A Method for Constructing Digital Twins of Multi-Site Shipbuilding Factories Based on Real-Time 3D Engine

Abstract

This invention provides a method for constructing digital twins of multi-factory shipbuilding plants based on a real-time 3D engine, comprising the following steps: S1: Construction of a rigid-flexible dual-anchoring system for the spatiotemporal reference of the entire process; S2: Granular layering of all-element twin entities and mapping of entity-component-system architecture; S3: Spatiotemporal alignment of multi-source heterogeneous data and engine injection; S4: Hierarchical asynchronous stream loading of twin scenes and customization of real-time rendering pipeline; S5: Rigid-flexible coupled physical simulation and virtual-real closed-loop verification of the entire process. This invention achieves accurate virtual-real mapping and business closed-loop driving of the entire shipbuilding process in multiple factories, and is applicable to the intelligent manufacturing process control scenario of multi-factory collaboration.

Description

Technical Field

[0001] This invention relates to the fields of digital twin technology and intelligent ship manufacturing technology, and in particular to a method for constructing a digital twin of a multi-site shipbuilding plant based on a real-time 3D engine. Background Technology

[0002] As the shipbuilding industry develops towards larger scale, clustering, and intelligence, the construction of large ships generally adopts a multi-factory, specialized division of labor production model. Different factories are responsible for different processes such as steel pretreatment, component forming and processing, section manufacturing, section painting, overall assembly and outfitting, dock loading, and berthing and sea trials. Efficient collaboration across factories and processes has become a core key factor restricting the overall efficiency, cost control, and quality management of shipbuilding. Digital twin technology, as a core supporting technology of intelligent manufacturing systems, can achieve accurate mapping of all elements, processes, and lifecycles between physical shipyards and virtual twin scenarios. It is a core technical means to solve the pain points of collaborative production in multi-factory shipbuilding.

[0003] Existing digital twin construction technologies for multi-site shipbuilding suffer from a series of unavoidable core defects in practical applications. First, the global spatiotemporal reference system is inconsistent. Conventional solutions use a separation of Cartesian coordinates and quaternions to describe rigid body pose transformations, while the deformation of flexible components is solved using a separate finite element model. This leads to a complete disconnect between the rigid body reference and the flexible body reference, resulting in unavoidable cumulative errors in coordinate transformations across multiple site scales. This easily leads to spatiotemporal misalignment and inaccurate entity mapping across site scenarios. Second, the alignment and injection capabilities of multi-source heterogeneous data are insufficient. Shipbuilding scenarios involve multiple heterogeneous data sources, including design, equipment, sensing, business, and environment data. Conventional solutions use simple alignment methods such as timestamp interpolation and hard coordinate transformation, which cannot completely eliminate the spatiotemporal reference bias implicit in the data. Furthermore, the commonly used middleware forwarding data transmission mode suffers from high transmission latency and poor data consistency, failing to achieve bidirectional closed-loop optimization of data and spatiotemporal reference. Third, physical simulation is severely disconnected from actual production conditions. Conventional shipbuilding simulations employ offline calculation methods that separate rigid body dynamics and flexible body finite element analysis, enabling only one-way data transfer and failing to achieve bidirectional force coupling between rigid and flexible bodies. This results in insufficient simulation accuracy and an inability to synchronize real-time on-site measured data for closed-loop model correction, hindering real-time production decision-making. Fourth, the digital twin system is completely separated from production operations. Conventional digital twin solutions use only a real-time 3D engine as a front-end visualization platform, with all production business logic running in external business systems. This prevents deep integration of the digital twin scenario with production operations, hindering the driving role of digital twins in multi-plant, cross-process collaborative production and preventing the core upgrade from visualization to business-driven processes. Furthermore, existing technologies lack a comprehensive algorithm system that deeply integrates the entire process from the underlying spatiotemporal benchmark to the upper-level business logic, failing to address these industry pain points at their architectural root. Summary of the Invention

[0004] This invention provides a method for constructing digital twins of multi-plant shipbuilding factories based on a real-time 3D engine. Addressing the aforementioned core deficiencies of existing multi-plant shipbuilding digital twin technologies, it starts from the underlying spatiotemporal reference architecture to construct a multi-plant shipbuilding digital twin system with a closed-loop process, full element mapping, and full business drive. This systematically solves the core problems in existing technologies, such as inconsistent spatiotemporal references, insufficient alignment accuracy of heterogeneous data, disconnect between physical simulation and on-site working conditions, and separation of twin scenes from production operations.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for constructing digital twins of multi-site shipbuilding plants based on a real-time 3D engine includes the following steps: S1: Construct a rigid-flexible dual-anchored global spatiotemporal benchmark for the entire shipbuilding process in multiple shipyards based on Clifford geometric algebra, and generate a global spatiotemporal benchmark library for multiple shipyards. S2: Based on the global spatiotemporal reference library of multiple shipyards, the twin entities of all shipbuilding elements in multiple shipyards are layered at 5 levels of granularity to complete the entity-component-system architecture mapping of the real-time 3D engine and generate a multi-factory twin entity architecture mapping library. S3: Acquire multi-source heterogeneous data of the entire shipbuilding process in multiple shipyards. Based on the global spatiotemporal benchmark library of multiple shipyards and the full-element twin entity architecture mapping library of multiple shipyards, the spatiotemporal manifold alignment algorithm of Hodge decomposition is used to perform full-link spatiotemporal alignment of multi-source heterogeneous data of the entire shipbuilding process in multiple shipyards. The aligned data is written into the components of the corresponding twin entity through the native injection channel of the real-time 3D engine, generating a running instance of spatiotemporal alignment injection engine for heterogeneous data in multiple shipyards. S4: Combining the multi-plant global spatiotemporal benchmark library, the multi-plant full-element twin entity architecture mapping library, and the multi-plant heterogeneous data spatiotemporal alignment injection engine running instance, construct hierarchical asynchronous stream loading rules bound to business status, customize a unified real-time rendering pipeline across plants, and generate real-time rendering and loading running instances of multi-plant twin scenes. S5: Based on the multi-plant global spatiotemporal benchmark library, the multi-plant full-element twin entity architecture mapping library, the multi-plant heterogeneous data spatiotemporal alignment injection engine running instance, and the multi-plant twin scene real-time rendering and loading running instance, it adopts the rigid-flexible body bidirectional coupling implicit dynamics solution algorithm based on Clifford algebraic tangent space mapping to construct a multi-plant full-process rigid-flexible coupling physical simulation system and generate a multi-plant full-process coupling physical simulation and virtual-real verification closed-loop system; S6: Based on a multi-plant global spatiotemporal benchmark library, a multi-plant full-element twin entity architecture mapping library, a multi-plant heterogeneous data spatiotemporal alignment injection engine running instance, a multi-plant twin scene real-time rendering and loading running instance, and a multi-plant full-process coupled physical simulation and virtual-real verification closed-loop system, the multi-plant cross-process collaborative business logic is built into the entity-component-system architecture of the real-time 3D engine, constructing a full-link business closed-loop system and generating a multi-plant cross-process collaborative twin business logic closed-loop system; S7: Conduct full-link accuracy verification of the multi-plant cross-process collaborative twin business logic closed-loop system, build a full lifecycle self-iterative optimization system, and complete the final optimization of the multi-plant shipbuilding factory digital twin system based on a real-time 3D engine.

[0006] In this specification, during the construction of the rigid-flexible dual-anchored global spatiotemporal reference in S1, the rigid body pose transformation and flexible body deformation solution are unified in the same algebraic space based on Clifford geometric algebra. The rigid body anchoring covers the fixed entities of all shipbuilding areas in the entire shipbuilding process. An independent local coordinate system is established for each shipbuilding area and the transformation parameters to the global coordinate system are calculated. The flexible anchoring covers all deformable components in the entire shipbuilding process. A flexible deformation constitutive model bound to the global coordinate system is established for each type of deformable component. All model nodes are bound to a globally unified spatiotemporal stamp.

[0007] In this manual, the five granularity levels in S2 are arranged from high to low as follows: global plant area level, single plant area workshop and workstation level, equipment and component level, part and operation action level, and parameter and sensor data level. Each level is bound to the corresponding spatiotemporal reference in the multi-plant area global spatiotemporal reference library.

[0008] In this specification, during the entity-component-system architecture mapping process in S2, a spatiotemporal reference core component is set for all twin entities. The spatiotemporal reference core component is an essential component for twin entities to create in the real-time 3D engine. At the same time, a corresponding engine system is established for each granular level, and the global spatiotemporal synchronization system is set as the highest scheduling priority.

[0009] In this manual, the multi-source heterogeneous data of the entire shipbuilding process in S3 includes design data, equipment data, sensor data, business data, and environmental data. Among them, design data is acquired from ship-specific design software, equipment data is acquired from the controllers and acquisition elements of production equipment, sensor data is acquired from on-site measurement equipment, environmental monitoring equipment, and positioning equipment, business data is acquired from production management-related systems, and environmental data is acquired from shipyard meteorological monitoring equipment, official tide release systems, and video surveillance systems.

[0010] In this specification, during the end-to-end spatiotemporal alignment process in S3, for static data, coordinate transformation and global spatiotemporal stamp binding are completed first, and then the data is matched to the corresponding twin entity component. For dynamic real-time data, global real-time spatiotemporal stamp binding and spatial coordinate alignment are completed first, and then the data is matched to the corresponding twin entity component. During the alignment process, spatiotemporal reference deviation data is separated, and the multi-plant global spatiotemporal reference library of S1 is corrected in reverse based on the deviation data. If the data alignment deviation exceeds the preset threshold, an abnormal feedback instruction is generated and sent back to the multi-plant global spatiotemporal reference library, triggering parameter review and correction, and then the alignment operation is re-executed.

[0011] In this manual, the hierarchical asynchronous stream loading rules in S4, which are bound to the business status, correspond one-to-one with the 5-level granularity layering. The global factory-level scene is set as the real-time 3D engine's resident memory content. The loading trigger conditions for other level scenes include both user perspective triggering and business status triggering. As long as either of the two triggering conditions is met, the loading of the corresponding scene will be executed. After the scene is loaded, the corresponding spatiotemporal reference core component is bound synchronously. If there is no business trigger and the user perspective leaves for more than a preset time, automatic unloading will be executed.

[0012] In this specification, during the construction of the rigid-flexible coupling physical simulation system for multiple plant areas and full processes in S5, the rigid body dynamics and flexible body deformation solutions are unified into the same tangent space based on the Clifford algebraic tangent space mapping. The rigid body dynamics solution and the flexible body deformation solution are executed synchronously. The motion state of the rigid body is input into the flexible body deformation solution equation in real time, and the reaction force generated by the flexible deformation is input into the rigid body dynamics solution equation in real time, realizing the bidirectional force coupling solution of the rigid body and the flexible body.

[0013] In this specification, during the construction of the full-link business closed-loop system in S6, an independent engine business logic system is built for each multi-plant cross-process collaborative business scenario. The execution flow of each business logic system is as follows: real-time business data reading, cross-process scheme pre-simulation verification, global spatiotemporal synchronization scheduling, twin entity status update, and business instruction reverse transmission to the on-site production management system, so as to realize the full-link closed-loop drive of the twin system for on-site production operations.

[0014] In summary, the present invention has at least the following beneficial effects: This invention achieves an integrated description of rigid body transformation and flexible deformation through a unified Clifford geometric algebra space, and constructs a global spatiotemporal reference system with rigid-flexible dual anchoring. It fundamentally solves the problems of inconsistent spatiotemporal references and cumulative coordinate transformation errors in multi-factory shipbuilding scenarios with cross-scale and cross-process requirements. It realizes full-process spatiotemporal anchoring of fixed facilities and deformable components throughout the entire factory area, ensuring accurate matching of the virtual and real spaces of the twin system.

[0015] This invention achieves spatiotemporal manifold alignment of multi-source heterogeneous data through Hodge decomposition, eliminating spatiotemporal reference bias components in the data, realizing high-precision spatiotemporal alignment of heterogeneous data and low-latency native engine injection, and establishing a two-way closed-loop optimization mechanism for data alignment and global spatiotemporal reference, ensuring real-time synchronization and accurate matching between twin data and entity models.

[0016] This invention achieves a unified implicit dynamic solution for rigid and flexible bodies through Clifford algebraic tangent space mapping, constructing a real-time physical simulation system with bidirectional force coupling. This solves the problems of traditional offline simulation being disconnected from on-site working conditions and insufficient simulation accuracy. At the same time, the simulation results can back-optimize the global spatiotemporal reference and data alignment rules, forming a self-optimizing closed loop of the simulation system, providing accurate simulation verification and decision support for multi-plant production collaboration.

[0017] This invention breaks through the inherent architecture of traditional digital twins that separate the front-end and back-end. It fully integrates the business logic of cross-process collaboration into the entity-component-system architecture of the real-time 3D engine, realizing the core upgrade of the twin system from visualization to business closed-loop drive. It effectively improves the efficiency of cross-process collaboration in multiple factories, eliminates the waiting time for process connection, and realizes intelligent control of the entire shipbuilding production process by the twin system.

[0018] This invention constructs a hierarchical asynchronous stream loading rule and a customized real-time rendering pipeline that are bound to business status, solving the problems of loading lag in large-scale twin scenes across multiple factories and spatiotemporal inconsistencies in rendering across factories. At the same time, it establishes a full lifecycle self-iterative optimization system to ensure that the accuracy, stability and business adaptability of the twin system can continuously meet the operational needs of the entire shipbuilding production process. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0020] Figure 1 This is a flowchart illustrating the method for constructing a digital twin of a multi-site shipbuilding factory based on a real-time 3D engine, which is involved in this invention.

[0021] Figure 2 This is a schematic diagram illustrating the process of constructing a rigid-flexible dual-anchored global spatiotemporal reference involved in this invention.

[0022] Figure 3 This is a schematic diagram illustrating the process of spatiotemporal alignment of multi-source heterogeneous data and native engine injection involved in this invention.

[0023] Figure 4 This is a schematic diagram of the rigid-flexible bidirectional coupling physical simulation and virtual-real closed-loop verification process involved in this invention. Detailed Implementation

[0024] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0025] The following disclosure provides many different implementations or examples for carrying out different structures of the embodiments of the present invention. To simplify the disclosure of the embodiments of the present invention, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. Furthermore, reference numerals and / or reference letters may be repeated in different examples of the embodiments of the present invention; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various implementations and / or arrangements discussed.

[0026] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] like Figure 1 As shown, this embodiment provides a method for constructing a digital twin of a multi-site shipbuilding factory based on a real-time 3D engine, including: S1: Construction of a rigid-flexible dual-anchoring system for the entire process's spatiotemporal reference The core of a digital twin system is a unified and dynamically adjustable spatiotemporal reference. Multi-plant shipbuilding scenarios suffer from industry pain points such as spatial coordinate mismatch across regions, processes, and scales, time synchronization deviations, and the lack of a unified spatiotemporal anchor for large-sized component deformation. Conventional solutions use Cartesian coordinates and quaternions to describe rigid body transformations separately, while flexible deformation is handled by a separate finite element model. This leads to inherent defects such as inconsistent rigid and flexible body references and large cumulative errors in cross-plant coordinate transformations. This step constructs a unified spatiotemporal reference construction algorithm for multi-plant rigid-flexible bodies based on Clifford geometric algebra, establishing a global spatiotemporal reference system with rigid-flexible dual anchoring covering the entire plant and all processes. The construction process of the rigid-flexible dual anchored global spatiotemporal reference is as follows: Figure 2 As shown.

[0028] The shipbuilding process covers all the following plant units: steel pretreatment and cutting plant, component forming and small assembly plant, section manufacturing and mid-stage assembly plant, section painting plant, final assembly and outfitting plant, dock assembly and joining plant, and quay mooring and sea trial plant. For all plant areas, a rigid body anchoring system is first established, and an independent local Cartesian coordinate system is created for each plant area. The origin of the coordinate system is locked to the permanent measurement benchmark target corresponding to the plant area. The X-axis is along the main production line of the plant area, the Y-axis is perpendicular to the main production line, and the Z-axis is vertically upward. A composite measurement method using a global navigation satellite system, total station, and laser tracker is employed to acquire the 6-DOF pose data of each factory area's reference target in the global coordinate system of shipbuilding engineering. The rigid body transformation parameters from the local coordinate system to the global coordinate system of each factory area are calculated. For each fixed workstation, equipment installation reference, yard positioning point, dock pier, and wharf mooring point within each factory area, the rigid body transformation sub-parameters corresponding to the global coordinate system are calculated one by one. All parameters are bound to a unique factory area number and reference target ID.

[0029] After rigid body anchoring is completed, a flexible anchoring system is constructed. All deformable component units in the entire shipbuilding process are steel plates, profiles, sub-assemblies, mid-assemblies, sections, main sections, and the entire hull. For each type of deformable component, based on material mechanical properties and the deformation laws of the shipbuilding process, a flexible deformation constitutive model bound to a global coordinate system is established. The model covers the complete solution logic for elastic deformation, plastic deformation, welding thermal deformation, and hoisting stress deformation of the component. For all deformation nodes of the flexible deformation constitutive model, a globally unified spatiotemporal stamp binding rule is set, meaning that the deformation data of each node must include the spatial coordinates of the corresponding global coordinate system and a globally unified timestamp, ensuring that the deformation data is completely aligned with the rigid body anchoring reference. Finally, a global spatiotemporal stamp generation rule is established, using the IEEE 1588 precision time protocol to set a unified time synchronization reference for all factory areas, all equipment, and all data sources. The timestamp accuracy is no less than 1 millisecond, ensuring complete synchronization of the time dimension of data throughout the entire process.

[0030] The core of this step is the multi-plant rigid-flexible body unified spatiotemporal benchmark construction algorithm based on Clifford geometric algebra. The complete model construction, training and application process of the algorithm is as follows: The algorithm model construction process first defines the Clifford geometric algebra fundamental operation space, as shown in the following formula: ; in , , is the orthonormal basis vector of 3D Euclidean space, corresponding to the X-axis, Y-axis, and Z-axis directions of the global coordinate system. The inner product operation in Clifford algebra describes the metric relationship of vectors. The outer product operation in Clifford algebra describes the relationship between the span of a vector space and its orientation.

[0031] Based on the aforementioned fundamental computational space, a unified multi-vector representation of the global spatiotemporal reference is defined as follows: ;in This is the global spatiotemporal reference multivector for the k-th plant area. k is the plant area number, ranging from 1 to 7, corresponding to the 7 plant areas. For the time scalar component of the k-th factory area, bind a globally unified timestamp of the IEEE 1588 Precision Time Protocol, in seconds. Let be the position vector component of the k-th plant area reference target in the global coordinate system, which is composed of a linear combination of 3 basis vectors, and the unit is meters. Let be the rotational two-vector component from the local coordinate system to the global coordinate system of the k-th plant area. It is composed of the linear combination of the outer product of two basis vectors and describes the rigid body rotation transformation. Let be the pseudoscalar component of the volume of the k-th plant area, which is composed of the outer product of three basis vectors and describes the spatial orientation and scale relationship.

[0032] Based on a global spatiotemporal reference multi-vector, a coordinate transformation model for rigid body anchoring is constructed, as shown in the following formula: ;in This is a multi-vector representing the position of the m-th fixed entity within the k-th factory area in the global coordinate system. m is the fixed entity number, covering all the aforementioned fixed entities. This is a multi-vector representing the position of the m-th fixed entity within the k-th factory area in the local coordinate system of the factory area. Let be the inverse of the global spatiotemporal reference multivector of the k-th plant area, satisfying 1 is the unit scalar in Clifford algebra.

[0033] Based on the same reference multiple vectors, a flexible anchored deformation constitutive model is constructed, as shown in the following formula: ;in This is a multi-vector representing the deformed position of the p-th deformable node of the n-th deformable component in the global coordinate system after deformation. n is the deformable component number, covering all deformable components. p is the deformable node number of the deformable component, covering all discrete deformable nodes of the component. This is the number of the current plant area where the nth deformable component is located, enabling the flexible deformation to be anchored to the rigid body reference of the corresponding plant area. The Clifford multivector representation of the deformation gradient tensor of the nth deformable member at the pth node includes the solution results for elastic deformation, plastic deformation, welding thermal deformation, and hoisting stress deformation. Let be the initial undeformed position multivector of the p-th node of the n-th deformable component, defined in the local coordinate system of the plant area to which the component belongs.

[0034] The algorithm model training process first involves collecting remeasurement data of the benchmark targets in each factory area, obtaining multi-vectors of the measured positions of fixed entities, and constructing a benchmark error loss function, as shown in the following formula: ;in This is the error loss function for the global spatiotemporal reference. This represents the total number of fixed entities within the k-th factory area. The modulus operation for Clifford multivectors is defined as the square root of the inner product of a multivector and its conjugate. This is a multi-vector representing the measured position of a fixed entity.

[0035] The baseline multivector for each plant area is optimized using the gradient descent method. Minimize the loss function The iteration termination condition is To ensure the positioning accuracy of the rigid body datum is no less than 0.1 mm, after optimizing the rigid body datum, collect measured deformation data of the deformable component and optimize the deformation gradient multivector. This ensures that the prediction accuracy of the flexible deformation model deviates from the measured value by no more than 0.5 mm.

[0036] This algorithm abandons the conventional approach of separating Cartesian coordinates and quaternions to describe rigid body transformations and using separate finite element models for flexible deformations. Instead, it unifies the pose transformation of the rigid body and the deformation solution of the flexible body within the same algebraic space using Clifford geometric algebra. This achieves dual anchoring of the rigid body reference and the flexible deformation, fundamentally solving the industry pain points of cumulative coordinate transformation errors and inconsistent rigid and flexible body references in multi-plant, multi-scale scenarios. The basic data comes from field measurements and compliant data sources. The spatial pose data of the plant reference targets comes from field measurements using global navigation satellite systems, total stations, and laser trackers. The material mechanical property parameters of deformable components come from shipbuilding material factory inspection reports. The global time synchronization reference comes from the IEEE 1588 precision time protocol master clock equipment. The final output is a multi-plant global spatiotemporal reference library, containing reference multi-vector sets for all plants, rigid body coordinate transformation rules, flexible constitutive models of deformable components for all processes, globally unified spatiotemporal stamp generation rules, and unique identifier rules for reference data.

[0037] S2: Granular Hierarchy of Full-Element Twin Entities and Entity-Component-System Architecture Mapping Conventional digital twin systems generally suffer from problems such as separation of entities from spatiotemporal references, disconnection between data and 3D models, and lack of unified hierarchical rules for scene scheduling. This leads to spatiotemporal misalignment and data and model asynchrony when switching between multiple factory scenes. This step takes the global spatiotemporal reference as an essential core element for all twin entities, providing a standardized real-time 3D engine carrier for subsequent heterogeneous data injection, real-time scene rendering, physical simulation calculation, and business logic operation.

[0038] The full-element twin entities of multi-plant shipbuilding encompass all types: spatial entities, product entities, personnel entities, material entities, and environmental entities. Spatial entities include plant buildings, plant roads, material storage yards, docks, wharves, lifting equipment, processing equipment, transfer equipment, and power pipelines. Product entities include steel plates, profiles, parts, sub-assemblies, intermediate assemblies, sections, complete sections, ship parts, outfitting components, pipe fittings, valves, and electromechanical equipment. Personnel entities include cutters, welders, assemblers, painters, crane operators, quality inspectors, dispatchers, and safety management personnel. Material entities include welding materials, paints, gases, fasteners, and auxiliary consumables. Environmental entities include wind, rain, snow, temperature, humidity, visibility, tidal levels, and sunlight.

[0039] Based on a global spatiotemporal reference library, all twin entities are layered into five levels of granularity. Each level is bound to its corresponding spatiotemporal reference, ensuring that all entities in the entire scenario are included in a unified spatiotemporal system. Level L1 is the global plant level, binding the global coordinate system to the rigid body transformation rules of each plant area, covering the overall spatial range of all plant areas. Level L2 is the single plant workshop and workstation level, binding the local coordinate system of the corresponding plant area to the global rigid body transformation sub-parameters, covering workshops, workstations, storage yards, docks, and wharves within a single plant area. Level L3 is the equipment and component level, binding the coordinate system of the corresponding workstation to the global rigid body transformation sub-parameters. Deformable components are simultaneously bound to their corresponding flexible constitutive models, covering single equipment and single product component units. Level L4 is the part and operation action level, binding the coordinate system of the corresponding equipment to the global spatiotemporal stamp, covering single part, single personnel operation action, and single equipment operation action unit. Level L5 is the parameter and sensor data level, binding the global spatiotemporal stamp to the spatial coordinates of the corresponding entity, covering the attribute parameters, real-time sensor data, and business data units of all entities.

[0040] After completing the granularity layering, based on the entity-component-system architecture of the real-time 3D engine, standardized mapping of all twin entities is achieved, with mapping rules being unique and consistent throughout the process. Each twin entity corresponds to a unique entity in the real-time 3D engine, and the entity ID is fully bound to the unique identifier rule of the baseline data. Fixed core components and extended components are set for each entity, among which the spatiotemporal baseline component is an essential core component for all entities. Entities without a spatiotemporal baseline component cannot be created in the engine. The spatiotemporal baseline component contains the corresponding rigid body transformation parameters, flexible constitutive model parameters, and global spatiotemporal stamp. Extended components are fixed according to entity type, namely geometric components, physical components, business attribute components, sensor data components, state machine components, and rendering components. A corresponding engine system is established for each granularity level. The system only performs data reading, writing, and logical operations on the target components of the entity at the corresponding level, namely the global spatiotemporal synchronization system, the factory scene scheduling system, the component deformation control system, the equipment operation control system, the personnel operation tracking system, the material flow system, the environmental simulation system, and the business logic scheduling system. Establish scheduling priority rules for the engine system, with the global spatiotemporal synchronization system having the highest priority. The scheduling of other systems must be based on the baseline data of the global spatiotemporal synchronization system to ensure that all logical operations are completed under a unified spatiotemporal baseline.

[0041] The twin entity list in this step comes from the shipyard's full-element ledger, the ship's design material list, and the factory's process layout drawings. The final output is a multi-factory full-element twin entity entity-component-system architecture mapping library, which includes a 5-level granularity twin entity set, an entity-component binding relationship set, engine system scheduling priority rules, and entity unique identifier and spatiotemporal reference binding rules.

[0042] S3: Spatiotemporal Alignment and Engine Injection of Multi-Source Heterogeneous Data Multi-factory shipbuilding scenarios present core challenges due to diverse data source types, significant format differences, and inconsistent spatiotemporal benchmarks. Conventional solutions employ middleware forwarding for data transmission, achieving data alignment through timestamp interpolation and hard coordinate transformation. This approach suffers from high latency, insufficient alignment accuracy, and the inability to completely eliminate spatiotemporal misalignments, and fails to achieve closed-loop optimization between data and the spatiotemporal benchmark. This step constructs a spatiotemporal manifold alignment algorithm for multi-source heterogeneous data based on Hodge decomposition. All incoming data is first aligned with the spatiotemporal benchmark across the entire link, and then directly written into the component memory of the corresponding entity via the real-time 3D engine's native channel. This achieves low-latency, high-consistency data injection while establishing a closed-loop feedback mechanism for data anomalies. The spatiotemporal alignment and native engine injection process for multi-source heterogeneous data is as follows: Figure 3 As shown.

[0043] The following are the sources of all heterogeneous data from multiple plants to be connected: The first category is design data, sourced from TRIBON, CADDS5, CATIA, and AVEVA Marine shipbuilding design software, including the full ship 3D model, production drawings, bills of materials, welding process documents, and assembly process documents. The second category is equipment data, sourced from programmable logic controllers, encoders, and limit switches of processing equipment, lifting equipment, and transfer equipment, including equipment operating status, operating parameters, location data, and fault alarm data. The third category is sensor data, sourced from laser trackers, total stations, 3D scanners, strain gauges, tilt sensors, temperature and humidity sensors, wind speed and direction sensors, water level sensors, ultra-wideband personnel positioning base stations, and RFID material readers / writers, including measured spatial location data, component deformation data, environmental parameter data, personnel location data, and material location data. The fourth category is business data, sourced from enterprise resource planning systems, manufacturing execution systems, product lifecycle management systems, and warehouse management systems, including production plans, process progress, material inventory, personnel scheduling, quality inspection results, cost data, and scheduling instructions. The fifth category is environmental data, which is obtained from the factory's weather station, the maritime safety administration's tide release system, and the factory's video monitoring system. It includes real-time meteorological data, tidal water level data, and video stream data.

[0044] All data must undergo spatiotemporal alignment before entering the engine injection stage. For all incoming static data, including design data and fixed attribute business data, spatial coordinate information is first extracted from the static data. Coordinate transformation is then performed using a global spatiotemporal reference library to align it to the global coordinate system. A fixed global spatiotemporal stamp is bound to the static data, which represents the data's effective time. The aligned static data is then matched to the corresponding geometric components and business attribute components of the entity in the entity-component-system architecture mapping library, completing the one-to-one binding between entities and data. For all accessed dynamic real-time data, including device data, sensor data, real-time business data, and environmental data, each piece of dynamic data is first stamped with a global real-time spatiotemporal stamp conforming to the S1 rule to ensure complete synchronization of the time dimension with the global benchmark. Then, spatial correlation information is extracted from the dynamic data, and spatial coordinate transformation and alignment are completed through the global spatiotemporal benchmark library. For the deformation measurement data of deformable components, node matching and spatiotemporal alignment with the corresponding flexible constitutive model are completed simultaneously. Finally, the aligned dynamic data is matched to the corresponding entity's sensor data component and state machine component in the entity-component-system architecture mapping library to complete the one-to-one binding of data and components. At the same time, a spatiotemporal alignment anomaly feedback mechanism is established. If the data cannot be spatiotemporally aligned or the alignment deviation exceeds the preset threshold, an anomaly feedback command is immediately generated and sent back to the multi-plant global spatiotemporal benchmark library to trigger the review and correction of the corresponding rigid body transformation parameters or flexible constitutive model. After the correction is completed, the alignment operation is re-executed to form a closed loop.

[0045] After spatiotemporal alignment is completed, the engine's native injection operation is executed. A dedicated data injection plugin is developed based on the real-time 3D engine's native software development kit, establishing a native injection channel at the engine's memory level. This channel is fully compatible with the entity-component-system architecture mapping library, allowing direct data reading and writing to the memory addresses of entity components. For aligned static data, the native injection channel writes it to the engine memory of the corresponding entity component in one go, completing the entity's initialization and creation. For aligned dynamic real-time data, the native injection channel incrementally writes it to the engine memory of the corresponding entity component in real-time, according to the global spatiotemporal stamp time tick, with the writing frequency fully synchronized with the engine's frame update frequency. A data injection verification mechanism is established. After each write operation, the data in the component's memory is read and verified. If the verification fails, a retransmission mechanism is immediately triggered to ensure the accuracy of the data injection.

[0046] The core of this step is a spatiotemporal manifold alignment algorithm for multi-source heterogeneous data based on Hodge decomposition. The algorithm and the spatiotemporal reference algorithm achieve bidirectional interaction. The output of the spatiotemporal reference algorithm directly determines the manifold metric of this algorithm, and the output of this algorithm corrects the spatiotemporal reference of S1 in reverse.

[0047] The algorithm model construction process begins by defining the metric tensor of the spatiotemporal manifold based on the baseline multivector, as shown in the following formula: ; Let be the metric tensor of the spatiotemporal manifold of the k-th plant area. The transpose of the k-th plant area reference multivector satisfies the symmetry of the inner product. A positive interaction between the spatiotemporal reference algorithm and this algorithm is implemented; the output of the spatiotemporal reference algorithm is... It directly determines the manifold metric rule, and thus the core parameters of subsequent decomposition operations.

[0048] Based on the metric tensor, a unified differential representation of multi-source heterogeneous data is defined, as follows: ;in Let q be the 1-differential representation of the q-th data from the s-th data source. s is the data source number, ranging from 1 to 5, corresponding to the aforementioned 5 types of heterogeneous data sources. q is the data entry number within the data source. These are the spatial component coefficients in differential form, corresponding to the spatial correlation information of the data in the directions of the three basis vectors. These are the time component coefficients in differential form, corresponding to the global spatiotemporal stamp information of the data. basis vectors The exterior differential operator. It is a time-varying exterior differential operator.

[0049] The core model of Hodge orthogonal decomposition is constructed as follows: ;in It is an exterior differential operator that describes the spatial and temporal rates of change of the differential form. The co-differential operator is defined as follows: . This is the Hodge star operator, which implements the orthogonal complement mapping of the differential form space. p is the order of the differential form, which is set to 1 here. For the 0-form scalar potential corresponding to the differential form, describe the irrotational components of the data. For the 2-form vector potential corresponding to the differential form, describe the discrete-free components of the data. For the harmonic components corresponding to the differential form, satisfying and , which describes the spatiotemporal reference deviation component of the data.

[0050] Based on the Hodge decomposition results, a spatiotemporally aligned harmonic component elimination model is constructed, as shown in the following formula: ;in To complete the 1-differential form of the qth data from the s-th data source after spatiotemporal alignment, the harmonic component of the spatiotemporal reference bias is eliminated, and the spatiotemporal manifold metric of the spatiotemporal reference algorithm is fully matched.

[0051] A reverse interactive correction model from this algorithm to the spatiotemporal benchmark algorithm is constructed, and the formula is as follows: ;in This is the correction amount for the reference multivector of the k-th plant area. This is the set of all data sources within the k-th factory area. Let be the total number of data entries for the s-th data source. For tensor product operations, the deviations of the harmonic components are mapped to the correction space of the reference multivector. The corrected reference multivector is: The correction result is sent back to S1, completing the reverse interaction closed loop.

[0052] The algorithm model training process first involves collecting standard verification data that has already undergone spatiotemporal alignment as the training set. The standard verification data consists of measured data of benchmark targets with precise global spatiotemporal stamps and global coordinates. An alignment error loss function is then constructed, as shown in the following formula: ;in This is the error loss function for spatiotemporal alignment. This is the 1-differential form of the standard verification data.

[0053] The conjugate gradient method is used to optimize the operator parameters of the Hodge decomposition and minimize the loss function. The iteration termination condition is To ensure that the time deviation of spatiotemporal alignment does not exceed 0.5 milliseconds and the spatial deviation does not exceed 0.1 millimeters, the component coefficient weights in differential form are optimized for different types of heterogeneous data to ensure that the alignment accuracy of all data types meets the requirements.

[0054] This algorithm abandons the heterogeneous data alignment mode of conventional methods that use timestamp interpolation and hard coordinate transformation. Instead, it accurately separates the harmonic components containing spatiotemporal reference deviations in heterogeneous data through Hodge orthogonal decomposition, directly eliminating these deviations to achieve spatiotemporal manifold alignment. This solves the core problems of spatiotemporal misalignment and insufficient alignment accuracy in multi-plant, multi-source heterogeneous data. This algorithm achieves deep bidirectional interaction with the Clifford spatiotemporal reference algorithm. The reference multi-vector directly determines the manifold measurement rules of this algorithm, and the deviation components separated by this algorithm can inversely correct the global spatiotemporal reference, forming a closed-loop optimization between the spatiotemporal reference and data alignment. After alignment in this step, the data is written to the component memory of the corresponding entity through the engine's native injection channel, providing accurate and low-latency aligned data for subsequent scene rendering and physical simulation, while ensuring complete matching between the data and the spatiotemporal reference throughout the entire process. The final output is a running example of the spatiotemporal alignment injection engine for heterogeneous data from multiple plants, including a full-data-source spatiotemporal alignment rule set, a real-time 3D engine native data injection channel, data anomaly feedback and spatiotemporal reference update mechanisms, and data injection verification rules.

[0055] S4: Hierarchical Asynchronous Streaming and Customization of Real-Time Rendering Pipeline for Twin Scenes Multi-factory shipbuilding twin scenarios are characterized by their wide scope, high model accuracy, and large data volume. Conventional solutions, employing a view-only loading mode and a general rendering pipeline, suffer from drawbacks such as engine loading lag, insufficient rendering frame rates, and spatiotemporal inconsistencies across different factory areas, failing to meet the requirements of real-time shipbuilding production twins. This step establishes hierarchical asynchronous streaming loading rules bound to business states based on a 5-level granularity layering and spatiotemporal benchmark. Simultaneously, a customized real-time rendering pipeline adapted to the shipbuilding scenario is implemented to achieve smooth operation and spatiotemporally consistent rendering across the entire scenario, providing a visual runtime platform for subsequent simulation calculations and business logic execution.

[0056] First, a hierarchical asynchronous loading rule is constructed, based on a 5-level granularity layer. This rule binds to both a spatiotemporal baseline and injected real-time business data, abandoning the conventional view-only loading model. For L1-level global factory-level scenarios, the content is set to reside in engine memory and is loaded upon engine startup without being unloaded, maintaining the synchronization of the global factory's spatial framework and spatiotemporal baseline. For L2-level single-factory workshop and workstation-level scenarios, content is loaded and unloaded asynchronously. Two triggering conditions are used: first, the user's view enters the corresponding factory's view frustum; second, the injected business data, such as the corresponding factory's process progress or scheduling instructions, triggers loading. Loading is executed if either condition is met. After loading, the corresponding spatiotemporal baseline component is bound synchronously. When no business triggers occur in the corresponding factory and the user's view leaves for more than a preset time, automatic unloading is performed, releasing engine memory. Level 3 (L3) device and component-level scenes are configured for two-level asynchronous loading and unloading. Loading is triggered only after the corresponding L2-level scene has been loaded. Triggering conditions include two categories: first, a change in the operating state of the corresponding device or an update in the process progress of the corresponding component; second, the user's view focusing on the corresponding device and component. Loading is executed as long as either condition is met. For deformable components, a flexible constitutive model is synchronously invoked during loading. Automatic unloading occurs when the corresponding device and component have no state updates and the user's view leaves for more than a preset time. Level 4 (L4) part and operation action-level scenes are configured for three-level asynchronous loading. Loading is executed only after the corresponding L3-level scene has been loaded and the user's view focusing on the corresponding entity. Unloading occurs immediately after the view leaves. Level 5 (L5) parameter and sensor data-level content is configured for real-time synchronous loading. Real-time writing is performed via the native injection channel only after the corresponding entity has been loaded. Data writing stops synchronously after the entity is unloaded, releasing memory resources.

[0057] After completing the loading rules, a customized unified real-time rendering pipeline across factory areas was developed. This pipeline is entirely based on the programmable rendering pipeline of the real-time 3D engine and is bound to the global spatiotemporal reference and entity rendering components throughout the process. First, a unified lighting calculation model based on the global spatiotemporal reference was established. Based on the real-time time of the global spatiotemporal stamp, the latitude and longitude of the corresponding factory area, and the injected environmental meteorological data, the sun's position, light intensity, and atmospheric scattering effects are calculated in real time, ensuring that the lighting effects of all loaded factory scenes are completely consistent, without any cross-factory lighting misalignment issues. For the detailed rendering requirements of shipbuilding's ultra-large components, a customized combined rendering strategy of hardware instantiation, view frustum culling, and distance attenuation was developed. Hardware instantiation rendering is used for components with repetitive structures, complete culling is performed on models outside the view frustum, and rendering precision is automatically reduced for distant models to ensure a stable rendering frame rate. For deformable components, a customized vertex deformation shader was developed. The shader directly reads the flexible constitutive model parameters within the corresponding entity spatiotemporal reference component, as well as the injected real-time deformation sensor data, driving the deformation rendering of the model vertices in real time, achieving real-time visualization of the deformation effect without additional model update operations. For transparent, metallic, and coated materials within the factory area, custom physically based rendering material spheres were developed to perfectly match the physical properties of steel, paint, glass, and other materials found on the shipbuilding site. Simultaneously, a particle rendering system synchronized with a spatiotemporal baseline was customized for special effects such as welding sparks, paint sprays, and rain / snow weather. A linkage mechanism between the rendering pipeline and the engine system was established; each frame update in the rendering pipeline first reads the baseline data from the global spatiotemporal synchronization system, ensuring that all rendered content is completed under a unified spatiotemporal baseline.

[0058] Finally, a global scene spatiotemporal synchronization mechanism is implemented. Through the global spatiotemporal synchronization system, the spatiotemporal reference of all loaded scenes is verified in real time to ensure that the time and space of scenes, entities, data, and effects in different plant areas are completely synchronized without any spatiotemporal deviations across plant areas. If a synchronization deviation occurs, a correction command is immediately triggered to adjust the parameters of the spatiotemporal reference component of the corresponding entity to ensure synchronization accuracy. The final output is a real-time rendering and asynchronous loading instance of multi-plant twin scenes, including a 5-level granular hierarchical asynchronous stream loading rule, a customized cross-plant unified real-time rendering pipeline, a global scene spatiotemporal synchronization operation mechanism, and scene loading and rendering verification rules, providing a stable visual operation platform for subsequent physical simulation and business collaboration.

[0059] S5: Full-process rigid-flexible coupling physical simulation and virtual-real closed-loop verification Conventional shipbuilding simulations typically employ offline computation, solving rigid body dynamics and flexible body finite element methods separately. This approach only achieves unidirectional data transfer, lacking bidirectional force coupling and suffers from insufficient simulation accuracy, disconnect from actual on-site conditions, and inability to synchronize on-site data in real time. Consequently, it struggles to meet the real-time simulation requirements of multi-plant, cross-process collaboration. This step constructs a rigid-flexible bidirectional coupled implicit dynamics solution algorithm based on Clifford algebraic tangent space mapping. A full-process rigid-flexible coupled physical simulation system is built within a real-time 3D engine. Simultaneously, a real-time closed-loop verification mechanism is established between simulation results and on-site measured data, achieving virtual-real synchronization and bidirectional correction between simulation and on-site conditions. This provides simulation support for subsequent cross-process business collaboration. The rigid-flexible bidirectional coupled physical simulation and virtual-real closed-loop verification process is as follows: Figure 4 As shown.

[0060] The simulation scenarios for all processes in the shipbuilding process are as follows: steel cutting and forming, small and medium assembly and section manufacturing, section transfer and yard operation, dock assembly and closure, and dock mooring and sea trial. Each scenario is bound to a corresponding factory area and spatiotemporal reference. The steel cutting and forming scenario covers simulations of thermal deformation during steel plate cutting, plastic deformation during steel plate bending, and deformation during profile cutting. The small and medium assembly and section manufacturing scenario covers simulations of welding deformation, component assembly accuracy, and stress deformation during section hoisting. The section transfer and yard operation scenario covers simulations of section transportation bump deformation, yard stacking and crushing deformation, rigid body dynamics during gantry crane hoisting, and flatbed truck transfer motion. The dock assembly and closure scenario covers simulations of rigid-flexible coupling deformation during overall section hoisting, docking accuracy at the closure joint, welding shrinkage deformation, and stress on dock piers. The simulation of dock mooring and sea trial processes covers the simulation of the ship's hydrostatic pressure, the stress on the mooring cables, the coupled motion of wind, waves and current, and the stability of the entire ship.

[0061] A rigid-flexible coupled physical simulation system is constructed, entirely based on an extension of the physics engine of a real-time 3D engine. First, a rigid-flexible coupled solver is developed based on the physics engine of the real-time 3D engine. The solver's time step is perfectly synchronized with the global spatiotemporal stamp beat, and its spatial reference is perfectly aligned with the global coordinate system. A rigid body simulation submodule is built to read the rigid body parameters of the corresponding entity's physical components and the injected real-time operating data of the equipment. Real-time rigid body dynamics solutions are performed on rigid entities such as lifting equipment, transport equipment, and fixed facilities. The solution results are written to the corresponding entity's physical components and state machine components in real time, synchronously driving scene updates in the rendering pipeline. A flexible simulation submodule is built to read the flexible constitutive model of the corresponding entity, the material parameters of the corresponding entity's physical components, and the injected real-time welding parameters, stress data, and deformation measurement data. Real-time flexible deformation solutions are performed on deformable components such as steel plates, segments, and assemblies. The solution results are written to the corresponding entity's spatiotemporal reference components and physical components in real time, synchronously driving vertex deformation shader updates in the rendering pipeline. After completing the rigid body and flexible submodule construction, a rigid-flexible coupled solution is executed. For scenarios involving both rigid body motion and flexible deformation, such as hoisting, transportation, and assembly, the solver simultaneously performs rigid body dynamics and flexible deformation solutions. The motion state of the rigid body is input into the flexible deformation solution equation in real time, and the reaction force of the flexible deformation is input into the rigid body dynamics solution equation in real time, achieving bidirectional coupled solution and ensuring that the simulation results match the actual stress state on site. A synchronization mechanism between simulation solution and engine frame update is established, ensuring that every solution operation of the coupled solver is fully synchronized with the real-time 3D engine frame update and global spatiotemporal synchronization system.

[0062] After completing the simulation system construction, a virtual-real closed-loop verification mechanism is implemented to establish a real-time comparison and two-way correction closed loop between simulation results and on-site measured data. First, the injected on-site measured data is read in real time, including the positional measurement data of the laser tracker, the shape and positional measurement data of the 3D scanner, the deformation measurement data of the strain gauges, and the operating status data of the equipment encoder. The measured data is compared with the simulation results in real time to calculate spatial position deviation, deformation deviation, and motion state deviation, generating deviation data. If the deviation value is within a preset threshold range, the deviation data is used as a correction amount to optimize the solution parameters of the corresponding flexible constitutive model and the material parameters of the corresponding physical components in real time, improving the accuracy of subsequent simulations. If the deviation value exceeds the preset threshold, an early warning command is immediately triggered. On the one hand, the early warning information is written into the state machine component of the corresponding entity, and a visual early warning is simultaneously displayed in the S4 twin scene. On the other hand, the deviation data and the early warning command are transmitted back to the corresponding business data source through the native injection channel, triggering the review and adjustment of the on-site production process. An iterative learning mechanism for deviation data is established, and all historical deviation data and correction parameters are stored in the simulation model knowledge base to continuously optimize the solution accuracy and convergence speed of the coupled solver.

[0063] The core of this step is a rigid-flexible body bidirectional coupled implicit dynamics solution algorithm based on Clifford algebraic tangent space mapping. It achieves bidirectional interaction with both the spatiotemporal benchmark construction algorithm and the multi-source heterogeneous data spatiotemporal manifold alignment algorithm. The output of the preceding algorithm directly determines the solution space and boundary conditions of this algorithm, and the output of this algorithm inversely corrects the core model parameters of the two preceding algorithms.

[0064] The algorithm model construction process first involves building a reference multi-vector based on the spatiotemporal reference, and then defining the tangent space mapping rule for unified solution of rigid-flexible bodies, as shown in the following formula: ; For the k-th plant area, a reference multi-vector The tangent space of the special Euclidean group SE(3) is used to uniformly describe the motion of rigid bodies and the deformation of flexible bodies. SE(3) is a special Euclidean group that describes all pose transformations of rigid bodies in three-dimensional space. It is the Lie algebra corresponding to the SE(3) group, which is composed of the angular velocity and linear velocity components of the rigid body. Lie algebra The motion spinor in the equation describes the generalized velocity of a rigid body. This facilitates positive interaction between the spatiotemporal reference construction algorithm and this algorithm. It directly determines the spatial reference for unified solution.

[0065] Based on the tangent space mapping rule, the rigid body dynamics equations in tangent space are constructed as follows: ;in Let be the generalized mass matrix of the rigid body, corresponding to the mass and moment of inertia of the rigid body entity. The rigid body entity covers all the aforementioned rigid bodies. It is the first derivative of the spinor of a rigid body, i.e., the generalized acceleration. Let be the matrix of Coriolis force and centrifugal force for a rigid body. It is the generalized force vector of gravity for a rigid body. The driving generalized force vector of the rigid body is calculated from the aligned real-time operating data of the device output by the multi-source heterogeneous data spatiotemporal manifold alignment algorithm, realizing the positive interaction between the multi-source heterogeneous data spatiotemporal manifold alignment algorithm and this algorithm. It is the generalized force vector of the force exerted by the flexible body on the rigid body, and the core interaction term of the bidirectional coupling between the rigid and flexible bodies.

[0066] Simultaneously construct the dynamic equations of the flexible body in tangent space, as follows: ;in This represents the mass matrix of a flexible body, corresponding to the mass distribution of deformable components. It is the second derivative of the generalized coordinates of the flexible body nodes, i.e., the node acceleration. The stiffness matrix of the flexible body is calculated from the parameters of the flexible constitutive model, realizing the positive interaction between the spatiotemporal reference construction algorithm and this algorithm. is the damping matrix of the flexible body. It is the first derivative of the generalized coordinates of the flexible body node, i.e., the node velocity. These are the generalized coordinates of the flexible body nodes. The generalized force vector of the external load of the flexible body is calculated from the aligned welding parameters, force data, and environmental data, realizing the positive interaction between the multi-source heterogeneous data spatiotemporal manifold alignment algorithm and this algorithm. It is the generalized force vector of the force exerted by the rigid body on the flexible body, and it is the core interaction term for the bidirectional coupling between the rigid body and the flexible body.

[0067] Construct a rigid-flexible body bidirectional coupling interaction model, as shown in the following formula: ; The Jacobian matrix of the contact point between the rigid body and the flexible body describes the mapping relationship between the generalized coordinates of the contact point of the flexible body and the spinor of the rigid body motion, derived from a spatiotemporal reference multivector. The calculation shows that the negative sign indicates that the action and reaction forces are in opposite directions, achieving bidirectional coupling of rigid and flexible body dynamics and eliminating the hysteresis error of conventional unidirectional transmission.

[0068] Based on the bidirectional coupling model, a unified implicit solution model for coupled dynamics is constructed, as shown in the following formula: Where t is the current time step. To determine the time step, it must be completely consistent with the global spacetime stamp tick. (Superscript) For the solution variables in the next time step, implicit trapezoidal integration is used for time discretization to ensure unconditional stability of the solution.

[0069] A reverse interactive correction model is constructed from this algorithm to the spatiotemporal reference construction algorithm, as shown in the following formula: ; This is the deformation gradient multi-vector correction amount for the nth deformable member at the pth node. The generalized coordinates of the nodes obtained by this algorithm. These are the aligned, measured coordinates of the deformed nodes output by the S3 algorithm. The corrected deformation gradient multivector is... The correction result is sent back to S1, completing the reverse interaction closed loop.

[0070] A reverse interactive correction model is constructed from this algorithm to the spatiotemporal manifold alignment algorithm for multi-source heterogeneous data, as shown in the following formula: ;in This is the alignment correction amount for the qth data from the s-th data source. This is the 1-differential form corresponding to the simulation results obtained by this algorithm. Correction amount. The data is then sent back to S3 to optimize the operator parameters of the Hodge decomposition, completing the reverse interactive closed loop.

[0071] The algorithm model training process first involves collecting mechanical test data of standard components as the training set. This standard test data includes the deformation and stress distribution data of the components under known loads. A simulation error loss function is then constructed, as shown in the following formula: ; This is the error loss function for coupled simulation. N is the total number of standard test components. This represents the total number of deformation nodes in the nth standard test component. These are the node deformation data obtained from standard tests.

[0072] Optimization of stiffness matrix of flexible body using Newton-Raphson method Damping matrix Minimize the loss function The iteration termination condition is To ensure that the accuracy of the simulated deformation prediction deviates from the experimental value by no more than 0.5 mm, the time step and integration parameters of the implicit solution are optimized to ensure the stability and real-time performance of the solution, and the solution frequency is fully synchronized with the frame update frequency of the real-time 3D engine.

[0073] This algorithm abandons the conventional shipbuilding simulation model of solving rigid body dynamics and flexible body finite element problems separately and only transmitting data in one direction. It maps the dynamic equations of both rigid and flexible bodies to the Clifford algebraic tangent space, realizing the synchronous implicit solution of bidirectional force coupling between rigid and flexible bodies. This solves the pain points of insufficient simulation accuracy and disconnect from actual on-site conditions in complex scenarios such as hoisting and assembly in conventional simulations. This algorithm achieves a deep integration and closed loop with algorithms S1 and S3. The spatiotemporal reference determines the spatial reference and time step of the solution. The aligned real-time data provides dynamic boundary conditions for the simulation, and the simulation results can inversely correct the data alignment rules of the flexible constitutive model of S1 and S3, realizing continuous self-optimization of the simulation model. The real-time simulation results output by this algorithm can not only drive the real-time visualization update of the digital twin scenario, but also provide pre-simulation verification support for subsequent cross-plant business collaboration, realizing the core upgrade of digital twin from visualization to simulation-driven. The final output is a multi-plant, full-process rigid-flexible coupled physical simulation and virtual-real verification closed-loop system, which includes a full-process coupled simulation solver, virtual-real data comparison and deviation calculation rules, constitutive model and physical parameter closed-loop correction mechanism, simulation early warning and business feedback rules.

[0074] S6: Construction of a Closed Loop for Cross-Process Collaborative Twin Business Logic Conventional digital twin systems generally suffer from an inherent pattern where the engine only provides visual display while business logic runs in an external system. This leads to a disconnect between the twin scenario and actual business operations, preventing the realization of closed-loop business drivers and hindering the realization of the core value of multi-plant, cross-process collaboration. This step fully integrates the business logic of multi-plant, cross-process collaboration into the entity-component-system architecture of the real-time 3D engine. Based on the output results of the entire preceding process, it constructs a closed-loop business logic across the entire chain, achieving a core upgrade of the digital twin system from visual display to business-driven operation.

[0075] The shipbuilding process encompasses all cross-plant and cross-process collaborative business scenarios, namely: material distribution and process connection collaboration from the steel pretreatment and cutting plant to the component forming and small assembly plant; parts delivery and assembly process collaboration from the component forming and small assembly plant to the section manufacturing and mid-assembly plant; section transfer and painting process collaboration from the section manufacturing and mid-assembly plant to the section painting plant; section delivery and assembly process collaboration from the section painting plant to the final assembly and outfitting plant; section transfer and dock loading sequence collaboration from the final assembly and outfitting plant to the dock loading and assembly plant; whole ship delivery and mooring test collaboration from the dock loading and assembly plant to the berthing and sea trial plant; global scheduling collaboration of equipment, personnel, and materials across the entire plant; and global management collaboration of quality, safety, schedule, and cost across the entire plant. Each scenario is bound to a corresponding plant and process node.

[0076] Based on an entity-component-system architecture, an independent engine business logic system is built for each collaborative business scenario. All business logic systems are incorporated into the engine system scheduling system, with scheduling priority lower than the global spatiotemporal synchronization system but higher than the rendering and simulation systems. The execution flow of each business logic system is unified throughout, achieving a closed-loop end-to-end.

[0077] The first step in the execution of the business logic system is data reading. After the business logic system starts, it first reads the real-time production progress, material status, equipment status, personnel status, quality inspection results and other business data of the corresponding process through the native injection channel. At the same time, it reads the state machine component data of the corresponding entity to obtain the execution status of the current business, ensuring that all business operations are carried out based on real-time and aligned accurate data.

[0078] The second step is pre-simulation verification, which calls upon the rigid-flexible coupled physical simulation and virtual-real verification closed-loop system to perform pre-simulation verification on the scheme for connecting processes across plant areas. This includes the feasibility of segmented transfer routes, the safety of hoisting operations, the accuracy matching of the closure joint, and the time matching degree of process connection. The pre-simulation verification results are output to identify risks and problems in the business plan in advance.

[0079] The third step is spatiotemporal synchronization scheduling. Based on the global spatiotemporal benchmark library, spatiotemporal synchronization scheduling is performed on the preceding and subsequent processes across different plant areas. Based on the pre-simulation results and real-time business data, the process plan is optimized and adjusted to ensure that the completion time and material delivery time of the preceding plant area are fully matched with the start time and production plan of the subsequent plant area, eliminating the waiting time for the connection between processes across different plant areas and improving the overall production efficiency.

[0080] The fourth step is entity status update, which writes the execution results of the business logic system into the state machine component and business attribute component of the corresponding entity in real time, and drives the twin scenario to perform visualization update in a synchronous manner, so as to realize the real-time twin mapping of business status and allow managers to intuitively grasp the business execution status of the entire plant and all processes.

[0081] The fifth step is business closed-loop feedback, which involves sending the optimized scheduling instructions, process adjustment plans, and early warning information output by the business logic system back to the corresponding business management systems such as enterprise resource planning system, manufacturing execution system, and warehouse management system through the native injection channel. This drives the adjustment and execution of on-site production operations, and at the same time, the on-site execution results are reintegrated into the twin system to complete the full closed loop of business logic.

[0082] After completing the construction of each business logic system, collaborative scheduling rules for the business logic systems within the engine are established. For business scenarios with upstream and downstream connections, the execution order and data interaction rules of the systems are set to ensure that the output results of the upstream business system serve as the input data for the downstream business system without data gaps. Simultaneously, an exception handling mechanism is established. When an execution exception occurs in a business scenario, it is immediately synchronized to all related business systems, triggering collaborative adjustments to ensure the stable operation of the entire business process. The final output is a multi-plant, cross-process collaborative twin business logic closed-loop system, including a collection of business logic systems for all collaborative scenarios, collaborative scheduling rules for business logic within the engine, a two-way interaction closed-loop mechanism for business systems, and collaborative handling rules for business exceptions.

[0083] S7: End-to-End Accuracy Verification and Self-Iterative Optimization of Twin Systems Building upon the outputs from S1 to S6, this step completes the full-link accuracy verification and full-lifecycle self-iterative optimization system construction of the multi-plant twin system, ensuring that the accuracy, real-time performance, and business effectiveness of the twin system always meet the needs of shipbuilding production, and finally completing the construction of a deliverable and operational twin system.

[0084] First, end-to-end accuracy verification is implemented, divided into three levels. Each level is based on a global spatiotemporal reference library, and the verification results directly correspond to the outputs of previous steps, achieving closed-loop correction throughout the process. The first level is spatial accuracy verification, which uses on-site laser trackers, total stations, and 3D scanners to conduct on-site measurements of benchmark targets, fixed equipment, and product components throughout the entire plant area. The measured spatial coordinates and shape data are obtained, and the measured data are compared with the spatial coordinates and simulation deformation data of the corresponding entities in the twin scene. The accuracy of rigid body coordinate transformation, flexible deformation simulation, and scene spatial positioning is calculated. If the accuracy exceeds the preset threshold, the corresponding correction command is triggered, the multi-plant global spatiotemporal reference library, entity component parameters, and simulation model parameters are updated, and the verification is re-executed until the accuracy meets the standard.

[0085] The second level is real-time verification. Based on global spatiotemporal stamps, end-to-end latency testing is performed on the entire data transmission and processing links. The testing links include on-site data acquisition, spatiotemporal alignment, engine native injection, scene rendering update, simulation solving, business logic operation, and instruction reverse transmission. The end-to-end latency of the entire link is verified to ensure that it meets the operation requirements of the real-time 3D engine and that the end-to-end latency is no more than 30 milliseconds. If the latency exceeds the threshold, the data injection channel, asynchronous loading rules and rendering pipeline, simulation solver and business logic scheduling rules are optimized in a targeted manner, and the verification is re-executed until the real-time performance meets the standard.

[0086] The third level is the verification of the effectiveness of the business closed loop. For all cross-plant collaborative business scenarios, a full-process business closed loop test is conducted one by one to verify whether the scheduling instructions and process adjustment plans output by the twin system are consistent with the pre-simulation results and business expectations after execution on site. This verifies the twin system's control over production progress, quality, safety, and cost, and confirms that the system can effectively improve cross-plant collaboration efficiency, reduce process connection costs, and reduce quality and safety risks. If the business closed loop effect does not meet expectations, the simulation model and business logic system rules are optimized, and the verification is re-executed until the business effectiveness meets the standards.

[0087] After completing accuracy verification, a full lifecycle self-iterative optimization system is constructed. First, a twin system self-iterative optimization engine is established, storing all deviation data, correction parameters, and optimization rules generated during the end-to-end verification process into the system knowledge base. The self-iterative optimization engine automatically learns and analyzes the data in the knowledge base at preset intervals, automatically updating the global spatiotemporal benchmark library (S1), the entity-component-system architecture mapping library (S2), the spatiotemporal alignment rules (S3), the rendering and loading rules (S4), the simulation model parameters (S5), and the business logic rules (S6). A verification mechanism for self-iterative optimization is established. After each automatic update, end-to-end accuracy verification is performed to ensure that the updated system performance and accuracy are not lower than before the update. If the verification fails, the system automatically rolls back to the previous stable version, ensuring system stability.

[0088] Finally, the system was deployed and released, and a multi-site shipbuilding digital twin system that can run independently was generated based on a real-time 3D engine. At the same time, system operation and maintenance rules and access management rules were set up to complete the entire process of building the multi-site shipbuilding digital twin system.

[0089] In some embodiments, the real-time 3D engine uses Unreal Engine 5, which implements hierarchical asynchronous streaming rules based on its built-in WorldPartition system, and uses its Niagara particle system to render welding sparks, painting sprays, and rain and snow weather effects synchronized with the global spatiotemporal reference. It also uses its programmable rendering pipeline (RPR) to customize a unified real-time rendering pipeline across different factory areas. In other embodiments, the real-time 3D engine uses the Unity engine, which implements hierarchical asynchronous streaming rules based on its Addressables system, and uses its customizable script rendering pipeline (SRP) to develop a unified real-time rendering pipeline across different factory areas. In still other embodiments, the real-time 3D engine can be the open-source Godot engine, which uses its GDExtension extension mechanism to develop a rigid-flexible coupling physical simulation system and the engine's native injection channel, ensuring that the technical solution of this invention can be directly adapted to the long-term support versions of mainstream commercial and open-source real-time 3D engines.

[0090] In some embodiments, the dedicated data injection plugin corresponding to the native injection channel of the real-time 3D engine is developed based on the native SDK of the corresponding real-time 3D engine. The plugin has built-in standardized core function interfaces, including at least a spatiotemporal reference data read / write interface, an entity component data write interface, a dynamic data frame synchronization interface, and a business instruction feedback interface. Among them, the spatiotemporal reference data read / write interface is uniquely bound to the global spatiotemporal synchronization system and is only open to system calls with the highest scheduling priority. The interface input parameters include the plant area number, the reference multi-vector correction amount, and the global spatiotemporal stamp. The interface output parameters include the reference data update result and the check code. The entity component data write interface is mapped one-to-one with the multi-plant full-element twin entity architecture mapping library. The interface input parameters include the entity unique ID, the component type identifier, the data frame to be written, and the global spatiotemporal stamp. The interface output parameters include the write result check code and the data consistency verification result. The write frequency of the dynamic data frame synchronization interface is completely synchronized with the frame update frequency of the real-time 3D engine. The business instruction feedback interface is adapted to the standard communication protocol of the on-site production management system. Based on the above interface definitions and the official interface specifications of the corresponding engine SDK, technical personnel in the relevant field can directly complete the code development and functional implementation of the plugin.

[0091] In some embodiments, the preset thresholds for spatiotemporal alignment deviations of multi-source heterogeneous data are set as follows: static design data spatial alignment deviation threshold ≤ 0.2mm, dynamic sensing data spatial alignment deviation threshold ≤ 0.5mm, and dynamic real-time data time alignment deviation threshold ≤ 1ms; the preset durations for automatic unloading of scenarios in hierarchical asynchronous stream loading rules are set as follows: L2 level single-plant workshop and workstation level scenarios with no business triggers and user view leaving the scene are set as follows: L2 level equipment and component level scenarios with no state updates and user view leaving the scene are set as follows: L3 level equipment and component level scenarios with no state updates and user view leaving the scene are set as follows: L4 level part and operation action level scenarios with user view leaving the scene are set as follows: L4 level part and operation action level scenarios are set as follows: L4 level rigid body-flexible coupling physical simulation results and field measured data are set as follows: component deformation simulation value and measured value deviation threshold ≤ 0.8mm, rigid body motion simulation value and measured value deviation threshold ≤ 0.3mm; the deviation threshold for global spatiotemporal reference library correction triggers is set as coordinate transformation cumulative deviation threshold ≤ 0.1mm. The above thresholds can be adaptively adjusted by technical personnel in the relevant field within the corresponding range according to the factory's production scale, component size, and process accuracy requirements, and all threshold settings meet the accuracy design goals specified in the instruction manual.

[0092] The embodiments described above are for illustrative purposes only and are not intended to limit the invention. Therefore, any changes in numerical values ​​or substitutions of equivalent elements should still fall within the scope of this invention.

[0093] The above detailed description will enable those skilled in the art to understand that the present invention can indeed achieve the aforementioned objectives and has complied with the provisions of the Patent Law.

[0094] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention. The above descriptions are merely preferred embodiments of the invention and are not intended to limit the invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.

[0095] It should be noted that the above description of the process is for illustrative purposes only and does not limit the scope of this specification. Those skilled in the art can make various modifications and changes to the process under the guidance of this specification. However, these modifications and changes remain within the scope of this specification.

[0096] The basic concepts have been described above. Obviously, for those skilled in the art who have read this application, the above disclosure is merely illustrative and does not constitute a limitation of this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are suggested in this application, and therefore, such modifications, improvements, and corrections still fall within the spirit and scope of the exemplary embodiments of this application.

[0097] Furthermore, this application uses specific terms to describe its embodiments. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different positions in this specification do not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application can be appropriately combined.

Claims

1. A method for constructing a digital twin of a multi-site shipbuilding factory based on a real-time 3D engine, characterized in that: Includes the following steps: S1: Construct a rigid-flexible dual-anchored global spatiotemporal benchmark for the entire shipbuilding process in multiple shipyards based on Clifford geometric algebra, and generate a global spatiotemporal benchmark library for multiple shipyards. S2: Based on the global spatiotemporal reference library of multiple shipyards, the twin entities of all shipbuilding elements in multiple shipyards are layered at 5 levels of granularity to complete the entity-component-system architecture mapping of the real-time 3D engine and generate a multi-factory twin entity architecture mapping library. S3: Acquire multi-source heterogeneous data of the entire shipbuilding process in multiple shipyards. Based on the global spatiotemporal benchmark library of multiple shipyards and the full-element twin entity architecture mapping library of multiple shipyards, the spatiotemporal manifold alignment algorithm of Hodge decomposition is used to perform full-link spatiotemporal alignment of multi-source heterogeneous data of the entire shipbuilding process in multiple shipyards. The aligned data is written into the components of the corresponding twin entity through the native injection channel of the real-time 3D engine, generating a running instance of spatiotemporal alignment injection engine for heterogeneous data in multiple shipyards. S4: Combining the multi-plant global spatiotemporal benchmark library, the multi-plant full-element twin entity architecture mapping library, and the multi-plant heterogeneous data spatiotemporal alignment injection engine running instance, construct hierarchical asynchronous stream loading rules bound to business status, customize a unified real-time rendering pipeline across plants, and generate real-time rendering and loading running instances of multi-plant twin scenes. S5: Based on a multi-plant global spatiotemporal benchmark library, a multi-plant full-element twin entity architecture mapping library, a multi-plant heterogeneous data spatiotemporal alignment injection engine running instance, and a multi-plant twin scene real-time rendering and loading running instance, it adopts a rigid-flexible body bidirectional coupling implicit dynamics solution algorithm based on Clifford algebraic tangent space mapping to construct a multi-plant full-process rigid-flexible coupling physical simulation system, and generate a multi-plant full-process coupling physical simulation and virtual-real verification closed-loop system.

2. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 1, characterized in that, Also includes: S6: Based on a multi-plant global spatiotemporal benchmark library, a multi-plant full-element twin entity architecture mapping library, a multi-plant heterogeneous data spatiotemporal alignment injection engine running instance, a multi-plant twin scene real-time rendering and loading running instance, and a multi-plant full-process coupled physical simulation and virtual-real verification closed-loop system, the multi-plant cross-process collaborative business logic is built into the entity-component-system architecture of the real-time 3D engine, constructing a full-link business closed-loop system and generating a multi-plant cross-process collaborative twin business logic closed-loop system; S7: Conduct full-link accuracy verification of the multi-plant cross-process collaborative twin business logic closed-loop system, build a full lifecycle self-iterative optimization system, and complete the final optimization of the multi-plant shipbuilding factory digital twin system based on a real-time 3D engine.

3. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 1, characterized in that, In the construction of the rigid-flexible dual-anchored global spatiotemporal reference in S1, the rigid body pose transformation and flexible body deformation solution are unified in the same algebraic space based on Clifford geometric algebra. The rigid body anchoring covers the fixed entities of all shipbuilding areas in the entire shipbuilding process. An independent local coordinate system is established for each shipbuilding area and the transformation parameters to the global coordinate system are calculated. The flexible anchoring covers all deformable components in the entire shipbuilding process. A flexible deformation constitutive model bound to the global coordinate system is established for each type of deformable component. All model nodes are bound to a globally unified spatiotemporal stamp.

4. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 1, characterized in that, In S2, the five levels of granularity are hierarchically arranged from high to low as follows: global plant area level, single plant area workshop and workstation level, equipment and component level, part and operation action level, and parameter and sensor data level. Each level is bound to the corresponding level of spatiotemporal reference in the multi-plant area global spatiotemporal reference library.

5. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 4, characterized in that, In the entity-component-system architecture mapping process in S2, a spatiotemporal reference core component is set for all twin entities. The spatiotemporal reference core component is an essential component for twin entities to create in the real-time 3D engine. At the same time, a corresponding engine system is established for each granular level, and the global spatiotemporal synchronization system is set as the highest scheduling priority.

6. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 1, characterized in that, The multi-source heterogeneous data in S3 covering the entire shipbuilding process across multiple shipyards includes design data, equipment data, sensor data, operational data, and environmental data. Among them, design data is acquired from ship-specific design software, equipment data is acquired from controllers and acquisition elements of production equipment, sensor data is acquired from on-site measurement equipment, environmental monitoring equipment, and positioning equipment, operational data is acquired from production management-related systems, and environmental data is acquired from shipyard meteorological monitoring equipment, official tide release systems, and video surveillance systems.

7. The method for constructing a digital twin of a multi-site shipbuilding factory based on a real-time 3D engine according to claim 1, characterized in that, In the end-to-end spatiotemporal alignment process in S3, for static data, coordinate transformation and global spatiotemporal stamp binding are completed first, and then the components are matched to the corresponding twin entity. For dynamic real-time data, global real-time spatiotemporal stamp binding and spatial coordinate alignment are completed first, and then the components are matched to the corresponding twin entity. During the alignment process, the spatiotemporal reference deviation data is separated, and the multi-plant global spatiotemporal reference library of S1 is corrected in reverse based on the deviation data. If the data alignment deviation exceeds the preset threshold, an abnormal feedback instruction is generated and sent back to the multi-plant global spatiotemporal reference library, triggering parameter review and correction, and then the alignment operation is re-executed.

8. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 4, characterized in that, In S4, the hierarchical asynchronous stream loading rules bound to business status correspond one-to-one with the 5-level granularity layering. The global factory-level scene is set as the real-time 3D engine's resident memory content. The loading trigger conditions for other level scenes include both user perspective triggering and business status triggering. As long as either of the two triggering conditions is met, the loading of the corresponding scene will be executed. After the scene is loaded, the corresponding spatiotemporal reference core component is bound synchronously. If there is no business trigger and the user perspective leaves for more than a preset time, automatic unloading will be executed.

9. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 1, characterized in that, In the construction of the rigid-flexible coupling physical simulation system for multiple plants and all processes in S5, the rigid body dynamics and flexible body deformation solutions are unified into the same tangent space based on the Clifford algebraic tangent space mapping. The rigid body dynamics solution and the flexible body deformation solution are executed synchronously. The motion state of the rigid body is input into the flexible body deformation solution equation in real time, and the reaction force generated by the flexible deformation is input into the rigid body dynamics solution equation in real time, realizing the bidirectional force coupling solution of the rigid body and the flexible body.

10. The real-time three-dimensional engine based multi-yard shipbuilding yard digital twin construction method according to claim 2, wherein, In the process of building the full-link business closed-loop system in S6, an independent engine business logic system is built for each multi-plant cross-process collaborative business scenario. The execution process of each business logic system is as follows: real-time business data reading, cross-process scheme pre-simulation verification, global spatiotemporal synchronization scheduling, twin entity status update, and business instruction reverse transmission to the on-site production management system, so as to realize the full-link closed-loop drive of the twin system for on-site production operations.

Images