of the rebar processing equipment. Based on the deviation between the actual processing force collected in real time and the processing force predicted by the model, a three-level response strategy is used to dynamically calibrate the equipment wear coefficient in the model. The three-level response strategy is as follows: when the absolute value of the predicted deviation does not exceed 3% of the theoretical value, the recursive least squares method is used to smoothly update the equipment wear coefficient; when the absolute value of the predicted deviation exceeds 3% but does not exceed 20% of the theoretical value, Kalman filtering is initiated for state estimation, and when the absolute value of the predicted deviation exceeds 8% of the theoretical value, an early warning signal is triggered; when the absolute value of the predicted deviation exceeds 20% of the theoretical value, it is determined that the equipment is seriously abnormal, and dynamic rescheduling is directly triggered.

[0068] The multi-objective intelligent scheduling and robust optimization module runs on the central controller. It is used to construct a multi-objective intelligent scheduling model based on the calibrated digital twin model and combined with the differentiated equipment changeover time matrix, generate an initial scheduling plan, calculate the robustness score of the scheduling plan through multi-scenario simulation and deduction, and output the optimal scheduling plan.

[0069] The dynamic rescheduling module runs on the central controller and is used to regenerate the production schedule using a rolling time-domain optimization strategy when a serious equipment malfunction or an urgent order is detected.

[0070] The quality scheduling closed-loop traceability module runs on the central controller and is used to assign a globally unique identifier to each steel bar. When a quality abnormality occurs, it reverses the location to the scheduling batch and equipment and feeds the traceability results back to the multi-objective intelligent scheduling and robust optimization module.

[0071] The intelligent matching module for scrap materials runs on the central controller and is used to construct a three-dimensional digital model of the scrap materials through three-dimensional scanning and to match available scrap materials based on a hierarchical classification strategy.

[0072] The central controller includes a processor and a memory. The memory stores a computer program. When the processor executes the computer program, it implements the functions of the above-mentioned modules. The central controller also interfaces with the upstream system through a standardized interface to automatically import order information and provide feedback on production scheduling results.

[0073] Furthermore, the physical space sensing module includes a sensor array and a data acquisition unit:

[0074] The sensor array includes torque sensors, vibration sensors, temperature sensors, and current sensors installed on each processing device;

[0075] The data acquisition unit has a data acquisition cycle of 100 milliseconds and transmits the data to the central controller via industrial Ethernet.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] The present invention provides a digital twin and dynamic intelligent scheduling method and system for steel bar processing. By dynamically calibrating the equipment wear coefficient using recursive least squares, Kalman filtering, and a fault-triggered three-level response strategy based on the deviation between the actual processing force collected in real time and the processing force predicted by the digital twin model, this method solves the problem of static digital twin models and lack of self-calibration capabilities, which leads to a decrease in model fidelity over time. This keeps the deviation between simulation results and actual execution within 5%, significantly improving the model's prediction accuracy and reliability. Furthermore, by establishing differentiated changeover time matrices for straightening and cutting machines, CNC bending machines, and threading machines, and incorporating them as independent constraints into a multi-objective intelligent scheduling model, this method addresses the problem of neglecting equipment changeover time constraints and frequent changeovers reducing capacity in existing scheduling methods. It effectively reduces ineffective waiting time and increases equipment utilization from approximately 70% to 82%-88%. Finally, by simulating the initial scheduling plan in a digital twin environment under various abnormal scenarios, including equipment failure and emergency order insertion, robustness scores are calculated and combined with Pareto bi-objective optimization and TOPSIS decision selection. By selecting the optimal solution, existing production scheduling schemes can be improved to address their lack of robustness assessment and inability to cope with uncertainties. This increases the average on-time delivery rate under abnormal scenarios from approximately 78% to approximately 90%, enhancing the resilience of production plans. Furthermore, by assigning a globally unique identifier to each rebar, a single-bar-level quality traceability digital thread is established. The quality defect rate is then fed back to the scheduling module to dynamically adjust equipment task allocation weights. This addresses the disconnect between existing quality traceability and scheduling, and the inability to achieve precise positioning and closed-loop optimization, enabling improvements from batch-level to single-bar-level optimization. Precise traceability at the root level and automatic avoidance of high-defect-rate equipment form a virtuous cycle of quality and production scheduling; by constructing a 3D digital model of scrap materials through 3D scanning, and adopting a three-level hierarchical retrieval strategy of material hard matching, shape classification (straight, adjustable straight, and curved), and comprehensive adaptability scoring for scrap material matching, it can solve the problems of extensive scrap material management, poor matching accuracy, and failure to consider the characteristics of curved shapes. It significantly improves the accuracy of scrap material identification and utilization, increases the scrap material utilization rate from 40%-50% to 65%-75%, and effectively reduces raw material costs. Attached Figure Description

[0078] Figure 1 This is a flowchart illustrating the digital twin and dynamic intelligent scheduling method for steel bar processing according to the present invention.

[0079] Figure 2 This is a schematic diagram of the architecture of the digital twin and dynamic intelligent scheduling system for steel bar processing of the present invention.

[0080] Figure 3 This is a flowchart of the three-level dynamic calibration process for equipment wear coefficient in this invention.

[0081] Figure 4This is a flowchart of the multi-objective intelligent scheduling and robustness evaluation process in this invention.

[0082] Figure 5 This is a flowchart of the three-dimensional digital model of the waste material and the three-level hierarchical retrieval and matching process in this invention. Detailed Implementation

[0083] The technical solutions in the embodiments of this disclosure 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, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0084] See Figure 1 In one embodiment of this application, a digital twin and dynamic intelligent scheduling method for steel bar processing is provided, the method comprising the following steps:

[0085] S1: Obtain the operating status parameters of the steel bar processing equipment in the physical space, the processing progress information of work-in-process, the material inventory data, and the order task information.

[0086] Step S1 specifically involves acquiring multi-dimensional real-time data of the steel bar processing workshop through a physical space sensing module. The physical space sensing module is deployed on-site in the steel bar processing workshop, and its core components include a sensor array and a data acquisition unit.

[0087] In some embodiments, the sensor array includes torque sensors, vibration sensors, temperature sensors, and current sensors installed on key parts of various processing equipment (such as straightening and cutting machines, CNC bending machines, threading machines, welding robots, etc.). The torque sensors collect load torque data of the equipment's spindle or actuator; the vibration sensors monitor the mechanical vibration spectrum during equipment operation; the temperature sensors monitor the operating temperature of the motor and bearings; and the current sensors record the real-time current value of the drive motor. These sensors together constitute a comprehensive sensing network for the equipment's operating status parameters, capable of reflecting the equipment's health status and processing load in real time.

[0088] In some implementations, the data acquisition unit is connected to the sensor array and configured to perform high-frequency data acquisition with a period of 100 milliseconds. The data acquired by the data acquisition unit includes not only the above-mentioned equipment operating status parameters, but also the processing progress information of work-in-process (such as the current process completion rate and flow position), material inventory data (including raw material balance, three-dimensional model information and warehouse location coordinates in the surplus material warehouse), and order task information issued by the upstream system (including steel bar specifications, quantity, delivery date requirements, etc.).

[0089] In some implementations, the data acquisition unit transmits the acquired multi-dimensional data to the central controller via Industrial Ethernet. The high bandwidth and low latency of Industrial Ethernet ensure that massive amounts of sensor data can be uploaded in real time and stably, providing an accurate data foundation for the subsequent construction of a high-fidelity device-level digital twin model. After receiving the data, the central controller performs timestamp alignment, noise filtering, and format standardization on the received data to form a unified data stream for subsequent processing modules to use.

[0090] S2: Construct an equipment-level digital twin model of the rebar processing equipment based on the operating status parameters, and dynamically calibrate the equipment wear coefficient in the equipment-level digital twin model based on the deviation between the actual processing force collected in real time and the processing force predicted by the digital twin model.

[0091] Step S2 specifically includes constructing an equipment-level digital twin model of the rebar processing equipment, and dynamically calibrating the equipment wear coefficient in the equipment-level digital twin model based on the deviation between the actual processing force collected in real time and the processing force predicted by the digital twin model.

[0092] In some implementations, the construction of the equipment-level digital twin model includes two levels: a workshop-level model and an equipment-level model. The workshop-level model reflects the spatial layout, material flow paths, and capacity matching relationships of each processing device. The equipment-level model establishes the kinematic and dynamic models for each individual rebar processing device. Taking a CNC bending machine as an example, a multi-axis linkage kinematic model is established, including three degrees of freedom: bending axis, feed axis, and rotation axis, capable of simulating the rebar deformation process under different bending angles. The predicted processing force... The calculation formula is:

[0093]

[0094] in, For a moment Predicted processing force, For a moment The equipment wear coefficient, The bending force is the theoretical value calculated using formulas from mechanics of materials under ideal conditions (i.e., new equipment, standard material conditions). For the bending angle, The diameter of the reinforcing bar. For the material's yield strength, This represents the model residuals.

[0095] like Figure 3 As shown, in some embodiments, the dynamic calibration of the equipment wear coefficient adopts a three-level response strategy, and the specific process is as follows:

[0096] Level 1 response: When prediction deviation occurs The absolute value does not exceed the theoretical value. When the value is 3%, it is considered a normal fluctuation. The forgetting factor recursive least squares (FFRLS) method is used to calculate the equipment wear coefficient. Perform a smooth update. The smooth update formula is:

[0097]

[0098] in, This is the gain matrix for the recursive least squares algorithm. Considering the differences in dynamic characteristics between different devices, the forgetting factor... The value of the forgetting factor varies: for CNC bending machines with large load fluctuations, the forgetting factor is set to 0.95 to be more sensitive to recent data; for straightening and cutting machines with relatively stable operation, the forgetting factor is set to 0.98. The recursive least squares method of the forgetting factor requires approximately 50 to 100 sampling points (i.e., 5 to 10 seconds) to converge to the true value, and the factory calibration value is used before convergence.

[0099] Second-level response: When the absolute value of the prediction deviation exceeds 3% but does not exceed 20% of the theoretical value, it is considered a significant deviation, and Kalman filtering is initiated for state estimation. A state-space model is established for the state estimation, and the state vector... ,in Let be the rate of change of the wear coefficient. The state equation and the observation equation are as follows:

[0100]

[0101]

[0102] Wherein, the state transition matrix Observation matrix , The process noise is given by: , For measuring noise, the covariance matrix of the measured noise is: ; and The value is determined through statistical analysis of the equipment's historical operating data. The recursive process of Kalman filtering includes two steps: prediction and update. The update cycle is consistent with the data acquisition cycle, which is 100 milliseconds. During the Kalman filtering stage, if the absolute value of the prediction deviation exceeds 8% of the theoretical value, an early warning signal is triggered to prompt maintenance personnel to check. At the same time, when the wear coefficient value estimated by Kalman filtering shows a monotonically decreasing trend for three consecutive sampling cycles, an early warning signal is also triggered.

[0103] Level 3 response: When the absolute value of the prediction deviation exceeds 20% of the theoretical value, it is judged as a serious equipment malfunction, and dynamic rescheduling is directly triggered to transfer or suspend the current task in order to avoid generating a large number of scraps.

[0104] In some implementations, a material batch variation coefficient is also introduced to further improve the accuracy of the digital twin model. Online calibration and decoupling strategy for the material batch variation coefficient and equipment wear; definition of material batch variation coefficient. This represents the ratio of the mechanical properties of the current batch of materials to those of the standard batch; the initial value is determined by the incoming inspection. Whenever a batch of materials begins processing, the system collects the actual processing force data of the first 20 products and calculates the average deviation for that batch. And update the coefficient of variation of the material for this batch:

[0105]

[0106] in, The learning rate is 0.6 to 0.8.

[0107] Since the actual deviation is affected by both equipment wear and material variation, this implementation adopts a layered calibration strategy for decoupling: high-frequency components (short-term fluctuations) are attributed to material variation, and low-frequency trends (long-term drift) are attributed to equipment wear. Specifically, a second-order Butterworth low-pass filter is used to extract the equipment wear trend, and a second-order Butterworth high-pass filter is used to extract the material variation component. The cutoff frequency of the low-pass filter is set to 0.001 Hz, meaning changes with a period greater than 1000 seconds are considered equipment wear; the cutoff frequency of the high-pass filter is set to 0.01 Hz, meaning changes with a period less than 100 seconds are considered material variation. The wear coefficient of each piece of equipment is stored in a database in time series, forming an equipment wear curve for predictive maintenance.

[0108] S3: Based on the calibrated equipment-level digital twin model, combined with the differentiated equipment changeover time constraints corresponding to the processing of steel bars of different specifications, a multi-objective intelligent scheduling model is constructed to generate an initial scheduling plan.

[0109] like Figure 4 As shown, in some embodiments, the construction process of the multi-objective intelligent scheduling model specifically includes the establishment of a differentiated transformation time matrix, the definition of a multi-objective optimization function, and a solution strategy based on the non-dominated sorting genetic algorithm (NSGA-II).

[0110] First, to address the production capacity loss caused by switching between different specifications of steel bars during steel bar processing, the system establishes differentiated changeover time matrices for straightening and cutting machines, CNC bending machines, and threading machines. .

[0111] in, and These represent the rebar specifications before and after the switch, respectively; matrix elements. Indicates from specifications Switch to Specifications The required standardized changeover time. Taking a CNC bending machine as an example, because different diameters and bending shapes of steel bars require different bending dies and adjustments to the robotic arm parameters, the changeover time of a CNC bending machine exhibits significant asymmetric and nonlinear characteristics. Through field measurements and historical data accumulation, the following changeover time mapping relationship was established: switching from a 10 mm diameter steel bar to a 12 mm diameter steel bar takes 8 minutes, from 10 mm to 16 mm takes 15 minutes, and from 10 mm to 20 mm takes 22 minutes; while switching from 12 mm to 16 mm takes 10 minutes, from 12 mm to 20 mm takes 18 minutes, and from 16 mm to 20 mm takes 12 minutes. Similarly, the changeover of a straightening and cutting machine mainly involves the replacement of the straightening wheel and the cutting tool, while the changeover of a threading machine mainly involves the replacement of the thread rolling wheel. The straightening and cutting machine and the threading machine each independently store their corresponding changeover time matrices. Each changeover time matrix is ​​directly embedded into the production scheduling model as a hard constraint, ensuring that the generated production scheduling plan is physically executable and conforms to the actual process time.

[0112] Secondly, a multi-objective optimization function is constructed to balance production efficiency, resource utilization, and changeover costs; the optimization objectives include minimizing the total completion time. Maximize the utilization rate of surplus materials and minimize total equipment changeover time The objective function of the multi-objective scheduling optimization model is defined as:

[0113]

[0114] in, Based on the completion time, The baseline changeover time is used as the benchmark. Simultaneously, the multi-objective production scheduling optimization model must meet the following hard constraints: equipment capacity constraint, meaning that the processing load of each piece of equipment in any given time period does not exceed the maximum theoretical capacity of each piece of equipment; process sequence constraint, meaning that straightening, cutting, bending, and threading processes must follow a strict sequential logical relationship; delivery date constraint, meaning that the final completion time of each order must not be later than the delivery date specified by the customer; and material availability constraint, ensuring that raw materials or matching surplus materials are ready before processing begins.

[0115] Finally, the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) is used to solve the multi-objective intelligent scheduling model. The NSGA-II algorithm, through fast non-dominated sorting and crowding distance calculation, can effectively distribute the solution set in the multi-objective space, avoiding getting trapped in local optima. In some implementations, the NSGA-II algorithm parameters are set as follows: population size is 200, and maximum number of iterations is 500. During the algorithm's operation, a random scheduling sequence is generated as the initial population. Offspring are produced through selection, crossover, and mutation operations, and an elite retention strategy is used to select the Pareto optimal solution set. This Pareto optimal solution set constitutes the basic initial scheduling scheme set for subsequent robustness evaluation, where each scheme represents a different trade-off between completion time, changeover time, and surplus material utilization.

[0116] S4: Perform multi-scenario simulation and deduction of the initial production scheduling scheme in a digital twin environment, calculate the robustness score of the production scheduling scheme, and optimize the initial production scheduling scheme based on the robustness score to obtain the optimal production scheduling scheme.

[0117] Step S4 specifically includes performing multi-scenario simulations of the initial production scheduling plan in a digital twin environment to evaluate and optimize the robustness of the production scheduling plan.

[0118] Specifically, the system first sets up various abnormal scenarios based on historical anomaly event statistics. These scenarios can include, but are not limited to, at least two of the following: equipment failure downtime, emergency order insertion, material delays, quality defects requiring rework, and employee absence. For example, an equipment failure downtime scenario could be set as a random downtime of a critical processing machine (such as a CNC bending machine) for 10 to 60 minutes; an emergency order insertion scenario could be set as a new order with a delivery time requirement shortened by 20% to 50%; and a material delay scenario could be set as a delay of 1 to 4 hours in the arrival of a batch of raw materials. The number of scenarios... The simulation objects can be determined based on the frequency of abnormal events in the workshop over the past 12 months. Usually, the top few types of abnormal events with the highest frequency of occurrence are selected as simulation objects.

[0119] Subsequently, an initial production scheduling plan was developed for each candidate. The system executes the above-mentioned steps in the digital twin model respectively. Simulation of various abnormal scenarios; recording the results of each scenario during the simulation. Actual completion time Number of delayed orders and lost working hours ;in, The planned completion time under normal circumstances. Total number of orders The total planned production capacity man-hours.

[0120] Next, calculate each scenario. The overall cost The comprehensive cost function is defined as follows:

[0121]

[0122] in, , , As a weighting factor, in some implementations, Take 0.4, Take 0.3, The formula above quantifies the time delay rate, order delay rate, and capacity loss rate of the production scheduling plan relative to the baseline plan under specific abnormal scenarios, taking 0.3 as an example.

[0123] Calculate the production scheduling plan based on the comprehensive cost of each scenario. robustness score The robustness score is defined as the arithmetic mean of the combined costs across all scenarios.

[0124]

[0125] The lower the robustness score, the more stable the corresponding production scheduling scheme is in the face of uncertainty disturbances, that is, the stronger the robustness.

[0126] After obtaining the efficiency metrics (total completion time under normal circumstances) of multiple candidate production scheduling schemes ) and robustness indicators (robustness score) After that, the system uses a Pareto bi-objective optimization strategy to generate a set of Pareto front solutions. Since minimizing completion time and maximizing robustness (i.e. minimizing robustness score) often conflict, the solutions on the Pareto front represent non-dominated solutions under different trade-offs.

[0127] To select the final optimal production scheduling plan from the Pareto front set of solutions, this implementation method employs the TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) multi-attribute decision-making method. The specific process is as follows:

[0128] First, a decision matrix is ​​constructed, where the rows of the decision matrix correspond to the candidate solutions on the Pareto front, and the columns of the decision matrix correspond to two evaluation dimensions: efficiency optimization (minimization). ) and robustness optimal (minimize) ).

[0129] Secondly, the decision-making data is standardized to eliminate the influence of dimensions.

[0130] Then, the positive and negative ideal solutions are determined. The positive ideal solution consists of the optimal value for each evaluation dimension (i.e., the minimum). and the smallest The negative ideal solution consists of the worst value of each evaluation dimension (i.e., the largest). and the largest ).

[0131] Next, the distance between each candidate solution and the ideal solution is calculated. and the distance from the negative ideal solution .

[0132] Finally, the relative similarity of each scheme is calculated. The calculation formula is: Relative closeness The larger the value, the closer the candidate solution is to the positive ideal solution and the farther it is from the negative ideal solution; the system selects the solution with the highest relative closeness as the final recommended optimal production scheduling solution; (where, " "This indicates that the closeness score is the final index value obtained after standardization and comprehensive evaluation, and is used to distinguish the subscript markers in the aforementioned comprehensive cost function."

[0133] In some implementations, the system may also provide a maximum cost criterion as a supplement to the conservative assessment, i.e., calculation It is used to identify the worst-performing solution in the worst-case scenario to assist managers in risk control; the final selected optimal production scheduling solution is issued to each processing equipment at the execution level after being confirmed by managers.

[0134] S5: If a serious equipment malfunction or an urgent order is detected during the execution of the optimal production scheduling plan, dynamic rescheduling is triggered to regenerate the production scheduling plan.

[0135] The dynamic rescheduling process described in step S5 specifically includes three stages: abnormal event identification, strategy matching, and rolling time-domain optimization.

[0136] When the physical space perception module detects in real time that the equipment operating status parameters deviate from the preset threshold or receives an emergency order instruction from an external system, the dynamic production rescheduling module first analyzes the trigger signal to identify the type of abnormal event. The system has a pre-set abnormal event library, which stores the characteristic signal patterns, priority levels, and corresponding initial response strategies for various typical abnormal events. For example, for a CNC bending machine failure shutdown event, the characteristic signal of the CNC bending machine failure shutdown event is that the signal collected by the torque sensor suddenly drops to 0. The system marks this event as a high-priority event, and the initial response strategy is to transfer the current task to other available equipment of the same type. For a straightening machine processing timeout event, the characteristic signal of the straightening machine processing timeout event is that the actual processing time exceeds the planned value by 20%. The system marks this event as a medium-priority event, and the initial response strategy is to adjust the processing rhythm of subsequent processes. For quality abnormal events, such as when the vision inspection system detects that the bending angle of 3 products exceeds the tolerance in a row, the system marks the quality abnormal event as a medium-priority event, and the initial response strategy is to suspend the corresponding equipment and trigger a recalibration process.

[0137] After determining the type of abnormal event, the dynamic rescheduling module uses a rolling time-domain optimization strategy to generate a new scheduling plan. The core of the rolling time-domain optimization strategy is to re-plan only the local process windows affected by the abnormal event, rather than globally reconstructing the entire production plan, thereby significantly reducing computational complexity and shortening response time. Specifically, the system takes the current moment as the starting point and selects processes within a fixed duration (e.g., 4 hours) as the optimization window. Within the optimization window, the system retains the original state of tasks that have not started and are not affected by the abnormality, and focuses on rescheduling tasks that have been assigned to faulty equipment or affected by urgent orders.

[0138] Within the optimization window, the system, based on the constraints of the multi-objective intelligent scheduling model, and combined with the current equipment availability, material inventory, and order delivery requirements, uses a non-dominated sorting genetic algorithm or other heuristic algorithms to quickly find local optima. For example, when a bending machine fails, the system assesses the current load rate of the remaining available bending machines, reallocates tasks originally assigned to the failed machine to alternative machines with lower loads, and simultaneously adjusts the start times of the preceding straightening process and the subsequent threading process to ensure logical connection and capacity balance between processes. For urgent order insertion scenarios, the system prioritizes the critical path processes of urgent orders within the optimization window and meets delivery requirements by compressing the buffer time of non-critical path tasks.

[0139] Through the aforementioned rolling time-domain optimization strategy, the dynamic rescheduling module can generate an adjusted scheduling plan within 3 to 8 minutes and distribute the new plan to each execution terminal in the execution layer.

[0140] In one embodiment of this application, the method further includes a quality scheduling closed-loop traceability step; the quality scheduling closed-loop traceability step is used to establish a digital twin file for the entire life cycle of each steel bar from raw material entry to finished product exit; specifically, during the scheduling stage, the system assigns a globally unique identifier to each steel bar, which is implemented in the form of a QR code and attached to the end of the raw material; during the steel bar processing, the processing parameters (such as bending angle, bending speed, pressure holding time, etc.) and quality inspection data of each steel bar are collected in real time, and the processing parameters and quality inspection data are bound to the globally unique identifier to form a single-bar-level quality traceability digital thread; the information associated with the quality traceability digital thread includes at least raw material batch information, production plan information, processing equipment number, process parameter combination, operator information, and final quality inspection results.

[0141] When a defect is detected during quality inspection, the system performs reverse tracing based on the aforementioned quality traceability digital thread. By scanning the QR code at the end of the rebar or retrieving process flow records (including processing equipment number, processing time, processing parameter combinations, etc.), the system can automatically locate the specific production batch, specific processing equipment and its process parameters at that time, specific operator, and operation time that caused the anomaly. For example, if the bending angle of a 16 mm diameter bent rebar is found to be out of tolerance, the system can trace back to the corresponding CNC bending machine and its operating status at that time within milliseconds, thereby quickly identifying whether the problem is caused by a sudden equipment failure, incorrect parameter settings, or material variation.

[0142] Preferably, the quality scheduling closed-loop traceability step automatically feeds back the traceability results to the multi-objective intelligent scheduling and robust optimization module to achieve closed-loop control of quality and scheduling. The system uses the 3-standard-deviation principle in statistical process control to determine the defect rate threshold. That is, when the product defect rate of a certain equipment exceeds the historical average plus 3 standard deviations, a weight adjustment mechanism is triggered. The specific adjustment strategy is: for every percentage point the defect rate exceeds the threshold, the task allocation weight of the corresponding equipment in subsequent scheduling is reduced by 5%, down to a minimum of 50% of the original weight. The above mechanism can effectively avoid continuing to allocate high-priority or high-precision orders during periods of equipment instability, thereby reducing batch quality risks.

[0143] In some implementations, the task allocation weight restoration mechanism is as follows: after the equipment has been repaired or calibrated, if 50 products are processed continuously without defects, the system automatically restores the task allocation weight of the equipment to its original value. The above dynamic weight adjustment mechanism ensures that the production scheduling plan can respond to quality fluctuations on the production site in real time, improving the robustness of the overall production system and the consistency of product quality. In addition, in batch processing scenarios, even if the ends of the steel bars are cut and the QR code labels are lost, the system can still accurately match and trace through the timestamps, equipment IDs and parameter fingerprints in the process flow records, ensuring the integrity of the traceability accuracy at the single-strand level.

[0144] like Figure 5 As shown, in one embodiment of this application, the method further includes a waste material intelligent matching step; the waste material intelligent matching step is used to digitally manage and intelligently match the waste materials generated during the steel bar processing; specifically, when the steel bar processing equipment generates waste materials after completing the cutting or bending process, the system uses a handheld 3D scanner or a fixed 3D scanning device deployed on-site in the workshop to scan the waste materials in real time; the scanning accuracy is controlled within 1 mm, and the scanning time for a single waste material is about 30 seconds; the original data obtained by scanning is 3D point cloud data, and the system performs noise reduction, registration and feature extraction processing on the point cloud data to construct a 3D digital model of the waste materials and store it in the waste material digital model library.

[0145] In some implementations, the system extracts geometric feature vectors from the three-dimensional digital model of the scrap material. The geometric features include at least the total length of the straight line segment, the maximum diameter, the equivalent cross-sectional area, the bending angle, the bending radius, the bending position coordinates, and the end flatness index. At the same time, the system also records attribute information such as the material grade, storage time, and current storage location coordinates of the scrap material to form a complete digital archive of the scrap material.

[0146] In order to efficiently retrieve surplus materials from the surplus material warehouse that meet the needs of the current production order, this implementation method adopts a three-level hierarchical retrieval and matching strategy.

[0147] The first level is a hard match between material and specifications. The system first filters out the surplus material whose material grade matches the required steel bars, and then checks whether the diameter of the filtered surplus material is within the allowable deviation range of the required specifications. In this embodiment, the allowable deviation is set to ±0.5 mm. If the material grade of the surplus material is inconsistent or the diameter exceeds the allowable deviation range, it will be directly eliminated and will not enter the subsequent matching process.

[0148] The second level involves shape classification and effective length calculation; based on the extracted bending angle and end flatness indicators, the system categorizes the leftover material that passed the first level of screening into three types:

[0149] Straight bar scrap: The bending angle is no more than 5 degrees and the flatness of the end is not less than 0.9; the straight bar scrap can be used directly as straight steel bars, and the straightening loss rate is set to zero.

[0150] Adjustable straightening scrap: The condition is that the bending angle is greater than 5 degrees but not more than 30 degrees; the adjustable straightening scrap must be straightened by a straightening machine before it can be used, and the system sets the straightening loss rate to 2% to 4% according to the grade of the rebar; for the adjustable straightening scrap, the system calculates its effective length. The calculation formula is:

[0151]

[0152] in The actual length measured in stock. To straighten the loss rate.

[0153] Bending type scrap: The condition is that the bending angle is greater than 30 degrees or the end is severely uneven; the bending type scrap is only suitable as raw material for specific bending parts and is not used for matching straight steel bars.

[0154] The third level is the overall fit score; for the candidate surplus materials that have passed the first two levels of screening, the system calculates their overall fit. The calculation formula is as follows:

[0155]

[0156] in, , , The weighting coefficients for length adaptability, shape adaptability, and storage location convenience are 0.6, 0.25, and 0.15, respectively, in this embodiment.

[0157] Specifically, length fit The calculation rules are as follows:

[0158] when hour, ;

[0159] when hour,

[0160]

[0161] when or hour, .

[0162] in, The required length of steel bars for the order.

[0163] Shape fit The calculation formula is:

[0164]

[0165] in The bending angle of the remaining material.

[0166] Warehouse location convenience The path distance from the current location of the Automated Guided Vehicle (AGV) to the waste material storage location is calculated and normalized to the range of 0 to 1. The shorter the distance, the better. The higher the value.

[0167] The system sorts all candidate scrap materials based on a comprehensive fit score, selects the scrap material with the highest score for inclusion in the output material list, and updates the material inventory data. If the remaining length of the selected scrap material exceeds the minimum usable length (e.g., 500 mm) after being cut and used, the system automatically generates a new scrap material record and updates the 3D scan model of the new scrap material; if the remaining length is less than the minimum usable length, it is marked as waste and enters the waste management process.

[0168] In addition, this implementation also includes a lifecycle management mechanism for the surplus material warehouse; each surplus material is set with a storage period, such as 90 days, when it is put into the warehouse; the system scans the surplus material warehouse regularly, and for surplus materials that have not been used after the storage period has expired, the status of the surplus material is automatically downgraded to waste and removed from the available surplus material warehouse, so as to avoid quality risks caused by rust or deformation and ensure the quality reliability of the surplus materials used in production scheduling.

[0169] like Figure 2 As shown, in one embodiment of the present invention, a digital twin and dynamic intelligent scheduling system for steel bar processing is also provided; the system includes a physical space perception module, an equipment-level digital twin modeling and dynamic calibration module, a multi-objective intelligent scheduling and robust optimization module, a dynamic rescheduling module, a quality scheduling closed-loop traceability module, a surplus material intelligent matching module, and a central controller.

[0170] The physical space perception module is deployed on-site in the steel bar processing workshop to collect multi-dimensional data of the physical workshop in real time. In some embodiments, the physical space perception module includes a sensor array and a data acquisition unit. The sensor array specifically includes torque sensors, vibration sensors, temperature sensors, and current sensors installed on key parts of various processing equipment (such as straightening and cutting machines, CNC bending machines, threading machines, welding robots, etc.). For example, a torque sensor is installed at the drive shaft of the CNC bending machine to monitor bending torque, a temperature sensor is installed on the motor housing to monitor overheating, and a vibration sensor is installed at the main stress points of the machine body to monitor abnormal vibration.

[0171] The data acquisition unit is used to preprocess and transmit analog or digital signals acquired by the sensor array. In some embodiments, the data acquisition period of the data acquisition unit is set to 100 milliseconds to ensure that high-frequency transient changes during equipment operation can be captured. The data acquired by the data acquisition unit is transmitted to the central controller through the industrial Ethernet protocol to ensure the real-time performance and stability of data transmission. The multi-dimensional data specifically includes the operating status parameters of each processing equipment (such as rotation speed, temperature, vibration amplitude, current magnitude, torque value, etc.), processing progress information of work-in-process, material inventory data (including surplus material information), and order task information.

[0172] The central controller is the core processing unit of the system, including a processor and a memory; the memory stores a computer program, and the processor executes the computer program to realize the functions of the above modules; in some embodiments, the central controller may be a high-performance industrial server or an edge computing gateway, with a multi-core processor and a large capacity of memory to support complex digital twin model operations and multi-objective optimization algorithm solutions.

[0173] The equipment-level digital twin modeling and dynamic calibration module, the multi-objective intelligent scheduling and robust optimization module, the dynamic rescheduling module, the quality scheduling closed-loop traceability module, and the surplus material intelligent matching module all run on the central controller. The above modules interact with each other through an internal bus or memory sharing mechanism. For example, the physical space perception module sends the real-time collected data to the equipment-level digital twin modeling and dynamic calibration module. After completing the model calibration, the equipment-level digital twin modeling and dynamic calibration module transmits the updated equipment status parameters to the multi-objective intelligent scheduling and robust optimization module. The multi-objective intelligent scheduling and robust optimization module generates a scheduling plan and sends it to the execution layer, and feeds back the scheduling results to the quality scheduling closed-loop traceability module and the surplus material intelligent matching module.

[0174] In addition, the central controller also interfaces with upstream ERP and MES systems through standardized interfaces (such as RESTful APIs). Through these standardized interfaces, the central controller automatically imports order information and BIM design data, and automatically feeds back the generated final production scheduling results to the upstream systems, achieving seamless integration of production planning and enterprise resource planning.

[0175] In this embodiment, the workflow of the digital twin and dynamic intelligent scheduling system for steel bar processing forms a complete closed loop from perception, modeling, calibration, scheduling, simulation verification, execution, monitoring, rescheduling, traceability to feedback.

[0176] Specifically, the physical space perception module continuously collects real-time data from the steel bar processing workshop, including the operating status parameters of each processing equipment, the processing progress information of work-in-process, material inventory data, and order task information, and transmits the data to the central controller; the equipment-level digital twin modeling and dynamic calibration module constructs a digital twin model based on the received real-time data, and updates the equipment wear coefficient and material batch variation coefficient online through the recursive least squares method and Kalman filtering three-level response strategy to ensure the high fidelity of the model.

[0177] Subsequently, the multi-objective intelligent scheduling and robust optimization module aims to minimize the total processing time, maximize the utilization rate of surplus materials, and minimize the total equipment changeover time. It generates a Pareto optimal solution set by combining a differentiated equipment changeover time matrix. The multi-objective intelligent scheduling and robust optimization module performs multi-scenario simulation and deduction on the initial scheduling scheme in a digital twin environment, calculates the robustness score, and uses Pareto bi-objective optimization combined with TOPSIS decision selection to select the final optimal scheduling scheme.

[0178] After the optimal production scheduling plan is issued and executed, the physical space perception module continuously monitors the production status. When events such as equipment failure, processing timeout, or quality anomalies are detected, the dynamic rescheduling module automatically identifies the event type, uses a rolling time-domain optimization strategy for local rescheduling, and generates an adjusted production schedule for execution. Simultaneously, the quality scheduling closed-loop traceability module establishes a unique digital identity for each rebar, linking it to information across the entire chain. When quality defects are discovered, it performs reverse tracing and feeds the tracing results back to the multi-objective intelligent scheduling and robust optimization module to adjust subsequent scheduling weights. Furthermore, the surplus material intelligent matching module continuously updates the surplus material 3D digital model library and automatically retrieves and matches available surplus materials during the scheduling process, thereby achieving efficient resource utilization.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them; although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A digital twin and dynamic intelligent scheduling method for steel bar processing, characterized in that, include: S1: Obtain the operating status parameters of steel bar processing equipment in the physical space, work-in-process processing progress information, material inventory data, and order task information; S2: Construct an equipment-level digital twin model of the steel bar processing equipment based on the operating status parameters, and dynamically calibrate the equipment wear coefficient in the equipment-level digital twin model based on the deviation between the actual processing force collected in real time and the processing force predicted by the digital twin model. S3: Based on the calibrated equipment-level digital twin model, combined with the differentiated equipment changeover time constraints corresponding to the processing of steel bars of different specifications, a multi-objective intelligent scheduling model is constructed to generate an initial scheduling plan; S4: Perform multi-scenario simulation and deduction of the initial production scheduling scheme in a digital twin environment, calculate the robustness score of the production scheduling scheme, and optimize the initial production scheduling scheme based on the robustness score to obtain the optimal production scheduling scheme. S5: If a serious equipment malfunction or an urgent order is detected during the execution of the optimal production scheduling plan, dynamic rescheduling is triggered to regenerate the production scheduling plan. The dynamic calibration of the equipment wear coefficient in the equipment-level digital twin model includes: Calculate the predicted deviation between the actual processing force and the predicted processing force. ; When the absolute value of the prediction deviation does not exceed 3% of the theoretical value, the recursive least squares method is used to smoothly update the equipment wear coefficient. When the absolute value of the prediction deviation exceeds 3% but does not exceed 20% of the theoretical value, Kalman filtering is initiated for state estimation; wherein, when the absolute value of the prediction deviation exceeds 8% of the theoretical value, an early warning signal is triggered. When the absolute value of the prediction deviation exceeds 20% of the theoretical value, it is determined to be a serious equipment malfunction, and dynamic production rearrangement is directly triggered.

2. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 1, characterized in that, The equipment-level digital twin model establishes a multi-axis linkage kinematic model for the CNC bending machine, including three degrees of freedom: bending axis, feed axis, and rotation axis; the predicted processing force... The calculation formula is: in, For a moment Predicted processing force, For a moment The equipment wear coefficient, This is based on the theoretical value of bending force under ideal conditions. For the bending angle, The diameter of the reinforcing bar. For the material's yield strength, This represents the model residuals.

3. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 2, characterized in that, The method of smoothing the update of the equipment wear coefficient using recursive least squares includes: The wear coefficient of the equipment is updated using the recursive least squares method with a forgetting factor. The update formula is as follows: in, This is the gain matrix of the recursive least squares algorithm. For a moment Prediction bias; For bending machines with large load fluctuations, the forgetting factor is set to 0.95; for straightening machines with relatively stable operation, the forgetting factor is set to 0.

98.

4. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 3, characterized in that, The initiation of Kalman filtering for state estimation includes: Establish a state-space model, state vectors ,in The rate of change of the wear coefficient; The state equation is: The observation equation is: Wherein, the state transition matrix Observation matrix , For process noise, To measure noise, when the wear coefficient value estimated by the Kalman filter shows a monotonically decreasing trend for three consecutive sampling periods, an early warning signal is triggered to prompt maintenance personnel to check.

5. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 1, characterized in that, The construction of the multi-objective intelligent scheduling model includes: Establish differentiated changeover time matrices for straightening and cutting machines, CNC bending machines, and threading machines respectively. ,in and Representing different steel bar specifications, matrix elements indicate specifications. Switch to Specifications Required changeover time; The optimization objectives are to minimize the total completion time, minimize the total changeover time, and maximize the utilization rate of surplus materials. The constraints of equipment changeover time, process sequence, equipment capacity, and material availability are treated as hard constraints. The multi-objective intelligent scheduling model is solved using a non-dominated sorting genetic algorithm to generate a Pareto optimal solution set, which serves as the initial scheduling solution.

6. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 1, characterized in that, The multi-scenario simulation and deduction of the initial production scheduling plan in the digital twin environment includes: In the digital twin model, multiple abnormal scenarios are set, including at least two of the following: equipment failure and downtime, emergency order insertion, material delays, quality abnormality rework, and staff absence. Simulations were run for each abnormal scenario, and the actual completion time, number of delayed orders, and capacity loss were recorded under each scenario. Calculate the overall cost under each scenario : in, This is the planned completion time under normal circumstances. For the context The actual completion time is as follows. For the context The number of delayed orders Total number of orders For the context Reduced production capacity and lost working hours The total planned production capacity working hours, , , These are the weighting coefficients. = 0.4, =0.3, = 0.3; Calculate production scheduling plan robustness score : in, This represents the total number of abnormal scenarios; The production scheduling scheme with the lowest robustness score is selected as the optimal production scheduling scheme.

7. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 1, characterized in that, It also includes a closed-loop traceability step for quality production scheduling: Assign a globally unique identifier to each steel bar; During the processing, the processing parameters and quality inspection data of each steel bar are collected in real time, and the processing parameters and quality inspection data are bound to the globally unique identifier to form a single-bar quality traceability digital thread. When a quality anomaly occurs, the production batch and processing equipment parameters that caused the anomaly are located in reverse based on the quality traceability digital thread. When the product defect rate of a certain piece of equipment exceeds a preset threshold, the task allocation weight of the equipment in subsequent production scheduling is reduced. For every percentage point the defect rate exceeds the threshold, the weight is reduced by 5%, down to a minimum of 50% of the original weight. After the equipment has been repaired and has continuously processed 50 defect-free products, the task allocation weight will automatically be restored to its original value.

8. The digital twin and dynamic intelligent scheduling method for steel bar processing according to claim 1, characterized in that, It also includes a smart matching step for surplus materials: Three-dimensional point cloud data of steel bar scraps are obtained through three-dimensional scanning equipment, and a three-dimensional digital model of the scraps is constructed. Extract the geometric features of the three-dimensional digital model of the scrap material, wherein the geometric features include at least the length of the straight line segment, the bending angle, and the flatness of the end; The scrap material is classified and graded based on the aforementioned geometric features. The classification and grading include: The first level is a hard match between material and specifications, requiring that the material grade be consistent and the diameter be within the allowable deviation range of the required specifications, wherein the allowable deviation is ±0.5 mm; The second level is shape classification. Leftover material with a bending angle of no more than 5 degrees and an end flatness of no less than 0.9 is classified as straight strip type, with a straightening loss rate of zero; leftover material with a bending angle of more than 5 degrees but no more than 30 degrees is classified as adjustable type, with a straightening loss rate of 2% to 4%; leftover material with a bending angle of more than 30 degrees is classified as curved type. For adjustable straightening scrap, the effective length equal to inventory length Multiply by (1 minus the straightening loss rate) ); The third level is the overall fit score, and the overall fit score... The calculation formula is: in, = 0.6, = 0.25, = 0.15; Length fit The calculation rule is: when hour, = 1; when hour, ;when or hour, = 0; Shape fit Warehouse location convenience The path distance of the automated guided vehicle is calculated and then normalized to the interval between 0 and 1. Based on the classification results, the surplus material with the highest overall suitability score is selected and included in the production material list, and the material inventory data is updated.

9. A digital twin and dynamic intelligent scheduling system for steel bar processing, characterized in that, include: The physical space perception module is deployed on-site in the steel bar processing workshop to collect multi-dimensional data of the physical workshop in real time. The multi-dimensional data includes the operating status parameters of each processing equipment, the processing progress information of work-in-process, material inventory data, and order task information. The equipment-level digital twin modeling and dynamic calibration module, running on the central controller, is used to construct a refined digital twin model of the rebar processing equipment. Based on the deviation between the actual processing force collected in real time and the processing force predicted by the model, a three-level response strategy is used to dynamically calibrate the equipment wear coefficient in the model. The three-level response strategy is as follows: when the absolute value of the predicted deviation does not exceed 3% of the theoretical value, the recursive least squares method is used to smoothly update the equipment wear coefficient; when the absolute value of the predicted deviation exceeds 3% but does not exceed 20% of the theoretical value, Kalman filtering is initiated for state estimation, and when the absolute value of the predicted deviation exceeds 8% of the theoretical value, an early warning signal is triggered; when the absolute value of the predicted deviation exceeds 20% of the theoretical value, it is determined that the equipment is seriously abnormal, and dynamic rescheduling is directly triggered. The multi-objective intelligent scheduling and robust optimization module runs on the central controller. It is used to construct a multi-objective intelligent scheduling model based on the calibrated digital twin model and combined with the differentiated equipment changeover time matrix, generate an initial scheduling plan, calculate the robustness score of the scheduling plan through multi-scenario simulation and deduction, and output the optimal scheduling plan. The dynamic rescheduling module runs on the central controller and is used to regenerate the production schedule using a rolling time-domain optimization strategy when a serious equipment malfunction or an urgent order is detected. The quality scheduling closed-loop traceability module runs on the central controller and is used to assign a globally unique identifier to each steel bar. When a quality abnormality occurs, it reverses the location to the scheduling batch and equipment and feeds the traceability results back to the multi-objective intelligent scheduling and robust optimization module. The intelligent matching module for scrap materials runs on the central controller and is used to construct a three-dimensional digital model of the scrap materials through three-dimensional scanning and to match available scrap materials based on a hierarchical classification strategy. The central controller includes a processor and a memory. The memory stores a computer program. When the processor executes the computer program, it implements the functions of the above-mentioned modules. The central controller also interfaces with the upstream system through a standardized interface to automatically import order information and provide feedback on production scheduling results.

10. The digital twin and dynamic intelligent scheduling system for steel bar processing according to claim 9, characterized in that, The physical space sensing module includes a sensor array and a data acquisition unit: The sensor array includes torque sensors, vibration sensors, temperature sensors, and current sensors installed on each processing device; The data acquisition unit has a data acquisition cycle of 100 milliseconds and transmits the data to the central controller via industrial Ethernet.