A manufacturing plant management system based on a digital twin platform

Abstract

This invention provides a manufacturing factory management system based on a digital twin platform, belonging to the field of factory management technology. It includes: a physical factory layer for real-time collection of equipment operation, process parameters, material consumption, and employee work reports; a digital twin platform layer deployed on a cloud or local server, including a data acquisition and integration module, a model building and management module, and a simulation and prediction module, constructing a high-fidelity, real-time synchronized digital twin and identifying dynamic bottlenecks based on constraint theory; and an application service layer including sub-modules such as intelligent scheduling and dynamic dispatching, human-machine collaborative work assignment, precise quality control, predictive equipment maintenance, and intelligent logistics and warehouse scheduling. This invention achieves real-time mapping and interaction between the physical factory and virtual space through the digital twin, solving problems such as data silos, delayed response, and disconnect between planning and execution in existing systems, thus realizing adaptive optimization and precise control of the production process.

Description

Technical Field

[0001] This invention relates to the field of factory management technology, and in particular to a manufacturing factory management system based on a digital twin platform. Background Technology

[0002] With the rapid development of the Industrial Internet and intelligent manufacturing, modern manufacturing engineering has widely deployed Manufacturing Execution Systems (MES), Enterprise Resource Planning (ERP) systems, and Warehouse Management Systems (WMS). However, after years of practical application and in-depth observation, significant limitations have been found in the existing management system architecture, especially when facing the flexible production demands of multi-variety, small-batch, and short-delivery periods, revealing the following thorny issues:

[0003] Although existing ERP and MES systems are functionally distinct, in practice, there is a serious disconnect between the two. Production orders issued by ERP are often based on static bills of materials (BOM) and fixed lead times, making it impossible to perceive in real time the status of equipment, tool wear, or fluctuations in operator skill on the production line.

[0004] When sudden anomalies occur in the workshop (such as equipment downtime or defective incoming materials), the information needs to go through a long feedback chain of "on-site reporting → team leader coordination → planner manual adjustment → ERP rescheduling". This delay is usually measured in hours or even days, which forces the planning department to rely on experience for extensive scheduling, while the on-site staff are forced to work overtime to catch up, resulting in the common management problem that plans cannot keep up with changes.

[0005] Although many factories have undergone automation upgrades, introducing a large number of CNC machine tools, robots, and AGVs, these devices often come from different manufacturers and have complex communication protocols (such as incompatible OPC UA, Modbus, and Profinet). Existing management systems mostly only collect equipment on / off status or simple production data, often failing to acquire and analyze in-depth process parameters (such as spindle load, vibration, and temperature profiles) in real time. This forces equipment maintenance departments to rely on periodic inspections or reactive repairs, unable to achieve true predictive maintenance. When equipment experiences hidden performance degradation, the management system remains unaware; problems only surface when scrap is generated or equipment is shut down. This "blind men and the elephant" approach to management results in significant hidden costs.

[0006] Most current systems rely on barcodes or RFID to record process flow, which is a "node-based" recording rather than "continuous" monitoring. For example, in heat treatment or chemical synthesis, slight fluctuations in furnace temperature can affect the hardness of the final product, but existing MES systems typically only record the start and end times, failing to accurately link the furnace temperature curve to specific product batches.

[0007] When the quality department discovers defects during the final inspection stage, a large number of defective products have often already been produced. Tracing back at this point is not only time-consuming and labor-intensive, but also makes it difficult to accurately pinpoint the specific process parameter deviations that caused the problem. Existing quality management modules are mostly reactive, lacking real-time, online process control (SPC) early warning capabilities.

[0008] The existing WMS and MES lack sufficient coordination, resulting in material distribution typically adopting a large-volume, low-frequency push-based model. The warehouse delivers large quantities of materials to the production line at once based on work orders, leading to inventory backlog at the production line, occupying a lot of physical space and increasing search time; or conversely, untimely delivery causes production line shutdowns due to material shortages.

[0009] Existing systems struggle to dynamically generate precise material pull instructions based on real-time production pace and consumption rates, resulting in either excess or interrupted internal logistics, hindering the timeliness of lean production.

[0010] Factories typically have multiple parallel management systems: PLM for product design, ERP for financial resources, MES for production processes, WMS for warehousing, and CMMS for equipment maintenance. Under traditional architectures, these systems often lack standardized data.

[0011] When managers need to make a decision (e.g., whether to accept an urgent order), they have to manually log into multiple systems to check capacity load, material availability, equipment health, and staff scheduling, and then make a comprehensive judgment based on experience. This "manual system integration" method is extremely inefficient and prone to errors, causing companies to react slowly and lack agility when facing market fluctuations.

[0012] In summary, while existing manufacturing plant management systems have achieved initial online operations, they are essentially still characterized by "partial automation but insufficient overall intelligence." They lack a "digital twin" capable of mapping the real-time operational status of the physical factory, making it impossible to pre-simulate, optimize, and control the physical production process in the virtual world. Therefore, there is an urgent need for a new management system that can break down data silos, achieve millisecond-level response times, and possess self-learning and adaptive capabilities to address the long-standing pain points plaguing manufacturing managers.

[0013] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0014] The purpose of this invention is to provide a new management system that can break down data silos, achieve millimeter-level response, and possess self-learning and self-adaptive capabilities.

[0015] To achieve the above objectives, the present invention provides the following solution: A manufacturing factory management system based on a digital twin platform includes: The physical factory layer includes multiple production equipment, multiple RFID readers and multiple sensors, which are used to collect real-time data on equipment operating status, process parameters, material consumption, and employee work reports. The digital twin platform layer, deployed on cloud servers or local servers, includes: The data acquisition and integration module is used to receive and process various types of data uploaded by the physical factory layer in real time through a multi-source heterogeneous data communication protocol, and to clean, transform and fuse the data to build a unified data foundation. The model building and management module is used to build the device's three-dimensional geometric model, physical model, behavioral model and rule model based on the unified data foundation, forming a high-fidelity, multi-scale, real-time synchronized digital twin; The simulation and prediction module is used to load the digital twin, perform virtual mapping, real-time simulation and dynamic extrapolation of the production process of the physical factory, and identify dynamic bottlenecks in the production process based on constraint theory to generate bottleneck prediction information. The application service layer interacts with the digital twin platform layer, including: The intelligent scheduling and dynamic dispatching submodule is used to generate optimized production plans and scheduling schemes based on the bottleneck prediction information, material availability status, real-time equipment capacity and order buffer penetration rate, and automatically trigger rescheduling when production anomalies occur. The human-machine collaborative work assignment submodule is used to combine the processes into process groups and assign them to machine positions and employees based on the scheduling scheme, the process cost table and the production line balance rate target, using an improved traversal algorithm based on the average total time of the process, and generate a human-machine flowchart. The quality precision control submodule is used to trace and analyze quality data throughout the product lifecycle, and combined with process parameter monitoring, to achieve online quality judgment and intelligent defect identification. The equipment predictive maintenance submodule is used to perform real-time assessment of the lifecycle values ​​of key equipment components based on the digital twin, and generate preventive maintenance warnings. The intelligent logistics and warehouse scheduling submodule is used to dynamically generate material pull instructions based on real-time production rhythm and schedule unmanned handling equipment to complete material delivery.

[0016] Optionally, the data acquisition and integration module further includes: using a multi-threaded user datagram protocol communication mechanism to process the work reporting data uploaded concurrently by multiple RFID readers; identifying continuous work reporting actions of the same RFID reader through IP address caching and data functionality caching to ensure zero packet loss rate of work reporting data in high-concurrency scenarios; and the multi-source heterogeneous data communication protocol includes OPC UA, Modbus, and Profinet, and supports preprocessing of underlying data through an edge computing gateway, uploading only key feature data to the digital twin platform layer.

[0017] Optionally, the model building and management module further includes a self-learning update unit for the digital twin model. This unit calculates the model deviation by comparing simulation results with actual production data, and uses machine learning algorithms to correct the parameters of the physical and behavioral models online, enabling the digital twin to adapt to equipment aging, performance degradation, and process adjustments in the physical factory. The model deviation is calculated using the root mean square error formula.

[0018] in, RMSE The root mean square error, n The number of sampling points. y i For the first i The actual production data value of each sampling point for i The simulation value of the digital twin model of each sampling point.

[0019] Optionally, the simulation and prediction module specifically includes: a bottleneck prediction engine based on constraint theory. This engine simulates the production process under different order combinations and equipment operating conditions in the digital twin. By analyzing the work-in-process accumulation time and equipment utilization rate of each process, it dynamically identifies drifting bottleneck processes and outputs bottleneck prediction information. The identification of the dynamic bottleneck is based on the real-time blocking degree index of the process, and its calculation formula is:

[0020] in, B i For the first i Real-time congestion level of each process step WIP i For the first i The current work-in-process quantity of each process. T i For the first i The average processing time per piece in each process, C i For the first i The real-time available capacity of each process, when Bi When the preset threshold is exceeded, the process is marked as a dynamic bottleneck.

[0021] Optionally, the intelligent scheduling and dynamic dispatching submodule specifically performs the following steps: First, it filters out scheduleable orders based on the material availability status of sales orders; second, it calculates the load of scheduleable orders on each production line, and, combined with the bottleneck prediction information, allocates orders to the optimal production line with the goal of maximizing production line balance and minimizing order delays; finally, it determines the order production sequence on the same production line based on the order buffer penetration rate, and automatically calculates the order's start and end dates; wherein, the formula for calculating the buffer penetration rate is:

[0022] in, w To buffer permeability, now For the current date, date T represents the delivery date of the sales order, and T represents the factory's standard delivery cycle.

[0023] Optionally, the improved traversal algorithm based on the average total time of processes used in the human-machine collaborative dispatch submodule includes: arranging all processes of the product in the order of the process flow, and pre-setting... m Individual camera positions and N Employees; Calculate the ideal value for the average total time of each process. ,in, T oi For the first i The total processing time for each machine station; a reasonable error range is set around the ideal value. d Within the stated error range, recursively traverse all possible process combinations and employee allocation schemes; calculate the production line balance rate for each scheme. q The scheme with the highest production line balance rate is selected as the final human-machine flowchart output; the formula for calculating the production line balance rate is:

[0024] in, q For production line balance rate, T i For the first i Average total time of each machine station's operation T max is the maximum average total time for the process among all machine positions, and m is the number of machine positions.

[0025] Optionally, the precise quality control submodule specifically includes: an online statistical process control unit, which monitors key process parameters in real time. When parameter fluctuations exceed control limits, it automatically triggers an alarm and links with the digital twin to trace back upstream processes and equipment that caused the parameter anomalies in a virtual environment; and a defect intelligent cutting unit, which, based on machine vision inspection data and big data analysis models, accurately locates defect positions in the production of strip steel or coil products, and automatically generates the optimal finished product cutting plan in conjunction with customer order requirements to isolate defective products. The optimization goal of the cutting plan is to maximize the yield of qualified products.

[0026] Optionally, in the predictive maintenance submodule, the lifecycle value is calculated based on a weighted average of equipment operating time factor, parts usage condition factor, parts replacement factor, and maintenance factor, and the calculation formula is as follows:

[0027] in, L This is the current lifecycle assessment value. L 0 is the initial value for the lifecycle. t To accumulate running time, t max To determine the maximum operating time, U is the usage condition coefficient, R is the replacement frequency coefficient, M is the maintenance frequency coefficient, and α, β, γ, and δ are the weight coefficients of the corresponding factors.

[0028] Optionally, the intelligent logistics and warehousing scheduling submodule specifically includes: a material pull model based on real-time production cycle time, which receives production instructions from the intelligent scheduling and dynamic scheduling submodule and, in conjunction with the material consumption rate at each machine position, calculates the material demand for a future period and automatically generates pull-style delivery tasks; and an unmanned handling equipment scheduling unit, which, based on the delivery task, equipment location, and real-time path conditions, uses a dynamic path planning algorithm to allocate the optimal driving path and task sequence for automated guided vehicles or unmanned overhead cranes to achieve timely material delivery; wherein, the material consumption rate is calculated using the following formula:

[0029] in, v j for j Material consumption rate per machine station n j For the time interval △ t The number of qualified products completed by the machine station.

[0030] Optionally, the system also includes a management cockpit visualization module. This module displays the real-time operating status of the physical factory in a three-dimensional visualization form through the data interfaces of each sub-module of the application service layer. This includes dynamic simulation images of digital twins, work-in-process levels of each production line, equipment health heatmaps, order production progress Gantt charts, and key performance indicator dashboards. It also allows managers to click on equipment in the virtual scene to obtain its detailed real-time data and historical operating records.

[0031] Compared with the prior art, the present invention has the following beneficial effects: 1. By building a unified data foundation and a multi-protocol compatible data acquisition and integration module, the data of ERP, MES, WMS and foundation equipment are deeply integrated, which solves the problems of information fragmentation and inconsistent standards between traditional management systems, and realizes real-time perception and monitoring of the physical factory in an all-round and multi-dimensional way.

[0032] 2. Innovatively, a bottleneck prediction engine based on constraint theory is introduced into the digital twin. By dynamically identifying drifting bottleneck processes through real-time congestion index, and combining material completeness and buffer penetration rate, intelligent scheduling and dynamic rescheduling are carried out, so that the production plan can adapt to the real-time changes of the production site, significantly improving order delivery rate and resource utilization.

[0033] 3. An improved traversal algorithm based on the average total time of each process is adopted for human-machine collaborative work assignment. Within an acceptable error range, the optimal process combination and personnel allocation scheme can be quickly solved. The generated production line balance rate is close to the theoretical optimal value. This effectively solves the problems of uneven resource allocation, work-in-process accumulation and uneven worker workload in the traditional work assignment mode, and significantly improves the overall efficiency of the production line.

[0034] 4. By linking the online statistical process control unit with the digital twin, real-time monitoring of process parameters and anomaly backtracking are achieved; combined with intelligent defect-splitting technology, defects in continuous production processes such as strip steel are accurately located and automatically isolated, transforming the quality management model from post-event traceability to pre-event early warning and in-event control, maximizing the yield of qualified products and significantly reducing quality costs.

[0035] 5. Establish a life cycle assessment model for key equipment components, comprehensively consider multiple factors such as operating time, usage status, maintenance and replacement of parts, calculate the remaining life of components in real time and generate preventive maintenance warnings, which changes the traditional periodic inspection or post-maintenance mode, effectively avoids unplanned downtime and extends the service life of equipment.

[0036] 6. Based on real-time production cycle time and material consumption rate at machine positions, precise material pull instructions are dynamically generated. Combined with dynamic path planning algorithms, unmanned handling equipment is scheduled, realizing the transformation from "push-type" large-volume delivery to "pull-type" just-in-time delivery, effectively reducing line-side inventory and eliminating downtime due to material shortages.

[0037] 7. Through a 3D visualization management dashboard, complex production operation data is presented intuitively in the form of dynamic simulation, heat map, Gantt chart and dashboard. Managers can gain real-time insight into the overall status of the factory, and perform click queries and decision simulations in the virtual environment, which significantly improves the ability to respond quickly to market fluctuations and emergency orders. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention.

[0040] Figure 2 A schematic diagram of a factory production key performance indicator dashboard provided in an embodiment of the present invention.

[0041] Figure 3 This is a schematic diagram of a dynamic simulation of a digital twin provided in an embodiment of the present invention.

[0042] Figure 4 The work-in-process (WIP) level diagram for each production line provided in the embodiments of the present invention.

[0043] Figure 5 A device health heat map provided for an embodiment of the present invention.

[0044] Figure 6 A Gantt chart of order production progress provided for an embodiment of the present invention.

[0045] Figure 7 A Key Performance Indicator (KPI) dashboard is provided for embodiments of the present invention. Detailed Implementation

[0046] 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.

[0047] The purpose of this invention is to provide a new management system that can break down data silos, achieve millimeter-level response, and possess self-learning and self-adaptive capabilities.

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, a specific application example from a garment manufacturing factory will be used to describe the invention in detail below. This embodiment is for illustrative purposes only and is not intended to limit the invention.

[0050] Example 1: Application case of a garment manufacturing plant: This garment factory is a medium-sized factory with three workshops: cutting, sewing, and finishing, totaling eight production lines. It mainly produces jeans and T-shirts. The factory has long faced the following problems: 1. The weekly plans issued by the ERP system do not match the actual production capacity, often leading to overtime work at the end of the month to catch up on work.

[0051] 2. The delivery of materials to the production line in large quantities at one time leads to a mountain of inventory at the production line, and the phenomenon of "difficulty in finding materials" and "production line stoppage due to material shortage" often coexist.

[0052] 3. When defective products are found during the final inspection, it is difficult to trace back to which specific process, which piece of equipment, or which batch of fabric had the problem.

[0053] 4. Repairing equipment such as sewing machines only after they malfunction severely impacts production rhythm.

[0054] 5. Uneven distribution of work processes between skilled and novice workers leads to low production line balance and serious backlog of work-in-process.

[0055] To this end, the factory has adopted the manufacturing factory management system based on a digital twin platform as described in this invention, such as... Figure 1-7 As shown.

[0056] System deployment and data infrastructure construction: 1. Physical factory layer transformation: Vibration and speed sensors were installed on all key equipment such as flat sewing machines and overlock sewing machines.

[0057] RFID readers are deployed at each workstation, and RFID name tags are issued to each employee. RFID tags are affixed to each cut piece / semi-finished product rack.

[0058] High-definition cameras (for machine vision) are installed at the cutting table and at the entrance of each process.

[0059] 2. Example of digital twin platform layer operation: Data Acquisition and Integration: The system collects sensor data from devices via the OPC UA protocol, smart meter data via the Modbus protocol, and processes employee work reports concurrently uploaded by RFID readers at the workstations via multi-threaded UDP communication. During the morning rush hour from 8:00 to 9:00 AM on a certain day, there were 326 work reports. The system achieved zero packet loss and 100% data accuracy thanks to IP address caching and functional caching technologies. The edge computing gateway preprocesses the vibration data, uploading only the feature values ​​after FFT transformation, rather than the original massive waveform data, reducing the amount of data uploaded by 85%.

[0060] Model Building and Management: The system built digital twins for 150 main pieces of equipment in the workshop. This included not only 3D models but also physical models (such as the thermodynamic model of a sewing machine motor) and behavioral models (such as vibration characteristic patterns when a thread breaks). The system initiated a self-learning update unit. After one month of operation, the temperature response model of a certain flatbed sewing machine's spindle motor was compared with actual data. The RMSE value was 2.3℃, exceeding the preset threshold of 1.5℃. The system automatically used machine learning algorithms to correct the model parameters online, reducing the RMSE to 1.1℃ and ensuring the high fidelity of the digital twin.

[0061] Core application process example: Scenario 1: Receiving urgent orders and intelligent scheduling: Scenario 1: Receiving urgent orders and intelligent scheduling: At 11:00 AM, the sales department received an urgent order for 500 new T-shirts, requiring delivery in 4 days.

[0062] 1. Bottleneck Prediction: The simulation and prediction module loads a digital twin to simulate the production process after adding new orders. This is achieved by analyzing the real-time bottleneck indicators of each process. B i It was discovered that the "collar-attaching" process (originally process 6) in the sewing workshop was not working properly in the simulation. WIP The number of pieces increased rapidly from 50 to 120, with equipment utilization approaching 100%, while the downstream bottom edge rolling process equipment remained idle. According to the formula... calculate, BThe value of 6 rose from 1.3 to 3.0, far exceeding the preset threshold of 2.0. The system marked the process as a dynamic bottleneck and generated bottleneck prediction information: the new order will cause the process to be severely blocked on the afternoon of the next day.

[0063] 2. Intelligent Scheduling: The intelligent scheduling submodule first checks the materials (fabric, neckline ribbing, etc.) for new orders to confirm they are complete. Then, based on bottleneck prediction information, the system doesn't simply schedule the order last. Instead, aiming to "maximize production line balance" and "minimize order delays," it proposes an optimization plan: delaying the production of another small order (100 pieces) of a different style by half a day, adjusting the "neckline attachment" process for the new order to the afternoon, and suggesting temporarily transferring two skilled workers from other production lines to support the bottleneck workstation. Simultaneously, based on the buffer penetration rate formula... Calculate the delivery date for this order (4 days later). w The value is 20%, which is considered medium urgency. The system automatically calculates the online date as the current day and the offline date as the afternoon of the third day, allowing for a one-day buffer.

[0064] Scenario 2: Dynamic work assignment and production line balancing for bottleneck processes: In response to the aforementioned bottleneck process of "leading", the human-machine collaborative dispatching submodule began to work.

[0065] Human-Machine Flowchart Generation: The "collar attachment" process for this new T-shirt includes three steps: "repairing the collar flat knitting machine," "making the collar clips," and "attaching the back collar patch," with total operation times of 13.8 seconds, 63.25 seconds, and 25.3 seconds respectively. This workstation has three sewing machines and is planned to be staffed with three employees. An improved traversal algorithm based on the average total time of each process is adopted, with ideal values... Centered on (13.8 + 63.25 + 25.3) / 3 = 34.12 seconds, and setting an error range d = 2 seconds, the algorithm quickly traversed dozens of process combinations and personnel allocation schemes. Ultimately, the scheme with the highest production line balance rate q was selected: Employee A was responsible for "repairing the collar knitting machine + attaching the collar patch" (total time 39.1 seconds), and employees B and C were each responsible for one machine to "make the collar clips" (each 63.25 seconds). Under this scheme, T max The time is 63.25 seconds (after dividing by the number of employees, the actual average time per person is 63.25 seconds). q = (39.1 + 63.25 + 63.25) / (3 × 63.25) = 87.3%. Although it does not reach 100%, it is the optimal value among all feasible solutions, which is much higher than the 65% of traditional random assignment.

[0066] Scenario 3: Real-time quality control and intelligent defect segmentation In the cutting workshop, a new batch of fabric is being cut.

[0067] 1. Online SPC Monitoring: Real-time data from the tension sensor of the cutting machine is uploaded to the quality precision control submodule. When the 50th piece is cut, the tension value suddenly exceeds the control limit. The online statistical process control unit immediately triggers an alarm and links with the digital twin. The system quickly backtracks in the virtual environment and finds that the tension anomaly at that moment highly matches the location of a "weft skew" defect recorded in the upstream "fabric inspection" process for this batch of fabric. The administrator immediately views the fabric image at that location in the virtual scene, confirms the problem, and avoids the subsequent batch scrapping of 450 pieces.

[0068] 2. Intelligent Defect Slitting: For a roll of printed fabric with defects, machine vision detects three consecutive stains. The intelligent defect slitting unit, combined with the customer order (requiring each roll to be cut into 100-meter segments), automatically generates the optimal slitting plan through a big data analysis model: isolating the defective points and dividing the originally continuous roll into three segments: 98 meters + 2 meters (waste) + 100 meters + 100 meters, maximizing the yield of qualified products to 99.5%, while manual slitting typically only achieves 95%.

[0069] Scenario 4: Predictive Maintenance and Smart Logistics 1. Predictive Maintenance: The predictive maintenance submodule continuously monitors the key component of the "up collar" process—the sewing machine rotary hook. According to the formula... The rotary hook's initial lifespan L0 = 10 million stitches, with a cumulative operating time t = 8.5 million stitches. Due to recent intensive use (usage condition coefficient U = 0.8), and two parts replacements (R = 0.2) and one maintenance (M = 0.1), the system's weighting coefficients are preset as α = 0.4, β = 0.3, γ = 0.2, and δ = 0.1. The system calculates in real-time that the current lifespan value L has dropped to 723,000 stitches, below the warning threshold of 800,000 stitches. It immediately generates a preventative maintenance warning, reminding maintenance personnel to replace the rotary hook during the lunch break, thus avoiding unexpected downtime during the afternoon peak production period.

[0070] 2. Intelligent Logistics Pull: Based on real-time production rhythm, the intelligent logistics submodule calculates the material consumption rate of the "feeding" process station. According to the formula... In the past hour, this station has completed 120 items. n j =120), then the consumption speed v j =120 pieces / hour. The system predicts that 240 pre-cut neckline pieces will be needed in the next 2 hours and automatically generates a pull-style delivery task. The scheduling unit plans the optimal route for the AGV, avoiding densely populated areas, and delivers the materials on time after 30 minutes. At this time, the materials at the machine station only have enough for half an hour, realizing JIT delivery and reducing line-side inventory by 70%.

[0071] Scenario 5: Management Cockpit and Decision Support The factory manager opened the visualization module of the management cockpit, and the 3D interface dynamically simulated the operation of the entire factory. The screen he saw included: Equipment health heatmap: The equipment at the "Shangling" workstation is displayed in yellow (warning status), reminding him of the maintenance warning that just occurred.

[0072] Work-in-process (WIP) level on the production line: In the sewing workshop, where work-in-process was piling up, the WIP level has dropped from 1,500 pieces in the morning to 900 pieces.

[0073] Order progress Gantt chart: Newly inserted urgent orders are shown in green, indicating that they are ahead of schedule and are expected to be completed half a day ahead of schedule.

[0074] KPI dashboard: The overall production line balance rate increased from 68% to 82% on the day, and the overall equipment efficiency increased by 15%.

[0075] The factory manager clicks on the virtual equipment at the workstation, and immediately sees the equipment's real-time vibration data, historical operating records, and the upcoming maintenance plan. Based on this, he can quickly make decisions such as whether to approve overtime and how to allocate personnel, reducing the entire process from hours to minutes.

[0076] In summary, this embodiment demonstrates through specific data and application processes that the system of the present invention can effectively solve long-standing pain points in manufacturing factory management, such as planning disconnect, chaotic logistics, difficulty in quality traceability, delayed maintenance, and uneven work assignment. It achieves adaptive optimization and precise control of the production process, significantly improving the factory's production efficiency and resource utilization.

[0077] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0078] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A manufacturing factory management system based on a digital twin platform, characterized in that, include: The physical factory layer includes multiple production equipment, multiple RFID readers and multiple sensors, which are used to collect real-time data on equipment operating status, process parameters, material consumption, and employee work reports. The digital twin platform layer, deployed on cloud servers or local servers, includes: The data acquisition and integration module is used to receive and process various types of data uploaded by the physical factory layer in real time through a multi-source heterogeneous data communication protocol, and to clean, transform and fuse the data to build a unified data foundation. The model building and management module is used to build the device's three-dimensional geometric model, physical model, behavioral model and rule model based on the unified data foundation, forming a high-fidelity, multi-scale, real-time synchronized digital twin; The simulation and prediction module is used to load the digital twin, perform virtual mapping, real-time simulation and dynamic extrapolation of the production process of the physical factory, and identify dynamic bottlenecks in the production process based on constraint theory to generate bottleneck prediction information. The application service layer interacts with the digital twin platform layer, including: The intelligent scheduling and dynamic dispatching submodule is used to generate optimized production plans and scheduling schemes based on the bottleneck prediction information, material availability status, real-time equipment capacity and order buffer penetration rate, and automatically trigger rescheduling when production anomalies occur. The human-machine collaborative work assignment submodule is used to combine the processes into process groups and assign them to machine positions and employees based on the scheduling scheme, the process cost table and the production line balance rate target, using an improved traversal algorithm based on the average total time of the process, and generate a human-machine flowchart. The quality precision control submodule is used to trace and analyze quality data throughout the product lifecycle, and combined with process parameter monitoring, to achieve online quality judgment and intelligent defect identification. The equipment predictive maintenance submodule is used to perform real-time assessment of the lifecycle values ​​of key equipment components based on the digital twin, and generate preventive maintenance warnings. The intelligent logistics and warehouse scheduling submodule is used to dynamically generate material pull instructions based on real-time production rhythm and schedule unmanned handling equipment to complete material delivery.

2. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The data acquisition and integration module further includes: using a multi-threaded user datagram protocol communication mechanism to process the work reporting data uploaded concurrently by multiple RFID readers; identifying continuous work reporting actions of the same RFID reader through IP address caching and data functionality caching to ensure zero packet loss rate of work reporting data in high-concurrency scenarios; and the multi-source heterogeneous data communication protocol includes OPC UA, Modbus, and Profinet, and supports preprocessing of underlying data through an edge computing gateway, uploading only key feature data to the digital twin platform layer.

3. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The model building and management module also includes a self-learning update unit for the digital twin model. This unit calculates the model deviation by comparing simulation results with actual production data, and uses machine learning algorithms to correct the parameters of the physical and behavioral models online, enabling the digital twin to adapt to equipment aging, performance degradation, and process adjustments in the physical factory. The model deviation is calculated using the root mean square error formula. in, RMSE The root mean square error, n The number of sampling points. y i For the first i The actual production data value of each sampling point for i The simulation value of the digital twin model of each sampling point.

4. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The simulation and prediction module specifically includes: a bottleneck prediction engine based on constraint theory. This engine simulates the production process under different order combinations and equipment operating conditions in the digital twin. By analyzing the work-in-process inventory time and equipment utilization rate of each process, it dynamically identifies drifting bottleneck processes and outputs bottleneck prediction information. The identification of the dynamic bottleneck is based on the real-time blocking degree index of the process, and its calculation formula is: in, B i For the first i Real-time congestion level of each process step WIP i For the first i The current work-in-process quantity of each process. T i For the first i The average processing time per piece in each process, C i For the first i The real-time available capacity of each process, when B i When the preset threshold is exceeded, the process is marked as a dynamic bottleneck.

5. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The intelligent scheduling and dynamic dispatching submodule specifically performs the following steps: First, it filters out scheduleable orders based on the material availability status of sales orders; second, it calculates the load of scheduleable orders on each production line, and, combined with the bottleneck prediction information, allocates orders to the optimal production line with the goal of maximizing production line balance and minimizing order delays; finally, it determines the order production sequence on the same production line based on the order buffer penetration rate, and automatically calculates the order start and end dates; wherein, the formula for calculating the buffer penetration rate is: in, w To buffer permeability, now For the current date, date For the delivery date of the sales order, T This is the factory's standard delivery cycle.

6. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The improved traversal algorithm based on the average total time of each process used in the human-machine collaborative dispatch submodule includes: arranging all processes of the product in the order of the technological flow, and pre-setting... m Individual camera positions and N Employees; Calculate the ideal value for the average total time of each process. ,in, T oi For the first i The total processing time for each machine station; a reasonable error range is set around the ideal value. d Within the stated error range, recursively traverse all possible process combinations and employee allocation schemes; calculate the production line balance rate for each scheme. q The scheme with the highest production line balance rate is selected as the final human-machine flowchart output; the formula for calculating the production line balance rate is: in, q For production line balance rate, T i For the first i Average total time of each machine station's operation T max is the maximum average total time for the process among all machine positions, and m is the number of machine positions.

7. A manufacturing plant management system based on a digital twin platform according to claim 1, characterized in that, The precise quality control submodule specifically includes: an online statistical process control unit, which monitors key process parameters in real time. When parameter fluctuations exceed control limits, it automatically triggers an alarm and links with the digital twin to trace back upstream processes and equipment that caused the parameter anomalies in a virtual environment; and a defect intelligent slitting unit, which, based on machine vision inspection data and big data analysis models, accurately locates defects in strip steel or coil product production and, combined with customer order requirements, automatically generates the optimal finished product slitting plan to isolate defective products. The optimization goal of the slitting plan is to maximize the yield of qualified products.

8. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, In the predictive maintenance submodule, the lifecycle value is calculated based on a weighted average of equipment operating time factor, parts usage condition factor, parts replacement factor, and maintenance factor, and the calculation formula is as follows: in, L This is the current lifecycle assessment value. L 0 is the initial value for the lifecycle. t To accumulate running time, t max To determine the maximum operating time, U is the usage condition coefficient, R is the replacement frequency coefficient, M is the maintenance frequency coefficient, and α, β, γ, and δ are the weight coefficients of the corresponding factors.

9. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The intelligent logistics and warehousing scheduling submodule specifically includes: a material pull model based on real-time production cycle time, which receives production instructions from the intelligent scheduling and dynamic scheduling submodule and, in conjunction with the material consumption rate at each machine position, calculates material demand for a future period and automatically generates pull-style delivery tasks; and an unmanned handling equipment scheduling unit, which, based on the delivery task, equipment location, and real-time path conditions, uses a dynamic path planning algorithm to allocate optimal travel paths and task sequences for automated guided vehicles or unmanned overhead cranes to achieve timely material delivery; wherein, the material consumption rate is calculated using the following formula: in, v j for j Material consumption rate per machine station n j For the time interval △ t The number of qualified products completed by the machine station.

10. A manufacturing factory management system based on a digital twin platform according to claim 1, characterized in that, The system also includes a management cockpit visualization module, which displays the real-time operating status of the physical factory in a three-dimensional visualization form through the data interfaces of each sub-module of the application service layer. This includes dynamic simulation images of digital twins, work-in-process levels of each production line, equipment health heatmaps, order production progress Gantt charts, and key performance indicator dashboards. It also allows managers to click on equipment in the virtual scene to obtain its detailed real-time data and historical operating records.

Images