A method for building a parameterized digital twin model of a factory with a computational kernel

Driven by a five-layer parameterized model architecture and a computational kernel, the correlation and collaboration issues of factory-level digital twin models are solved, achieving seamless integration and intelligent decision support throughout the entire lifecycle, and improving model building efficiency and intelligence.

CN123117410APending Publication Date: 2026-11-17SHUXUZHI (GUANGDONG) TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610976972.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2026-03-30
Filing Date
2026-07-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing factory-level digital twin models have weak correlation with the production process, poor coordination, limited functionality, are difficult to build, have long cycles and high costs, and lack intelligent decision support.

Method used

It adopts a five-layer parameterized model architecture, including a data layer, parameter layer, computing kernel layer, model layer and application layer. The computing kernel serves as the central driver, enabling seamless integration and intelligent decision support throughout the entire lifecycle.

Benefits of technology

It enables seamless collaboration throughout the entire factory lifecycle, greatly improves the efficiency of model building and iteration, reduces technical barriers and labor costs, enhances the intelligence and practicality of digital twins, and supports refined management and intelligent decision-making.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117410A_ABST
    Figure CN123117410A_ABST
Patent Text Reader

Abstract

This invention discloses a method for building a parameterized digital twin model of a factory with a computational kernel, belonging to the field of digital twin construction technology. The core of the method lies in constructing a parameterized model architecture integrating a computational kernel, which includes a data layer, a parameter layer, a computational kernel layer, a model layer, and an application layer. By abstracting the physical entities and business rules of the factory into parameter sets and utilizing the computational kernel for logical operations and data-driven processes, the method achieves automatic generation and synchronous mapping of the entire lifecycle model, from layout design, production line planning, logistics simulation to operational debugging. This invention solves the technical problems of existing digital twin technologies, such as limited functionality, weak coordination with the production process, and difficulty in supporting full lifecycle management. It can significantly reduce the difficulty of model building, mitigate design risks, accelerate factory construction, and reduce overall construction and operating costs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of digital twin and industrial simulation technology, specifically to a method for building a factory digital twin parametric model with a computing kernel, applicable to the full lifecycle management of a factory from planning and design, construction and commissioning to production and operation. Background Technology

[0002] Digital twins, as a core technology for realizing cyber-physical systems (CPS), have been widely used in the field of smart manufacturing in recent years. They enable real-time monitoring of states, process simulation, and optimization support for decision-making by constructing virtual mappings of physical entities. However, in the actual construction and application of factory-level digital twins, existing technologies still have the following significant shortcomings: The model has weak correlation with the production process and poor coordination: Traditional 3D modeling and simulation models are mostly "static" or "offline" models, with insufficient data connection depth with real production systems (such as MES and SCADA), making it difficult to reflect the dynamic changes of the production site in real time, resulting in the twin being disconnected from the physical entity.

[0003] Limited functionality and low integration: The layout model and logistics simulation model in the planning stage and the production monitoring model in the operation stage are often independent "information islands" built by different software, with inconsistent data formats and standards, making it impossible to form a unified digital twin that spans the entire life cycle.

[0004] Model building is difficult, time-consuming, and costly: Any changes to the factory (such as production line restructuring or process adjustment) require professionals to manually modify and reprogram the 3D model and simulation logic extensively, resulting in slow response speed and inability to adapt to rapidly changing market demands.

[0005] Lack of intelligent decision support: Most digital twin systems are limited to visualization and monitoring. Their internal models lack powerful computing and reasoning capabilities and cannot perform in-depth analysis, predictive maintenance, or autonomous optimization based on real-time data and production rules, making it difficult to effectively support high-level intelligent decision-making.

[0006] Therefore, there is an urgent need for a method to build a digital twin model that can organically integrate all stages of the factory's entire life cycle, can respond quickly to changes through parameter-driven approaches, and has intelligent computing capabilities. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for building a parameterized digital twin model of a factory with a computing kernel, so as to solve the problems of poor coordination between digital twins and production processes, single function, difficulty in model reconstruction, and inability to support refined management of the entire life cycle in the prior art.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A method for building a parameterized digital twin model of a factory with a computational kernel is presented. Its core is the construction of a five-layer parameterized model architecture, with the computational kernel layer acting as the "brain" for central drive. This five-layer architecture includes: Data layer: As the foundation, it integrates static data such as factory BIM data, CAD drawings, and equipment attributes, as well as dynamic real-time data such as production orders, equipment status, and material information obtained from MES, SCADA, and IoT platforms.

[0009] Parameter layer: As the driving source, it defines a structured parameter system, including global parameters that affect the whole (such as factory size and production cycle time) and local parameters that describe specific objects (such as the power of a machine tool or the speed of a conveyor belt). All parameters can be accessed and modified by the computation kernel.

[0010] The computational kernel layer, as the core innovation, integrates logic operations, simulation algorithms, rule engines, and data processing modules. It is responsible for interpreting parameters, executing business logic, running simulation calculations, and coordinating data flow and state synchronization between the data layer and the model layer.

[0011] Model layer: As the presentation layer, it is dynamically generated and updated by the computational kernel based on parameters. It is not a pre-made static model, but includes: Parametric 3D geometric model (factory building and equipment layout) Parametric process simulation model (machining and assembly process) Parametric production line simulation model (equipment linkage, cycle time analysis) Parametric factory simulation model (entire plant material and personnel flow) Parametric digital twin model (a virtual entity that is synchronized with the physical factory in real time) Application layer: As the interactive interface, it provides dedicated functional modules for users at different stages (planning engineers, operation and maintenance personnel, and managers), such as layout designers, simulation analyzers, real-time monitoring panels, and decision support dashboards.

[0012] The method specifically includes a parameterized setup process for key stages throughout the factory's entire lifecycle: Layout design phase: Users only need to input or modify global layout parameters (such as length, width, height, and area division), and the calculation kernel will call the geometry engine and rule library to automatically generate or update the corresponding 3D factory layout model and verify the rationality of the design in real time.

[0013] Production line and logistics planning stage: Users set equipment parameters (type, quantity, processing time) and logistics parameters (path, carrier, frequency). The computing kernel drives the discrete event simulation engine to automatically build production line simulation logic, perform capacity verification, bottleneck analysis and logistics optimization, and directly output simulation results.

[0014] Factory operation, commissioning, and visualization phase: The computing kernel establishes a real-time data connection with the on-site IoT gateway, mapping sensor readings and equipment status to the numerical changes of corresponding parameters in the model layer. This enables real-time and accurate driving of equipment animation, material movement, and status indication in the 3D scene, achieving "what you see is what you get" commissioning and monitoring.

[0015] Factory operation and management phase: Supports "simulation and twin synchronous mapping". Managers can perform "what-if" analysis at the application layer. For example, if order scheduling parameters are modified, the computing kernel will immediately perform a rapid simulation in the twin model to predict the production status in the future. Conversely, when the twin model detects low equipment efficiency, the rule engine can provide reverse suggestions to optimize process parameters.

[0016] Compared with the prior art, the present invention has the following significant advantages: Full lifecycle coverage and high degree of collaboration: Through a unified parameterized architecture and computing kernel, planning, simulation, operation and maintenance are seamlessly connected, breaking down data and functional barriers and realizing a true full lifecycle digital thread.

[0017] Significantly improves model building and iteration efficiency: By adopting parameter-driven approach, changes to the factory can be automatically and quickly reflected in all relevant models simply by adjusting the corresponding parameters, reducing model reconstruction work from the "day / week" level to the "minute / hour" level, greatly lowering the technical threshold and labor costs.

[0018] Effectively avoid early design risks: Rapid parametric simulation and verification of multiple planning schemes in a virtual environment can identify potential problems in layout, production capacity, logistics, etc. in advance, reducing rework and cost waste during the physical construction phase.

[0019] Enhancing the intelligence and practicality of digital twins: The built-in computing kernel transforms digital twin models from passive visual objects into proactive intelligent agents with analytical, computational, predictive, and optimization capabilities, providing strong support for refined factory management, energy conservation and consumption reduction, and intelligent decision-making.

[0020] Reduce overall construction and operation costs: This approach achieves "one model for the whole process," avoiding the huge expenses of purchasing and integrating multiple independent software systems at different stages, while continuously creating value for the factory by optimizing operational efficiency. Attached Figure Description

[0021] Figure 1 The following is an architectural diagram of a factory digital twin parameterized model with a computing kernel as described in an embodiment of the present invention.

[0022] Figure 2 The flowchart of the construction and mapping process of the factory digital twin parameterized model throughout its entire lifecycle, as described in this embodiment of the invention.

[0023] Figure 3 : A schematic diagram of the automatic generation process of parametric three-dimensional geometric models according to an embodiment of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Example

[0025] Take the construction of a digital twin project for a new energy vehicle battery factory as an example.

[0026] System initialization and architecture setup: The system platform of this invention is deployed, constructing a five-layer architecture. Static data such as factory BIM design drawings and equipment manuals are imported into the data layer.

[0027] In the parameter layer, define the following global parameters: Total capacity = 10 GWh / year, Plant area = 50,000 m², Production cycle time = 120 seconds / piece. Define the following local parameters: Coating machine speed = 10 m / min, Assembly robot model = XX123, AGV quantity = 15 units.

[0028] At the computational kernel layer, configure the corresponding geometry generation algorithm and discrete event simulation engine, and set business rules, such as "if the oven temperature exceeds the set value by 5%, trigger an alarm".

[0029] Layout and production line planning stage (corresponding to steps 2 and 3 of claim 1): In the application layer's "layout designer," the planning engineer modified the factory area to 55,000 m² and adjusted the workshop's length-to-width ratio to 2:1. Upon receiving the parameter changes, the computational kernel layer immediately invoked the geometry engine to automatically generate a new 3D factory model based on building codes (such as column load-bearing capacity and fire escape requirements), and completed the initial equipment placement.

[0030] Subsequently, engineers optimized the production cycle time parameter from 120 seconds to 110 seconds in the "production line simulator." The computational kernel layer drove the simulation engine to rerun the production line simulation based on the new cycle time, and quickly output a report showing that the originally planned 15 AGVs could not meet the material delivery needs, indicating a bottleneck. Based on this, the engineers modified the AGV quantity parameter to 18 units, and the simulation results showed that the bottleneck was eliminated. The entire process required no manual adjustment of the 3D scene or rewriting of the simulation script.

[0031] Operation and monitoring phase (corresponding to steps 4 and 5 of claim 1): After the factory is built and put into operation, the data layer receives production orders from MES and equipment status data from the IoT platform in real time.

[0032] The computational kernel layer continuously monitors this data. When it receives real-time data such as "Coating machine_actual speed = 9.5 m / min", it immediately updates the coating machine_speed parameter in the parameter layer and drives the running speed of the 3D coating machine model in the model layer to slow down synchronously. On the 3D visualization interface, the administrator can intuitively see the abnormal operation of the equipment.

[0033] Simultaneously, the rule engine in the computational kernel layer is triggered, executing pre-set diagnostic logic and, combined with historical data, pushing a warning message to the decision dashboard in the application layer: "Coating machine speed decreased by 5%, predicted remaining service life is less than 30 days, preventive maintenance is recommended." This exemplifies the synchronous mapping between simulation and twin: real-time twin drives visualization, while rule-based simulation prediction guides future actions.

[0034] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention. Instruction manual illustrations

[0035] To more intuitively illustrate the technical solution of the present invention, a detailed description will be provided below in conjunction with the accompanying drawings and embodiments. These drawings are schematic diagrams of embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0036] Figure 1 The following is an architectural diagram of a factory digital twin parameterized model with a computing kernel as described in an embodiment of the present invention.

[0037] Figure 2 The flowchart of the construction and mapping process of the factory digital twin parameterized model throughout its entire lifecycle, as described in this embodiment of the invention.

[0038] Figure 3 : A schematic diagram of the automatic generation process of parametric three-dimensional geometric models according to an embodiment of the present invention. Detailed Implementation

[0039] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment takes a digital twin project for building a new energy vehicle battery factory as an example.

[0040] Step S1-1: Construct the data layer After system deployment, the first step is to build the data layer. This layer serves as the foundation of the entire system, integrating data from different sources and in different formats through API gateways and data adapters.

[0041] Static data import: The factory's BIM (Building Information Modeling) files, CAD drawings provided by equipment suppliers, and equipment technical parameter tables (such as dimensions, power, and processing range) are imported into the system through parsing tools to form a "static basic database".

[0042] Establish dynamic data channels: Configure data interfaces with upper-level information systems.

[0043] It interfaces with the MES (Manufacturing Execution System) to obtain production orders, process routes, and bills of materials (BOM).

[0044] It interfaces with SCADA (Supervisory Control and Data Acquisition) and IoT platforms to collect real-time data on equipment status (running, shutdown, fault), sensor readings (temperature, pressure, vibration), AGV position, material inventory, etc., forming a "dynamic real-time database".

[0045] Step S1-2: Define the parameter layer The parameter layer is the "source" that drives everything. We define a structured parameter system for the battery factory.

[0046] Global parameters: Top-level parameters that affect the entire plant.

[0047] Total capacity = 10 GWh / year The factory area covers 50,000 m² Target production cycle time = 120 seconds / piece Energy budget = 5000 MWh / year Local parameters: Parameters belonging to a specific area, production line, or equipment.

[0048] Coating workshop parameter set: Coating machine_model="CC-2024" Coating machine - Maximum speed = 15 m / min Coating machine - Set speed = 10 m / min (initial value) Oven Rated Temperature = 180 °C Assembly line parameter set: Number of robots = 8 Tightening gun torque = 25 N·m Logistics parameter set: AGV Quantity = 15 units AGV Rated Speed ​​= 1.5 m / s Material buffer zone_capacity = 200 Step S1-3: Configure the computation kernel layer This is the "brain" of the invention, which we have configured with three core modules: Logical operation unit: responsible for performing basic mathematical operations and geometric calculations (such as coordinate transformation and collision detection).

[0049] Rules engine: Loads predefined business rules. For example: Rule 1: IF "Coating Machine_Actual Speed" < "Coating Machine_Set Speed" * 0.9 THEN Trigger "Equipment Inefficiency" alarm.

[0050] Rule 2: IF "Current Inventory" < "Safety Stock" THEN Call the "Generate Purchase Order" service.

[0051] Simulation algorithm library: It includes a discrete event simulator (for simulating production line cycle time and logistics), a dynamics simulator (for robot motion simulation), and a fluid / thermodynamics simulator (for simulating ovens and cooling systems).

[0052] Step S1-4: Initialize the model layer and application layer The model layer is initially empty, waiting for the computation kernel to drive it.

[0053] The application layer has developed interfaces for multiple functional modules, such as "3D visualization view", "parameter configuration panel", "simulation report dashboard" and "real-time monitoring screen".

[0054] Step S2-1: Parametric 3D layout generation (corresponding to claim 2, see...) Figure 3 ) User actions: The planning engineer modifies global parameters in the "Layout Designer" at the application layer. The factory area is adjusted from 50,000 m² to 55,000 m², and the workshop aspect ratio is changed from the default 1.5:1 to 2:1.

[0055] Kernel operations: Changes in the parameter layer are captured by the computation kernel layer.

[0056] The logic unit calls the geometric constraint solver, which performs automatic calculations based on a pre-loaded library of building codes (such as the requirements for passage width and column spacing in the "Code for Fire Protection Design of Buildings") and a library of equipment dimensions.

[0057] The solver first determines the outer contour of the main plant, and then automatically divides the functional areas such as coating, stacking, and assembly according to the production cycle and the number of equipment, and plans the main passage and fire passage.

[0058] Model Generation: The computational kernel outputs the calculation results to the model layer, and the parametric 3D geometric model is automatically updated: a new factory building model with an aspect ratio of 2:1 is generated, the internal areas are clearly divided, and the initial automated layout of the equipment is completed. Engineers can rotate and examine the model in real time in the 3D view to check the rationality of the layout.

[0059] Step S2-2: Parametric production line and logistics simulation (corresponding to claim 3) User action: An engineer wants to optimize production capacity, so he optimizes the global parameter "target production cycle time" from 120 seconds to 110 seconds.

[0060] Kernel operations: The discrete event simulator in the computational kernel layer is triggered. It reads all relevant local parameters (equipment speed, quantity, loading and unloading time, etc.) from the parameter layer and reconstructs the simulation logic based on the new cycle time.

[0061] The simulator virtually runs the entire battery production process: from feeding, coating, baking to assembly and testing.

[0062] After the simulation is completed, the computational kernel analyzes the simulation data and generates a simulation analysis report.

[0063] Results feedback and optimization: The report shows that when the cycle time is increased to 110 seconds, the existing 15 AGVs cannot transport the battery cells from the stacking machine to the assembly line in time, causing the assembly line to wait and becoming a system bottleneck.

[0064] Based on the report, the engineer changed the "AGV_Quantity" in the logistics parameter set from 15 to 18.

[0065] The computing kernel ran the simulation again, and the new report confirmed that the bottleneck had been eliminated and predicted that the capacity could be increased by 8.3%.

[0066] The value of this stage lies in accurately identifying logistics planning deficiencies through parameter adjustments and rapid simulations before physical construction begins, thus avoiding costly production line modifications later on.

[0067] Step S4-1: Parametric Digital Twin and Real-time Visualization Data synchronization: After the factory goes into production, the data layer receives data from SCADA in real time: Coating machine_actual speed = 9.5 m / min.

[0068] Kernel mapping: Upon receiving this data, the data processing unit in the kernel layer immediately updates the value of the corresponding "Coating Machine_Set Speed" parameter in the parameter layer (from 10 to 9.5).

[0069] This parameter change event is sent to the model layer.

[0070] Model Update: Upon receiving an event, the parametric digital twin model in the model layer immediately drives the coating machine equipment model in the 3D scene, causing its animation speed to be reduced to 95% synchronously. On the 3D visualization screen in the application layer, the administrator can see the abnormal speed reduction of the equipment in real time and intuitively.

[0071] Step S5-1: Simulation and twin synchronization mapping and intelligent decision-making (corresponding to claims 4 and 5) Positive drive (What-if analysis): User action: The sales department receives an urgent order. The production manager enters a new "simulated order quantity" parameter in the "production scheduling simulation" interface at the application layer.

[0072] Kernel computation: The computation kernel layer starts with the current real factory state (from twin data), calls the discrete event simulator, and quickly simulates the production situation for the next 24 hours based on the new order parameters.

[0073] Output: The simulation report predicted that this order would cause an overload of a critical piece of equipment. Based on this, the manager adjusted the production schedule, avoiding potential production disruptions.

[0074] Reverse-driven (optimization and early warning): Kernel autonomous calculation: Following step S4-1, when the "coating machine_actual speed" continues to be lower than the set value, the rule engine is triggered.

[0075] Rule execution: The rule engine checks historical maintenance data and performance curves and executes diagnostic logic.

[0076] Intelligent feedback: The computing kernel pushes an advanced warning message to the administrator through the application layer decision dashboard: "Based on the performance degradation model analysis, the remaining service life of coating machine 01 is predicted to be less than 30 days, with a confidence level of 92%. It is recommended to arrange preventive maintenance in the next maintenance window (this Friday), and the expected replacement part is the drive belt." Parameter optimization suggestion: At the same time, the system may also automatically generate a parameter optimization suggestion, such as "It is recommended to temporarily fine-tune the 'Coating Machine_Set Speed' to 9.5 m / min to maintain stable quality until maintenance is completed".

[0077] Through the above detailed embodiments, it is clearly demonstrated how the present invention, through the core design of parameter-driven and computational kernel, organically integrates all aspects of the entire life cycle of a factory into a unified digital twin, achieving a leap from passive visualization to proactive intelligent decision-making.

Claims

1. A method for constructing a parameterized digital twin model of a factory with a computational kernel, characterized in that, Includes the following steps: Step 1: Construct a parametric model architecture for a factory digital twin, the architecture including: The data layer is used to store the factory's static basic data (BIM, product / equipment 3D models, CAD layout, etc.) and dynamic operational data (MES orders, processing parameters, equipment quantity, process routes, etc.). The parameter layer is used to define and store global parameters (capacity, cycle time, area, etc.) and local parameter sets (equipment speed, robot model, number of AGVs, etc.) that drive the digital twin model. The computation kernel layer integrates a logic operation unit, a rule engine, and a data processing unit, which is used to perform simulation calculations, logical reasoning, and data mapping based on the parameters of the parameter layer and the preset business logic. The model layer, driven by the computational kernel layer, generates and updates various parameters, including parametric 3D geometric models, parametric process simulation models, parametric production line simulation models, and parametric factory simulation models. The application layer provides human-machine interfaces and functional modules (layout planner, simulation analyzer, monitoring screen, etc.) for all stages of the factory's life cycle. Step 2: During the factory layout design phase, input global layout parameters into the parameter layer. Based on the parameters and preset layout rules, the computation kernel layer automatically generates and renders the corresponding parametric 3D geometric model in the model layer and stores it in the data layer, which can be called as model library data. Step 3: In the production line and logistics planning stage, input equipment, process and logistics parameters into the parameter layer. The computation kernel layer calls the built-in simulation algorithm to drive the model layer to generate a parameterized process simulation model (joint node granularity) and a parameterized production line simulation model (covering factors such as production line, logistics, and personnel), and perform simulation verification. Step 4: During the factory operation debugging and visualization stage, the computing kernel layer will perform bidirectional mapping between the real-time collected production data and the parameter layer. The digital twin model can quickly simulate the future operation of the factory, discover factory operation risks in advance, and realize the synchronous operation and three-dimensional visualization of the parameterized digital twin model. Step 5: In the factory operation and management phase, control commands are issued or operating parameters are modified through the application layer. The parameter layer receives the changes and triggers the calculation kernel layer to recalculate, thereby updating the simulation state of the model layer, guiding the physical factory to optimize and adjust, and realizing synchronous mapping and iterative optimization between simulation and twin. The method according to claim 1, characterized in that, In step 2, the automatic generation of the parametric 3D geometric model is as follows: The computational kernel layer automatically generates a 3D model of the factory buildings, workshop layout, and equipment foundation by solving geometric constraints based on the input global parameters such as floor area, number of floors, column spacing, and floor height, or global parameters extracted from the CAD detailed drawings of the standard template, combined with the predefined building code library and equipment size library. The method according to claim 1, characterized in that, In step 3, the generation of the parameterized process simulation model and the parameterized production line simulation model is as follows: The computational kernel layer automatically constructs the production line operation logic and logistics path based on the input process flow, equipment processing cycle, loading and unloading time, failure rate process parameters, AGV speed, and path node logistics parameters, and outputs a simulation analysis report on capacity, bottleneck, and utilization rate. The method according to claim 1, characterized in that, The computational kernel layer has bidirectional driving capability: in forward driving, modifications to the parameter layer drive updates to the model layer; in reverse driving, abnormal data captured by the model layer during simulation or twin operation is used by the rule engine of the computational kernel layer to optimize the parameter settings of the parameter layer. The method according to claim 1, characterized in that, The synchronous mapping in step 4 refers to: the operation kernel layer establishing an association between the real-time database and the key parameters in the parameter layer. When the real-time data changes, the corresponding entities in the 3D model (such as equipment status and material position) are immediately driven to change synchronously through the event triggering mechanism, and the playback and traceability of historical status are supported.