A data integration management method and system for a digital twin factory of an automobile

Abstract

The application discloses a kind of data integration management methods and systems for automobile digital twin factory, it is related to data integration technical field, including, through industrial internet of things equipment and edge computing node, collect full-element real-time data stream and carry out pre-processing, utilize knowledge graph to construct semantic mapping rule, output multimodal data set;Build multiscale digital twin model, generate plant state matrix, the multiscale digital twin model includes geometric layer, physical layer and metabolic layer;Constitute quadratic unconstrained binary optimization model, obtain scheduling scheme by quantum annealing algorithm solution;Drive metabolic layer to carry out energy demand prediction, trigger dynamic adjustment based on model predictive control, output real-time regulation instruction;Through digital thread synchronization real-time regulation instruction to physical equipment, refresh multiscale digital twin model state.The application realizes the accurate mapping of digital twin model by fusing geometric space calibration, physical mechanism simulation and dynamic metabolic analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data integration, and in particular to a data integration management method and system for a digital twin factory of an automobile. BACKGROUND

[0002] The prior art mainly realizes device data acquisition based on an industrial Internet of Things, constructs a virtual mapping in combination with three-dimensional modeling and discrete event simulation technology, and adopts a traditional optimization algorithm (such as a genetic algorithm) for production scheduling. The system architecture is designed in multiple layers, real-time data are processed by edge computing nodes, simulation analysis is performed on a cloud platform, and virtual-real interaction is realized by means of digital threads. In terms of energy management, a static energy consumption model and multi-modal data fusion technology are used for device calibration and energy consumption prediction, and containerized deployment provides a basis for high-concurrency computing.

[0003] However, the prior art still has limitations. The semantic alignment and dynamic fusion efficiency of multi-source heterogeneous data are insufficient, resulting in high delay in updating the twin model and difficulty in meeting real-time requirements. The calculation complexity of the traditional scheduling algorithm rises too fast when dealing with multi-objective optimization problems, and the dynamic adjustment capability is limited. Energy metabolism analysis and production scheduling lack coordination, and the energy efficiency optimization effect is insufficient. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a data integration management method for a digital twin factory of an automobile to solve the problem of insufficient coordination between energy metabolism analysis and production scheduling and insufficient energy efficiency optimization effect.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a data integration management method for a digital twin factory of an automobile, which comprises collecting full-factor real-time data streams and performing preprocessing through industrial Internet of Things devices and edge computing nodes, constructing semantic mapping rules using a knowledge graph, and outputting a multi-modal data set;

[0008] A multi-scale digital twin model is constructed based on the multi-modal data set, a factory state matrix is generated, and the multi-scale digital twin model comprises a geometric layer, a physical layer and a metabolism layer;

[0009] A quadratic unconstrained binary optimization model is constructed according to the factory state matrix, and a scheduling scheme is obtained by solving using a quantum annealing algorithm;

[0010] In combination with the scheduling scheme, energy demand prediction is driven in the metabolism layer, triggering dynamic adjustment based on model predictive control, and outputting real-time control instructions;

[0011] Real-time control instructions to physical devices are refreshed by digital thread synchronization to refresh the multi-scale digital twin model state.

[0012] As a preferred scheme of the data integration management method for the automobile digital twin factory, the geometry layer refers to importing a three-dimensional model of the production line and dynamically calibrating the spatial pose of the equipment by combining laser point clouds; the physical layer refers to embedding a heat conduction partial differential equation and discrete event simulation logic to simulate material flow and process rhythm; and the metabolism layer refers to establishing an energy metabolism network model, quantifying energy demand by material flow analysis, and predicting energy demand by a dynamic Bayesian network.

[0013] As a preferred scheme of the data integration management method for the automobile digital twin factory, the knowledge graph uses a resource description framework triple to represent entity relationships.

[0014] As a preferred scheme of the data integration management method for the automobile digital twin factory, the quantum bit coding mode of the quadratic unconstrained binary optimization model is that each order is assigned a quantum chain, the chain length is equal to the number of optional workstations, and device conflict constraints are realized through cross-coupling terms.

[0015] As a preferred scheme of the data integration management method for the automobile digital twin factory, the dynamic adjustment based on model predictive control includes,

[0016] The process start time window is adjusted based on the time-of-use signal of electricity prices to transfer loads.

[0017] The outlet pressure value of the air compressor is adjusted by a PID controller to optimize energy efficiency.

[0018] The energy storage battery is discharged during peak load periods of the power grid and charged during valley periods for energy storage scheduling.

[0019] As a preferred scheme of the data integration management method for the automobile digital twin factory, the refreshing of the multi-scale digital twin model state includes dynamically rendering an energy consumption heat map in the HSV color space, mapping energy types in the hue dimension, mapping energy consumption intensity thresholds in the saturation dimension, and superimposing real-time control trajectories in a three-dimensional twin scene.

[0020] As a preferred scheme of the data integration management method for the automobile digital twin factory, the multi-scale digital twin model and the quadratic unconstrained binary optimization model are deployed in a container cluster to support parallel computing including energy metabolism network model simulation and quantum annealing solution.

[0021] In a second aspect, the present application provides a data integration management system for a digital twin factory of an automobile, comprising a data acquisition module, a digital twin module, a quantum annealing module, an energy prediction module and an instruction synchronization module.

[0022] The data acquisition module is configured to acquire full-factor real-time data streams and perform preprocessing through an industrial Internet of Things device and an edge computing node, construct semantic mapping rules using a knowledge graph, and output a multi-modal data set.

[0023] The digital twin module is configured to construct a multi-scale digital twin model based on the multi-modal data set, generate a factory state matrix, and the multi-scale digital twin model comprises a geometric layer, a physical layer and a metabolic layer.

[0024] The quantum annealing module is configured to construct a quadratic unconstrained binary optimization model according to the factory state matrix, and obtain a production scheduling scheme by solving the model using a quantum annealing algorithm.

[0025] The energy prediction module is configured to drive the metabolic layer to perform energy demand prediction in combination with the production scheduling scheme, trigger dynamic adjustment based on model predictive control, and output real-time control instructions.

[0026] The instruction synchronization module is configured to synchronize the real-time control instructions to physical devices through digital threads, and refresh the state of the multi-scale digital twin model.

[0027] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the data integration management method for a digital twin factory of an automobile according to the first aspect of the present application is implemented.

[0028] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, any step of the data integration management method for a digital twin factory of an automobile according to the first aspect of the present application is implemented.

[0029] The present application has the beneficial effects that: the edge side knowledge graph is used to realize real-time semantic alignment of multi-source heterogeneous data, reduce twin model update delay, and improve real-time performance of virtual-real mapping; quantum chain coding is used to convert high-dimensional scheduling problems into a solvable form of quantum annealing, break through the computational complexity bottleneck of traditional algorithms in multi-objective optimization, and enhance dynamic scheduling response capability; through the collaborative optimization mechanism of metabolic layer dynamic prediction and model predictive control, closed-loop control of energy demand analysis, time-sharing scheduling and energy efficiency regulation is realized, and collaborative optimization of energy consumption cost and production efficiency is achieved; combined with digital thread synchronization and multi-dimensional visualization technology, while ensuring the time consistency of virtual and real systems, the device pose calibration accuracy and the visualization analysis capability of energy flow trajectory are improved, forming a closed-loop optimization system from data fusion, simulation deduction to dynamic regulation. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0031] Fig. 1 The figure is a data integration management system architecture diagram for the automobile digital twin factory in embodiment 1.

[0032] Fig. 2 The figure is a multi-scale digital twin model construction flowchart in embodiment 1.

[0033] Fig. 3 The figure is a quantum annealing scheduling algorithm flowchart in embodiment 1.

[0034] Fig. 4 The figure is a dynamic regulation and synchronization flowchart in embodiment 1. DETAILED DESCRIPTION

[0035] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0036] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0037] Second, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, characteristic, or combination of features and / or characteristics described herein that can be included in at least one implementation of the present application. The various appearances of "in one embodiment" or "an embodiment" in the specification are not necessarily all referring to the same embodiment.

[0038] Embodiment 1, Reference Figs. 1-4 The embodiment provides a data integration management method for a digital twin factory of an automobile, and comprises the following steps:

[0039] S1: Collect full-factor real-time data streams through industrial Internet of Things devices and edge computing nodes, and perform preprocessing, utilize a knowledge graph to construct semantic mapping rules, and output a multi-modal data set.

[0040] Specifically, the following steps are included:

[0041] S1.1: Collect full-factor real-time data streams through industrial Internet of Things devices and edge computing nodes.

[0042] Specifically, the industrial Internet of Things devices refer to laser scanners, sensors and RFID readers.

[0043] The edge computing nodes refer to embedded device platforms.

[0044] The full-factor real-time data streams include real-time geometric data, physical data and metabolic data.

[0045] Furthermore, laser point cloud scanning device space poses are used to control point cloud data density, and geometric data is obtained.

[0046] The temperature, pressure and rotating speed of the running equipment are collected, the process beat is recorded through discrete event simulation logic, the logistics flow state is collected, and the physical data is obtained.

[0047] Current, voltage, gas flow, environmental temperature and humidity and carbon dioxide concentration are collected as metabolic data.

[0048] S1.2: Preprocess the full-factor real-time data streams.

[0049] Specifically, noise is filtered based on a sliding window algorithm for data cleaning, missing data is processed through linear interpolation, data of different dimensions is normalized through Min-Max normalization, time sequence characteristics are extracted through FFT transformation, and redundant data is compressed using the LZ4 algorithm, so that the compression ratio is greater than or equal to 4:1, and the transmission bandwidth occupancy rate is reduced.

[0050] S1.3: Adopt RDF triples to drive semantic mapping through a knowledge graph.

[0051] Specifically, the RDF triple (subject-predicate-object) is used to represent the association relationship of equipment, process and energy, the mapping rule is set by using SPARQL query language, the semantic level is unified, and the RDF triple is stored in Neo4j graph database.

[0052] The heterogeneous data conflicts are solved by the ontology mapping algorithm, and the automatic matching of, for example, equipment coding (OPC UA tag and MES ID) is realized.

[0053] S1.4: Integrate and output the multi-modal data set.

[0054] Specifically, the multi-modal data set is generated by integrating the pre-processed full-factor real-time data stream and the knowledge graph.

[0055] The multi-modal data set includes structured data, unstructured data and semantic metadata. The structured data is stored in a time series database (such as InfluxDB) and is indexed by equipment ID and timestamp. The unstructured data is stored in an object storage (such as MinIO) and is associated by metadata tags. The semantic metadata is stored in a graph database (such as Neo4j) and supports SPARQL queries.

[0056] Further, the structured data refers to a parameter set in JSON format. The unstructured data refers to laser radar point cloud in pcd format, which is associated with structured data through space-time tags. The semantic metadata refers to knowledge graph triplets.

[0057] A standardized interface is provided through a REST API to support on-demand calling by downstream, and the interface response format is JSON-LD (compliant with RDF specification).

[0058] Preferably, laser point cloud, sensors and RFID readers are used to obtain geometric, physical and metabolic data, and sliding window algorithm, FFT transform and LZ4 compression technology are used to effectively improve the collection quality and reduce the transmission bandwidth occupation. Semantic mapping is constructed through a knowledge graph, RDF triplets are used to represent the relationship between equipment, process and energy, and SPARQL rules and ontology mapping algorithm are used to realize automatic matching and fusion of heterogeneous data. Multi-modal data is integrated, including structured data stored in a time series database, unstructured data stored in an object storage, and semantic metadata stored in a graph database, and a standardized interface is provided through a REST API to support flexible queries in JSON-LD format.

[0059] S2: Construct a multi-scale digital twin model based on the multi-modal data set to generate a plant state matrix, the multi-scale digital twin model including a geometric layer, a physical layer and a metabolic layer.

[0060] Specifically, the following steps are included:

[0061] S2.1: Construct geometry layer and dynamic calibration based on multi-modal dataset.

[0062] Specifically, import the production line three-dimensional CAD file based on BIM standard (format IFC or glTF), obtain the physical structure, equipment layout and geometric characteristics of the production line, including the three-dimensional pose of the production line and the three-dimensional vertex coordinates of the production line.

[0063] Real-time match point cloud data and production line three-dimensional pose by Iterative Closest Point algorithm (ICP), the objective function expression is:

[0064]

[0065] Where R is the ICP algorithm objective function, N is the number of matching point pairs, n is the index of the matching point pair, U is the rigid transformation matrix (including rotation and translation), p n is the three-dimensional vertex coordinates of the production line of the matching point pair n, q n is the corresponding point of the point cloud of the matching point pair n.

[0066] Obtain the equipment pose matrix based on the ICP algorithm objective function.

[0067] S2.2: Physical layer simulation based on multi-modal dataset.

[0068] Specifically, simulate heat conduction and discrete events. Solve the three-dimensional unsteady heat conduction equation to obtain the temperature distribution state vector at each time step, the expression is:

[0069]

[0070] Where ρ is the material density, c is the specific heat capacity, T is the temperature, t is the time, is the partial derivative symbol, is the vector differential operator, k is the thermal conductivity, Q source is the heat source term (such as welding current loss).

[0071] Discretize the grid resolution by finite element method (FEM) to ≤5mm.

[0072] Further, use discrete event simulation (DES) to construct state transition logic, use event scheduling algorithm to maintain priority queue (minimum heap implementation), and obtain process state vector.

[0073] S2.3: Metabolic layer network modeling and prediction based on multi-modal dataset.

[0074] Specifically, construct a directed graph, where the directed graph nodes represent energy conversion equipment and the directed graph edges represent energy flow, and the mass balance equation is constrained, the expression is:

[0075]

[0076] where V is the set of directed graph nodes, F ij is the flow from node i to node j, F ji is the flow from node j to node i, i is a node of the directed graph, j is another node of the directed graph, S i is the sink item of node i. The state transition probability is predicted by time-sliced dynamic Bayesian network (DBN), and the energy demand matrix is obtained, expressed as:

[0077]

[0078] where P is the state transition probability, X t is the energy demand state at time t, X t-1 is the energy demand state at time t-1, M is the number of state variables, m is the index of the number of state variables, is the m-th specific state variable in X t at time t, is the set of parent nodes.

[0079] S2.4: Generate the factory state matrix.

[0080] Specifically, the geometric layer outputs the device pose matrix, the physical layer outputs the process state vector, and the metabolic layer outputs the energy demand matrix.

[0081] The outputs of the multi-scale digital twin model are aligned according to the time window to construct the factory state matrix The factory state matrix S contains the device pose matrix, the process state vector and the energy demand matrix, where the row index H of the factory state matrix corresponds to the device / process number, and the column index W is the timestamp sequence.

[0082] Preferably, in the geometric layer, three-dimensional dynamic calibration is performed using BIM standards and ICP algorithm to ensure high-precision matching of device pose and improve the accuracy of spatial modeling. In the physical layer, finite element method (FEM) is used to simulate heat conduction process, and discrete event simulation (DES) is used to optimize process scheduling, realizing fine simulation of factory environment and process flow. In the metabolic layer, energy flow is modeled based on directed graph, and dynamic Bayesian network (DBN) is used to predict energy demand state, enhancing the intelligence of factory energy management. The outputs of the multi-scale model are integrated to construct the factory state matrix, aligning the time dimension information, and realizing unified representation of device, process and energy state.

[0083] S3: According to the factory state matrix, a quadratic unconstrained binary optimization model is constructed, and a scheduling scheme is obtained by quantum annealing algorithm.

[0084] Specifically, the following steps are included:

[0085] S3.1: Perform production scheduling problem modeling and quadratic unconstrained binary optimization model conversion.

[0086] Specifically, based on the factory state matrix, the production scheduling optimization objectives are set, including minimizing total production time, maximizing equipment utilization, and adding a penalty conflict constraint.

[0087] Obtain order information from the database, and minimize the total production time based on the order information, with the expression being:

[0088]

[0089] where f1 is the function of minimizing total production time, G is the number of orders, g is the order number index, t g is the processing time of order g.

[0090] Maximize equipment utilization, with the expression being:

[0091]

[0092] where f2 is the function of maximizing equipment utilization, D is the number of devices, d is the device number index, T work,d is the effective working time of device d, T total is the total working time of the device.

[0093] Penalty conflict constraint: if the orders are assigned to the same device and overlap in time, a penalty term is applied.

[0094] Further, the multi-objective optimization problem is converted into a quadratic unconstrained binary optimization model (QUBO), with the expression being:

[0095] B1 = αf1 + βf2 + γP conflict ;

[0096] where B1 is the quadratic unconstrained binary optimization model, α is the weight coefficient of minimizing total production time (in this example, it can be taken as 0.4), β is the weight coefficient of maximizing equipment utilization (in this example, it can be taken as 0.4), γ is the weight coefficient of the penalty conflict constraint (in this example, it can be taken as 0.2), satisfying α + β + γ = 1, P conflict is the penalty term.

[0097] S3.2: Perform quantum bit encoding.

[0098] It should be noted that each order corresponds to a quantum chain, and the chain length is equal to the number of available workstations. One-hot encoding is used to constrain each order to only select one workstation.

[0099] The device conflict is represented by cross-coupling, and a coupling term is added when multiple orders are assigned to the same device. The time overlap rate is calculated by the overlap function to constrain the device conflict.

[0100] S3.3: Construct the Hamiltonian.

[0101] Specifically, the quadratic unconstrained binary optimization model is converted into an Ising model Hamiltonian suitable for quantum annealing, expressed as:

[0102]

[0103] where B2 is the total Hamiltonian, is the Pauli operator acting on quantum bit u, is the Pauli operator acting on quantum bit v, h u is the local magnetic field, J uv is the quantum bit interaction strength, and z is the spin z component operator identifier.

[0104] S3.4: Configure annealing parameters and demapping.

[0105] It should be understood that in this example, the annealing time is set to 20 μs, the annealing path uses reverse annealing, the initial temperature is 100 mK, the final temperature is 10 mK, the D-Wave quantum annealing device is used, the embedding algorithm uses the Chimera topology structure, the reading frequency is 1000 times, and the lowest energy solution is taken.

[0106] The bit string is decoded, and the constraint satisfaction is verified. If there is a conflict, a greedy local search is started for post-processing.

[0107] S3.5: Output the scheduling scheme.

[0108] It should be understood that the scheduling scheme contains order ID, station ID, start time and end time in each row.

[0109] Preferably, a trade-off between minimizing total production time and maximizing device utilization is made, while a penalty term is added to avoid order conflicts and improve the rationality of production scheduling. One-hot quantum bit encoding is used to ensure that each order is uniquely assigned to a station, and the cross-coupling mechanism is used to effectively handle device conflicts, making the scheduling scheme more reasonable. By constructing the Ising model Hamiltonian, the optimization problem is mapped to quantum annealing hardware, and the D-Wave device is used for solving, so that the algorithm can efficiently search for the global optimal solution, especially suitable for large-scale and complex scheduling problems. Through the optimization configuration of reverse annealing path, temperature control, etc., combined with the post-processing strategy (greedy local search), the effectiveness and stability of the solution are ensured, thereby improving the execution quality of the scheduling scheme and optimizing the factory production efficiency.

[0110] S4: Combined with the production scheduling scheme, drive the metabolic layer to predict energy demand, trigger dynamic adjustment based on model predictive control, and output real-time regulation instructions.

[0111] Specifically, the following steps are included:

[0112] S4.1: Energy demand prediction modeling.

[0113] Specifically, based on the workstation allocation and process schedule in the production scheduling scheme, the energy metabolic network model in the metabolic layer is driven. This network associates order processing parameters (such as welding current, painting duration) with device energy consumption characteristics (such as air compressor power curve), and calculates the real-time energy demand of each process through material flow analysis.

[0114] A time-sliced dynamic Bayesian network (DBN) is used, with process start time, device load rate, and environmental temperature and humidity (from metabolic layer sensor data) in the production scheduling scheme as input variables to predict energy demand fluctuations within the next hour. The network nodes include: parent nodes: process type, device operating status, environmental temperature; child nodes: electricity demand, gas instantaneous flow, compressed air consumption. The state transition probability is trained through historical data.

[0115] S4.2: Model predictive control (MPC) optimization.

[0116] Specifically, according to the time-of-use electricity price signal (such as 1.2 yuan / kWh during peak hours and 0.4 yuan / kWh during valley hours), high-energy-consumption processes (such as electrophoretic coating) are adjusted to the valley price period for time window optimization, while time buffers are inserted into non-critical path processes (such as material handling) to provide flexibility for load transfer, ensuring that the total production cycle remains unchanged.

[0117] According to the predicted compressed air demand curve, the PID controller dynamically adjusts the air compressor outlet pressure set value. For the welding workshop cooling system, a feedforward-feedback composite control is used. The environmental temperature is used as the feedforward signal to adjust the cooling water flow in real time.

[0118] S4.3: Generate real-time regulation instructions.

[0119] Specifically, the regulation instructions are packaged as JSON format messages, including the device ID and workstation number of the control object, pressure set value adjustment, process start and stop, and timestamp.

[0120] Preferably, a dynamic Bayesian network is used to build an energy demand prediction model, which combines production plans and environmental parameters to accurately predict future energy consumption and provide data support for optimization and regulation. Based on model predictive control (MPC), load distribution is optimized, high-energy consumption process operation timing is adjusted in combination with time-of-use electricity price, and equipment parameters are dynamically adjusted to improve energy utilization efficiency. The optimized regulation instructions are standardized and packaged to ensure the accuracy and executability of energy consumption control. The overall scheme effectively reduces energy costs, reduces peak load, improves the energy adaptability and stability of the production process, and thus enhances the intelligent level and economic benefits of manufacturing.

[0121] S5: Synchronize real-time regulation instructions to physical devices through digital threads to refresh the state of the multi-scale digital twin model.

[0122] Specifically, the following steps are included:

[0123] S5.1: Perform communication protocol adaptation through digital thread synchronization execution.

[0124] Specifically, communication protocol adaptation and configuration are performed, OPC UA Pub / Sub mode is adopted, QoS level is configured to at least one delivery (QoS1), transmission is performed through MQTT protocol, AES-256 load encryption is performed, REST API is used to interface with MES / SCADA, and the request header includes a digital signature (HMAC-SHA256).

[0125] Further, digital thread synchronization execution includes multi-channel instruction distribution. The main channel transmits critical instructions (such as device emergency stop) based on 5G private network; the backup channel transmits non-real-time instructions (such as production scheduling adjustment) through industrial Ethernet (Profinet) with 30% bandwidth reservation; edge caching stores the last 30 minutes of instructions locally on the device gateway, and executes cached instructions in priority order when the network is interrupted.

[0126] The device controller returns an acknowledgement message within 5 seconds after instruction execution, including actual execution parameter values, execution status code (success / failure / timeout), and real-time sensor readings (for verification). If no feedback is received, retransmission is performed with exponential backoff (first retry in 1 second, maximum of 3 retries).

[0127] S5.2: Refresh the state of the multi-scale digital twin model.

[0128] Specifically, the geometry layer collects device poses every 10 seconds through a laser scanner, triggering point cloud matching (ICP algorithm). When the pose offset is > 2mm, the device coordinates in the three-dimensional model are automatically updated.

[0129] The physical layer real-time injection device sensor data (temperature, pressure) to the heat conduction simulation model, update the temperature field distribution every 5 seconds. The discrete event simulation engine dynamically adjusts the event queue according to the actual process progress, and if the simulation and the physical object deviation > 3 seconds, trigger the calibration alarm.

[0130] The metabolic layer synchronizes energy metering data (electricity meter, gas meter readings) every 1 minute, updates the flow parameters in material flow analysis. The dynamic Bayesian network recalculates the state transition probability according to the latest energy consumption data, and updates the 30-minute prediction curve rolling.

[0131] Preferably, through multi-channel communication and encrypted transmission, the reliability of control instructions is ensured, and the safety and real-time performance of the execution instructions are improved. The edge cache and adaptive retransmission mechanism are used to enhance the robustness under network abnormal conditions. Through real-time updating of the multi-scale digital twin model, the data synchronization of the geometric, physical and metabolic layers is ensured, and the real running state of the equipment is always reflected, improving the accuracy of prediction and control. The automation degree and operation stability of the intelligent manufacturing system are improved, precise control, energy optimization and abnormal warning are realized, and a powerful support is provided for efficient, low-carbon and intelligent production environment.

[0132] The embodiment also provides a data integration management system for an automobile digital twin factory, comprising: a data acquisition module, a digital twin module, a quantum annealing module, an energy prediction module and an instruction synchronization module; the data acquisition module is used for collecting full-factor real-time data streams through industrial Internet of Things devices and edge computing nodes and pre-processing, using a knowledge graph to construct semantic mapping rules, and outputting a multi-modal data set; the digital twin module is used for constructing a multi-scale digital twin model based on the multi-modal data set, generating a factory state matrix, and the multi-scale digital twin model comprises a geometric layer, a physical layer and a metabolic layer; the quantum annealing module is used for constructing a quadratic unconstrained binary optimization model according to the factory state matrix, and solving to obtain a production scheduling scheme through a quantum annealing algorithm; the energy prediction module is used for combining the production scheduling scheme to drive the metabolic layer to perform energy demand prediction, triggering dynamic adjustment based on model predictive control, and outputting real-time control instructions; and the instruction synchronization module is used for synchronizing the real-time control instructions to physical devices through digital threads, and refreshing the state of the multi-scale digital twin model.

[0133] The embodiment also provides a computer device suitable for the case of the data integration management method for an automobile digital twin factory, comprising: a memory and a processor; the memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions to realize the data integration management method for an automobile digital twin factory as proposed in the above embodiment.

[0134] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved by WIFI, a carrier network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0135] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to implement the method for implementing data integration management for a digital twin factory of an automobile according to the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0136] In summary, the present application realizes real-time semantic alignment of multi-source heterogeneous data through edge side knowledge graph, reduces twin model update delay, and improves real-time performance of virtual-real mapping; adopts quantum chain encoding to convert high-dimensional scheduling problem into a solvable form of quantum annealing, breaks through the computational complexity bottleneck of traditional algorithms in multi-objective optimization, and enhances dynamic scheduling response capability; through the collaborative optimization mechanism of metabolic layer dynamic prediction and model predictive control, realizes closed-loop control of energy demand analysis, time-sharing scheduling and energy efficiency regulation, and achieves collaborative optimization of energy consumption cost and production efficiency; combined with digital thread synchronization and multi-dimensional visualization technology, while ensuring the time consistency of virtual and real systems, the accuracy of equipment pose calibration and the visualization analysis capability of energy flow trajectory are improved, forming a closed-loop optimization system from data fusion, simulation deduction to dynamic regulation.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A data integration and management method for automotive digital twin factories, characterized in that: include, By using industrial IoT devices and edge computing nodes, real-time data streams of all elements are collected and preprocessed, semantic mapping rules are constructed using knowledge graphs, and multimodal datasets are output. A multi-scale digital twin model is constructed based on a multi-modal dataset to generate a factory state matrix. The multi-scale digital twin model includes a geometric layer, a physical layer, and a metabolic layer. The geometric layer refers to importing a 3D model of the production line and dynamically calibrating the spatial pose of the equipment using laser point clouds; the physical layer refers to embedding partial differential equations of heat conduction and discrete event simulation logic to simulate material flow and process cycle time; the metabolic layer refers to establishing an energy metabolism network model, using material flow analysis to quantify energy demand, and using dynamic Bayesian networks to predict energy demand. Based on the factory state matrix, a quadratic unconstrained binary optimization model is constructed, and the production scheduling plan is obtained by solving the quantum annealing algorithm. The quantum bit encoding method of the quadratic unconstrained binary optimization model is as follows: each order is assigned a quantum chain, the chain length is equal to the number of selectable workstations, and equipment conflict constraints are achieved through cross-coupling terms. Combined with the production scheduling plan, the metabolic layer is driven to predict energy demand, triggering dynamic adjustment based on model prediction control, and outputting real-time control commands. The dynamic adjustment based on model predictive control includes, Load transfer is carried out based on the time window for adjusting the operation of the electricity price time-of-use signal; Energy efficiency is optimized by adjusting the air compressor outlet pressure using a PID controller. During peak grid load periods, energy storage batteries are discharged and charged during off-peak periods for energy storage scheduling. The digital twin model updates its state by synchronously sending control commands to physical devices in real time via digital threads.

2. The data integration and management method for automotive digital twin factories as described in claim 1, characterized in that: The knowledge graph uses a resource description framework triple to represent entity relationships.

3. The data integration and management method for automotive digital twin factories as described in claim 1, characterized in that: The process of refreshing the multi-scale digital twin model state includes dynamically rendering an energy consumption heatmap using the HSV color space, mapping the energy type to the hue dimension, mapping the energy intensity threshold to the saturation dimension, and overlaying and displaying the real-time control trajectory in the three-dimensional twin scene.

4. The data integration and management method for automotive digital twin factories as described in claim 1, characterized in that: The multi-scale digital twin model and the quadratic unconstrained binary optimization model are deployed on a container cluster, supporting parallel computing including energy metabolism network model simulation and quantum annealing solution.

5. A data integration management system for automotive digital twin factories, based on the data integration management method for automotive digital twin factories as described in any one of claims 1 to 4, characterized in that: This includes a data acquisition module, a digital twin module, a quantum annealing module, an energy prediction module, and an instruction synchronization module; The data acquisition module is used to collect real-time data streams of all elements through industrial IoT devices and edge computing nodes, preprocess them, construct semantic mapping rules using knowledge graphs, and output multimodal datasets. The digital twin module is used to construct a multi-scale digital twin model based on a multi-modal dataset and generate a factory state matrix. The multi-scale digital twin model includes a geometric layer, a physical layer, and a metabolic layer. The quantum annealing module is used to construct a quadratic unconstrained binary optimization model based on the factory state matrix, and obtain the production scheduling plan by solving the quantum annealing algorithm. The energy prediction module is used to combine the production scheduling plan to drive the metabolic layer to predict energy demand, trigger dynamic adjustment based on model prediction control, and output real-time control instructions. The instruction synchronization module is used to synchronize and control instructions to physical devices in real time through digital threads, and refresh the state of the multi-scale digital twin model.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the data integration management method for automotive digital twin factories as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the data integration management method for automotive digital twin factories as described in any one of claims 1 to 4.

Images