Supply chain traceability system based on digital twin and blockchain

Abstract

The present invention discloses a supply chain full-process traceability system based on digital twins and blockchain, which belongs to the field of digital twins and blockchain technology, including: a data acquisition layer, which acquires raw data from each link in real time and pre-processes the raw data using edge computing equipment; a twin model construction layer, which constructs digital twin models corresponding to each link of the supply chain, and generates link-level optimization parameters through simulation analysis; a blockchain network layer, which verifies and stores shared data and key traceability information between digital twin models; a collaborative optimization module, which dynamically adjusts cross-link collaborative strategies based on a multi-agent reinforcement learning algorithm; a data traceability layer, which responds to external traceability query requests and generates a full-link visual traceability map. This application uses a collaborative algorithm based on multi-agent reinforcement learning to ensure the automation and intelligence of information interaction between models in each link, and can also provide users with personalized, multi-terminal traceability services.

Description

Technical Field

[0001] The present invention relates to the field of digital twin and blockchain technology, and in particular to a full-process supply chain traceability system based on digital twin and blockchain. Background Art

[0002] Supply chain management, as a comprehensive management area, integrates multidisciplinary knowledge and encompasses the efficient coordination of the entire process, from raw material procurement and production operations to product delivery to end customers. However, traditional supply chain management models suffer from numerous shortcomings in data collection, storage and sharing, process monitoring, and quality traceability, making them unable to meet the requirements of modern enterprises for supply chain efficiency, transparency, and collaboration.

[0003] First, in terms of data collection, traditional supply chain management relies primarily on manual record-keeping or simple information system input. Limited by manpower and time, manual record-keeping makes it difficult to achieve real-time and comprehensive data collection and is prone to human error. Furthermore, existing information systems have limited functionality and struggle to integrate deeply with production equipment and logistics systems. This results in limited data collection scope and frequency, information lags, and impacts overall decision-making efficiency and responsiveness.

[0004] Secondly, when it comes to data storage and sharing, traditional technologies primarily rely on centralized database architectures. This architecture carries the risk of node failure; issues with core nodes can paralyze the entire system. Furthermore, centralized management methods lack effective security mechanisms to safeguard data authenticity and integrity, making data susceptible to tampering and lacking credibility. More seriously, data formats and standards are inconsistent across different enterprises or departments, leading to fragmented information development efforts and widespread "information silos," hindering efficient collaboration and information exchange across the supply chain.

[0005] Third, when it comes to product quality traceability and problem tracking, traditional methods typically rely on paper records or limited electronic records, with information scattered across different departments and business processes. This lacks a unified integration mechanism. Paper documents are easily lost or damaged, making them difficult to access. Electronic records, however, suffer from inconsistent formats and fragmented data, making the traceability process cumbersome, time-consuming, and inaccurate. This makes it difficult to quickly pinpoint the root cause of quality issues, severely impacting a company's emergency response capabilities and customer satisfaction.

[0006] In summary, the traditional supply chain management model has been unable to adapt to the current complex and changing market environment and the needs of high-quality development of enterprises, and there is an urgent need for system optimization and technological upgrades. Summary of the Invention

[0007] The purpose of this invention is to provide a full-process supply chain traceability system based on digital twins and blockchain to address the shortcomings of traditional supply chain management models in data collection, storage and sharing, process monitoring, and quality traceability.

[0008] To achieve the above objectives, this application adopts the following technical solutions:

[0009] The present application provides a full-process supply chain traceability system based on digital twins and blockchain. The system includes a data acquisition layer, a twin model construction layer, a blockchain network layer, a collaborative optimization module, and a data traceability layer, wherein:

[0010] The data collection layer, based on sensors and data collection devices deployed at each link of the supply chain, acquires raw data from each link in real time and pre-processes the raw data using edge computing devices;

[0011] The twin model construction layer is used to construct digital twin models corresponding to each link of the supply chain based on the preprocessed data. Each digital twin model is updated based on real-time data to dynamically reflect the operating status of the physical entity in the corresponding link, and generates link-level optimization parameters through simulation analysis;

[0012] The blockchain network layer is configured as a consortium chain network composed of nodes from all supply chain participants. It uses a hierarchical consensus mechanism to verify and store shared data and key traceability information between digital twin models, and associates cross-link data through a globally unique identifier.

[0013] The collaborative optimization module is coupled between the twin model construction layer and the blockchain network layer. Based on the multi-agent reinforcement learning algorithm, it drives each digital twin model to interact with the blockchain network through state information and optimization parameters, and dynamically adjusts the cross-link collaborative strategy according to the consensus results on the chain;

[0014] The data traceability layer is used to respond to external traceability query requests, retrieve cross-link data associated with the target global unique identifier in the blockchain network, and generate a full-link visual traceability map in combination with various digital twin models.

[0015] Preferably, the supply chain links include production links and transportation links;

[0016] The participants in the supply chain include core enterprises and suppliers as the main nodes, and logistics companies and sellers as the secondary nodes.

[0017] Preferably, the twin model construction layer includes:

[0018] A virtual production equipment model is constructed based on the initialization data of the production process, used to simulate the operating status of the production equipment, and autonomously optimize the production parameters using a reinforcement learning algorithm, and pass the production parameters to the collaborative optimization module;

[0019] A twin model of the transportation process, integrating geographic information and real-time transportation data, is used to simulate vehicle positions and transportation routes, and to generate alternative routes based on road condition predictions and upload them to the collaborative optimization module;

[0020] The model collaboration interface realizes real-time data exchange between production and transportation models based on the collaborative optimization module, and shares optimization parameters and strategy execution results through the blockchain network.

[0021] Preferably, the collaborative optimization module includes:

[0022] A state perception unit, configured to receive historical optimization parameters, real-time operating status, and simulation analysis results of the virtual production equipment model and the transportation process twin model;

[0023] Strategy learning unit, based on reinforcement learning mechanism Learning update rules, treating each digital twin model as an agent, and calculating the optimal behavior strategy for each agent based on the received data;

[0024] The on-chain feedback unit is used to encapsulate the optimal behavior strategy into a blockchain transaction, which is then written into the on-chain ledger after verification by a hierarchical consensus mechanism.

[0025] Preferably, the blockchain network layer includes:

[0026] Smart contract module, including transaction settlement contracts and goods delivery contracts, is used to automatically execute contract terms when supply chain operations meet preset conditions;

[0027] The identifier management module is used to generate a globally unique identifier and bind the globally unique identifier to the first blockchain transaction to form a data association anchor.

[0028] Preferably, the smart contract module further includes:

[0029] The cross-model collaborative response unit is used to automatically call the target model interface to generate a collaborative processing solution when any digital twin model detects an abnormal event, and write the event processing record into the blockchain network.

[0030] Preferably, the layered consensus mechanism includes a Byzantine Fault Tolerant consensus algorithm used between primary nodes and an improved practical Byzantine Fault Tolerant algorithm used between secondary nodes and primary nodes.

[0031] Preferably, the improved practical Byzantine fault tolerance algorithm comprises the following steps:

[0032] Select a subset of primary nodes as the verification node subset, and set the fast verification threshold K and synchronization period S;

[0033] When a secondary node generates data, it generates a message and submits the message to a subset of validating nodes in the primary node for preliminary verification;

[0034] If within a short period of time, K nodes in the verification node subset successfully verify the message, a quick confirmation message is sent to the secondary node;

[0035] After receiving the quick confirmation message, the secondary node directly writes the data into the local ledger;

[0036] When the synchronization period S arrives, the secondary node synchronizes the ledger with the primary node and repairs the differences.

[0037] Preferably, the data tracing method includes the following steps:

[0038] Receive a traceability query request sent by a user, wherein the request includes a globally unique identifier;

[0039] Retrieving corresponding historical transaction data in the blockchain network according to the globally unique identifier, and distributing the historical transaction data to the corresponding digital twin model;

[0040] Using each digital twin model to reverse simulate the historical transaction data and generate a visual map of the entire supply chain process;

[0041] Dynamically filter data based on user roles to display customized traceability information.

[0042] Preferably, the key traceability information includes product completion time, production batch and quality inspection results in the production link, and real-time location and vehicle transportation route in the transportation link.

[0043] The present invention has the following beneficial effects:

[0044] The present invention achieves efficient collection and preprocessing of raw data by deploying a variety of sensors and equipment in various links of the supply chain; it also uses advanced technologies such as industrial Internet of Things edge intelligent collaboration, digital twin and geographic information system integration to create an intelligent model with real-time self-learning and optimization, and active intervention capabilities, and realizes comprehensive digital mapping of physical entities in the supply chain; and with the help of collaborative algorithms based on multi-agent reinforcement learning, it ensures the automation and intelligence of information interaction between models in each link; at the same time, in terms of blockchain network construction, based on the design of layered consensus mechanism and targeted smart contracts, it ensures data consistency, reliability and automated processing of business processes; it can also provide users with personalized, multi-terminal traceability services, and realize transparent and traceable management of the entire supply chain process. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 This is a structural diagram of a full-process supply chain traceability system based on digital twins and blockchain provided in an embodiment of the present application;

[0047] Figure 2 This is a flowchart of the Byzantine fault-tolerant consensus algorithm in an embodiment of the present application;

[0048] Figure 3 This is a flowchart of the improved practical Byzantine fault-tolerant algorithm in the embodiment of the present application;

[0049] Figure 4 It is a flow chart of the data tracing method in an embodiment of the present application. DETAILED DESCRIPTION

[0050] To make the technical solution of the present application clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. The terms "first", "second", etc. in the claims and description of the present application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged where appropriate. This is merely a way of distinguishing objects of the same attributes when describing the embodiments of the present application. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, so that a process, method, system, product, or device comprising a series of units is not necessarily limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0051] This embodiment provides a supply chain traceability system based on digital twins and blockchain. Figure 1 As shown in the figure, the system includes data acquisition layer, twin model construction layer, blockchain network layer, collaborative optimization module and data traceability layer.

[0052] Among them, the data collection layer is based on sensors and data collection equipment deployed in various links of the supply chain, obtains raw data from each link in real time, and uses edge computing devices to pre-process the raw data.

[0053] Specifically, various sensors and data acquisition devices are deployed in various links of the supply chain, including but not limited to temperature sensors and humidity sensors for recording parameters of product production links, GPS positioning devices for recording logistics transportation locations, and radio frequency identification tag readers for recording product raw material information, etc., so as to collect original data of production, transportation, warehousing, sales and other supply chain links in real time, such as raw material sources, production progress, output, product offline time, batch number, transportation trajectory, inventory quantity, etc., and preliminarily organize and classify all collected original data. At the same time, the Industrial Internet of Things (IIoT) technology is used to connect these sensors and data acquisition devices to edge computing devices, and the edge computing devices are used to pre-process the data after preliminary organization and classification. The pre-processing specifically includes data cleaning, deduplication, feature extraction and format standardization, etc., aiming to eliminate invalid noise data and duplicate records, thereby significantly reducing the data transmission load and effectively improving the overall data quality.

[0054] Among them, the twin model construction layer is used to construct digital twin models corresponding to each link of the supply chain based on the preprocessed data. Each digital twin model is updated based on real-time data to dynamically reflect the operating status of the physical entity in the corresponding link, and generate link-level optimization parameters through simulation analysis.

[0055] Furthermore, the twin model building layer includes:

[0056] A virtual production equipment model is constructed based on the initialization data of the production process, used to simulate the operating status of the production equipment, and autonomously optimize the production parameters using a reinforcement learning algorithm, and pass the production parameters to the collaborative optimization module;

[0057] A twin model of the transportation process, integrating geographic information and real-time transportation data, is used to simulate vehicle positions and transportation routes, and to generate alternative routes based on road condition predictions and upload them to the collaborative optimization module;

[0058] The model collaboration interface realizes real-time data exchange between production and transportation models based on the collaborative optimization module, and shares optimization parameters and strategy execution results through the blockchain network.

[0059] In this embodiment, the main focus is on building digital twin models of the production and transportation links.

[0060] Specifically, for the production link, a virtual production equipment model is constructed based on the production equipment parameters and process data contained in the preprocessed data. This model can not only simulate the equipment operating status, production efficiency and product quality fluctuations in the actual production process, but also simulate the production progress, output and product offline time in real time.

[0061] By simulating the operating status of equipment in the actual production process, such as start-up, stop, and failure, the model can predict possible problems with the equipment in advance, optimize equipment maintenance plans, and reduce downtime; by simulating the production process to evaluate production efficiency and identify production bottlenecks, it can optimize the production process, improve overall production efficiency, and reduce production costs; by simulating the fluctuations in product quality during the production process, it can predict the defective rate, adjust production parameters in advance, reduce the defective rate, and improve product quality; by simulating the production progress in real time, it can grasp the product progress in real time and ensure the smooth execution of the production plan; by predicting output in advance, it can optimize resource allocation and meet market demand; by accurately predicting the time when products are offline, it can optimize supply chain management, reduce inventory backlogs, and improve customer satisfaction.

[0062] At the same time, the virtual production equipment model also has adaptive control capabilities based on reinforcement learning. It uses reinforcement learning algorithms and real-time data to train and optimize production parameters and dynamically adjust production strategies. Specifically, it continuously receives the latest equipment operation data and uses comparative analysis technology based on generative adversarial networks to conduct in-depth comparative analysis with its own simulated operation results. When a deviation is found between the actual data and the simulation results, the self-learning mechanism is activated and the model parameters are optimized and adjusted using a deep transfer learning algorithm. Assuming that the actual production efficiency is lower than the model's expectations, the complex connections between multiple sources of information such as equipment operation and raw material quality data are deeply analyzed with the help of knowledge graph technology to identify the key factors affecting production efficiency. The equipment performance parameters and production process parameters in the model are adjusted through the intelligent decision-making system to enable the model to more accurately reflect the actual production situation, or to autonomously adjust the production strategy based on the real-time monitored equipment operation data.

[0063] For the logistics and transportation links, the digital twin and geographic information system fusion technology is used to build a transportation process twin model based on the transportation vehicle information, route planning information and real-time transportation data contained in the data pre-processed by the edge computing device. This model is used to display the transportation vehicle location, driving speed, estimated arrival time and other information in real time. Through the introduction of spatiotemporal big data analysis algorithms, real-time predictions are made on factors such as road conditions changes and weather impacts during the transportation process. If congestion is predicted to occur on a certain road section, a quantum heuristic optimization algorithm is used to quickly plan an alternative route for the transportation vehicle. At the same time, it coordinates surrounding vehicle resources to realize dynamic allocation and collaborative scheduling of transportation tasks, thereby improving transportation efficiency. It also predicts vehicle failure risks through deep learning of historical transportation data and real-time environmental data, arranges maintenance plans in advance, and reduces transportation delays caused by vehicle failures, thereby providing guarantees for logistics and transportation reliability.

[0064] By constructing digital twin models of each link in the supply chain, we can achieve comprehensive digital mapping of the physical entities in the supply chain, providing intuitive and visual basis for traceability throughout the entire process.

[0065] Among them, the blockchain network layer is configured as a consortium chain network composed of nodes from all supply chain participants. It adopts a layered consensus mechanism to verify and store shared data and key traceability information between digital twin models, and associates cross-link data through a globally unique identifier.

[0066] Key traceability information includes product completion time, production batches and quality inspection results in the production process, as well as the real-time location and transportation route of vehicles in the transportation process.

[0067] In this embodiment, a multi-party blockchain network is designed using a consortium chain architecture, and core enterprises, suppliers, sellers, and logistics companies in the supply chain are added to the network as nodes of the consortium chain. Among them, core enterprises and suppliers serve as primary nodes, and sellers and logistics companies serve as secondary nodes. Each node has independent data storage and processing capabilities, and a hierarchical consensus mechanism is used to ensure the consistency and reliability of data between nodes. At the same time, data processed and verified by the digital twin model will also be packaged into blocks according to a specific format and timestamp and stored on the blockchain.

[0068] Blockchain uses asymmetric encryption technology to ensure the security and immutability of data, and any modification of data by any node must be verified by a layered consensus mechanism, ensuring the authenticity and integrity of data throughout the entire supply chain process.

[0069] Among them, the layered consensus mechanism includes the Byzantine fault-tolerant consensus algorithm used between primary nodes and the improved practical Byzantine fault-tolerant algorithm used between secondary nodes and primary nodes.

[0070] Specifically, if Figure 2 As shown in Figure 2, the Byzantine fault-tolerant consensus algorithm mainly includes the following steps:

[0071] S110. Parameter Setting: Define the specific number of key nodes, such as core enterprises and public key suppliers. The specific number N is determined based on the actual scale of the supply chain and the complexity of the business. Considering network conditions, set the maximum allowable delay time T for message propagation between nodes to control the time range for reaching consensus. For example, in a high-speed and stable internal enterprise network environment, T is set to a relatively small value, such as 100 milliseconds. If there are more unstable factors in the network, T needs to be appropriately increased. Based on the reliability assessment of the main nodes, set the tolerable proportion of faulty or malicious nodes, f, such as f = N / 3 (rounded down), to ensure that the consensus mechanism can still operate normally when a certain number of nodes have problems.

[0072] S111, Message Broadcast: When a major node generates new key data (e.g., a core enterprise determines a new raw material procurement plan), the data and related transaction information are packaged into a message in a specific format and broadcast to all other major nodes;

[0073] S112, message verification: Each receiving node verifies the legitimacy of the received message, including whether the data format is correct and whether the signature is valid. If the verification passes, the node stores the message in the local cache and sends a confirmation message to other nodes;

[0074] S113, consensus reached: When a node receives more than 2f+1 confirmation messages (including messages sent by itself), it considers that the message has reached consensus, writes the relevant data into the local ledger, and updates the blockchain status.

[0075] The Byzantine Fault Tolerant consensus algorithm ensures that consensus can be reached quickly even when there are a large number of nodes and some nodes fail or are subject to malicious attacks, thus ensuring the consistency and reliability of key data.

[0076] like Figure 3 As shown in the figure, the improved practical Byzantine fault tolerance algorithm mainly includes the following steps:

[0077] S210. Parameter setting: Select a subset of primary nodes as the verification node subset M; set a fast verification message threshold, K, so that when a secondary node receives K identical messages from the verification node subset, it immediately considers consensus reached and enters the data processing flow; define the time interval for data synchronization between the secondary node and the primary node, such as every 5 minutes, to ensure the real-time data of the secondary node;

[0078] S211, Message Submission: After a secondary node (such as a logistics provider) generates data (such as a shipping status update), it submits the message to a subset of validating nodes in the primary node.

[0079] S212, Fast Verification: After receiving the message, the verification node performs preliminary verification. If K nodes in the verification node subset successfully verify the same message within a short period of time (e.g., 1 second), a fast confirmation message is sent to the secondary node. After receiving the fast confirmation message, the secondary node can directly write the data to the local ledger.

[0080] S213, regular synchronization: When the synchronization period S arrives, the secondary node synchronizes all data with the primary node, compares the ledger differences, repairs inconsistent data, and ensures data consistency.

[0081] The improved Practical Byzantine Fault Tolerance algorithm improves data transmission and storage efficiency by reducing unnecessary consensus verification steps. It also significantly enhances the system's efficiency and fault tolerance by introducing fast verification and regular synchronization mechanisms. This improvement is particularly suitable for distributed systems that require fast response and high throughput, such as logistics and supply chain management. Through these optimizations, the system can maintain data consistency while improving overall performance and user experience.

[0082] Specifically, the blockchain network layer includes:

[0083] Smart contract module, including transaction settlement contracts and goods delivery contracts, is used to automatically execute contract terms when supply chain operations meet preset conditions;

[0084] The identifier management module is used to generate a globally unique identifier and bind the globally unique identifier to the first blockchain transaction to form a data association anchor.

[0085] The blockchain network layer also has a smart contract module, which writes smart contract code for business processes such as transaction settlement and goods handover in the supply chain. When the smart contract meets the preset conditions, the smart contract is automatically triggered to execute, realizing the automation of business processes, reducing manual intervention, improving the efficiency and transparency of supply chain business processing, and further ensuring the safe and efficient flow of data throughout the entire supply chain process on the blockchain.

[0086] The transaction settlement contract includes parameters such as the two parties to the transaction, such as suppliers and core enterprises, the settlement amount, settlement conditions such as settlement within X days after product acceptance, and payment methods such as digital currency payment addresses. When the settlement conditions are met, the contract execution is automatically triggered, and the funds are transferred from the payer's account to the payee's account.

[0087] The cargo delivery contract includes parameters such as the parties to the cargo delivery, cargo information (such as name, quantity, specifications, etc.), delivery address, and delivery time. When the cargo is transported to the designated delivery location and the time is in compliance with the contract, and the recipient confirms that the cargo is correct, the contract automatically records the completion of the delivery and updates the status of the cargo in the supply chain.

[0088] The smart contract module also includes:

[0089] The cross-model collaborative response unit is used to automatically call the target model interface to generate a collaborative processing solution when any digital twin model detects an abnormal event, and write the event processing record into the blockchain network.

[0090] Assuming that the virtual production equipment model predicts through a reinforcement learning algorithm that equipment Machine-001 will fail in 30 minutes, it first generates an abnormal event {GUID: "PROD-20250421-002", failure type: "equipment downtime risk", remaining normal operation time: 25 minutes}, and then calls the "transportation resource reallocation" interface of the transportation process twin model through the smart contract to notify it that the PROD-20250421-002 batch of products originally planned to be produced by Machine-001 will be delayed by 2 hours; after receiving the notification, the transportation process twin model immediately adjusts the route of the transport vehicle TRUCK-008 (originally scheduled to arrive at the factory at 15:00), delaying its departure to 17:00, and temporarily allocates the idle vehicle to other urgent tasks.

[0091] Suppose the transportation process twin model predicts, through spatiotemporal big data analysis, that a certain road section (for example, the 3km mark of Highway X) will be congested in one hour (with a probability >80%). It first generates a congestion warning event {GUID: "PROD-20250421-001", congestion period: 16:00–18:00, recommended detour route: National Highway Y}. Then, using a multi-agent reinforcement learning algorithm, it sends a request to the production model: "Should the loading time of batch PROD-20250421-001 be adjusted to after 17:30?" Upon receiving this request, the virtual production equipment model, based on the current production schedule (estimated completion time at 16:30), uses knowledge graph analysis to conclude that a one-hour loading delay will not affect subsequent processes. It then returns an approval instruction, which the transportation process digital twin uses to reschedule the departure time of vehicle TRUCK-007.

[0092] The blockchain network layer also has an identifier management module, which assigns a unique identification code, namely a globally unique identifier (GUID), to each batch or single product in the product production process. As a "digital ID card" throughout the entire supply chain, this GUID is associated with data such as raw material batches and process parameters in the production process, stored in the initial blockchain transaction of the production process, and synchronized with the blockchain nodes of logistics, warehousing and other links as the product flows, realizing cross-link data association and anchoring. For example, the production link generates a transaction Tx-Production, including the fields {GUID: "PROD-20250421-001", raw material batch: "RM-202504-05", process version: "V3.2", production time: "2025-04-21 09:00"}; the logistics link generates a transaction Tx-Logistics, including the fields {GUID: "PROD-20250421-001", transportation period: "2025-04-21 14:00~2025-04-22 08:00", temperature and humidity data: "[22°C, 55%RH]", vehicle ID: "TRUCK-007"}.

[0093] GUIDs are also used to establish a hash chain link between data across different stages, ensuring data immutability and traceability. Specifically, when building the digital twin virtual model, GUIDs are used to map virtual equipment operating data from the production stage (such as equipment numbers and process parameters) to the transportation model from the logistics stage (such as vehicle numbers and route nodes). For example, in the production model, a product batch produced by machine-001 corresponds to the GUID PROD-20250421-001; in the logistics model, the cargo transported by vehicle TRUCK-007 has the GUID PROD-20250421-001. Cross-validation of spatiotemporal data (production time and transportation period) ensures the accuracy of data association. When a product is rolled off the production line, the edge computing device synchronizes the GUID and production data to the digital twin model of the logistics and transportation stage through a smart contract, triggering the transportation model's transportation resource scheduling. The data collected from the logistics and transportation stage is then written to the blockchain network by the edge computing device and linked to the transaction associated with the corresponding GUID. If the production time is later than the shipping time, it is determined that there is a data error or process problem.

[0094] Among them, the collaborative optimization module is coupled between the twin model construction layer and the blockchain network layer. Based on the multi-agent reinforcement learning algorithm, it drives each digital twin model to interact with the state information and optimization parameters through the blockchain network, and dynamically adjusts the cross-link collaborative strategy according to the on-chain consensus results.

[0095] Specifically, the collaborative optimization module includes:

[0096] The state perception unit is used to receive the historical optimization parameters, real-time operation status and simulation analysis results of the virtual production equipment model and the transportation process twin model;

[0097] Strategy learning unit, based on reinforcement learning mechanism Learning update rules, treating each digital twin model as an agent, and calculating the optimal behavior strategy for each agent based on the received data;

[0098] The on-chain feedback unit is used to encapsulate the optimal behavior strategy into a blockchain transaction, which is then written into the on-chain ledger after verification by the hierarchical consensus mechanism.

[0099] In order to meet the automation and intelligent requirements of information interaction of digital twin models in each link, this embodiment also adopts a collaborative algorithm based on multi-agent reinforcement learning. Specifically, in the collaborative algorithm based on multi-agent reinforcement learning, the digital twin model corresponding to each supply chain link is an agent, such as the virtual production equipment model is the production link agent, and the transportation process twin model is the logistics and transportation link agent, and the reinforcement learning mechanism is introduced. Learning update rules for policy optimization is expressed as follows:

[0100] ,

[0101] in, Indicates that the status Take action The current value estimate of the corresponding production link agent or logistics link agent performs a certain action in a specific state, such as the production link agent in the event of equipment failure (state ) when selecting to send fault information to the logistics transport link (action )of value, Indicates "update with" or "replace with."

[0102] is the learning rate, which usually ranges from 0 to 1 and determines the effect of newly acquired information on the original The degree of update of the value, such as Indicates new information The impact of value updates is relatively small and relies more on historical experience.

[0103] Is to perform an action The rewards obtained from the environment are related to the design of the reward mechanism. For example, the production agent receives a +10 reward for reasonably adjusting the production speed to avoid goods backlog. .

[0104] is a discount factor, also ranging from 0 to 1, used to balance the importance of current rewards and future rewards, for example It means that you place more emphasis on future rewards and tend to choose actions that can bring long-term benefits.

[0105] Is to perform an action The new state that is transferred to later, such as the logistics transportation link agent reallocating vehicle resources (action ) After that, the vehicle position, transportation task allocation, etc. change, forming a new state .

[0106] In the new state All possible moves middle The one with the largest value, that is, the agent's estimate of the value of the optimal action in the future, represents the best choice, such as in the new transportation task allocation state In this case, the intelligent agent in the logistics transport link can choose to optimize the transport route, adjust the vehicle load and other actions. The action with the largest value corresponds to The value is .

[0107] By continuously repeating this update process, the production link intelligent agent and the logistics and transportation link intelligent agent can gradually learn what actions (such as adjusting the production rate, selecting the optimal distribution route, etc.) to take under different conditions (such as inventory levels, order quantities, traffic conditions, etc.) based on real-time collected data and business rules to maximize the accumulated rewards, thereby realizing the automation and intelligence of the information interaction of the digital twin models in each link. For example, the production link intelligent agent and the logistics and transportation link intelligent agent found through calculation that notifying the logistics to prepare for transportation in advance within a certain period of time before the completion of the product can maximize the accumulated rewards. Then, through the information interaction mechanism and collaborative process established by these two links, when the production link predicts that a batch of products is about to be completed, the information will be synchronized to the transportation process twin model. The model will arrange suitable vehicles on standby in advance based on the vehicle location, transportation capacity and real-time status of the transportation route, optimize the transportation route planning, ensure that the products can be quickly loaded and shipped after completion, reduce the waiting time of products in the warehouse, and thus improve the overall efficiency of the supply chain. Among them, business rules refer to the operating specifications set within the enterprise, such as quality control standards and safety regulations.

[0108] The purpose of calculating the maximum cumulative reward in this embodiment is to guide the agents in each link to find the behavioral strategies that are most conducive to improving the overall efficiency of the supply chain. In a complex supply chain system, different links are interconnected and face various dynamic changes, such as equipment failures and road congestion. By maximizing the cumulative reward, the agents can identify the action plan that is most beneficial to the overall supply chain's long-term benefits from a wide range of possible behavioral choices. For example, if the production agent can obtain high cumulative rewards by reasonably adjusting production speed and promptly communicating and coordinating with the logistics and transportation links, it indicates that these actions are helping to improve the overall operational efficiency of the supply chain, including reducing inventory backlogs, lowering transportation costs, and improving on-time delivery rates.

[0109] Among them, the data traceability layer is used to respond to external traceability query requests, and retrieve cross-link data associated with the target global unique identifier in the blockchain network, and combine various digital twin models to generate a full-link visual traceability map.

[0110] Specifically, the data tracing method includes the following steps:

[0111] Receive a traceability query request sent by a user, wherein the request includes a globally unique identifier;

[0112] Retrieve the corresponding historical transaction data in the blockchain network based on the globally unique identifier, and distribute the historical transaction data to the corresponding digital twin model;

[0113] Using each digital twin model to reverse simulate the historical transaction data and generate a visual map of the entire supply chain process;

[0114] Dynamically filter data based on user roles to display customized traceability information.

[0115] like Figure 4As shown in the figure, when a link in the supply chain requires data query or traceability, a user sends a traceability query request to the blockchain network through the traceability system's client. Furthermore, users can enter or scan a product's GUID through the front-end interface to issue a traceability query request. This request contains the GUID and user identity information (such as consumer, regulator, or company internal staff). This GUID is used as an index in the blockchain network to locate the corresponding block data. Each data item is then pushed to the digital twin model of the corresponding link based on the event type. Based on the received data, the digital twin model starts from the final state and reversely traces the Zhengege supply chain process. Through the collaborative work of the digital twin models, the entire process of the product from its birth to the current state is restored. A visual map of the entire process is then constructed using a graph database such as Neo4j. This visualization displays the actual status of each link in the supply chain and the changes in historical data. Through interactive operations, users can view detailed information about the entire product process, from raw material procurement to final sale, including the specific time, location, operation content, and responsible personnel of each link, thus achieving transparent and traceable management of the entire supply chain process.

[0116] The data traceability layer can also choose different display methods and information emphases based on user roles. For example, for manufacturers, the focus can be on process parameter changes and quality inspection data during the production process; for consumers, the display interface can be simplified to highlight the product's raw material source, production date, shelf life, and quality certification information. Furthermore, the client supports multi-terminal access, including but not limited to PC, mobile, and smart wearable devices.

[0117] This embodiment is based on the data collection layer. By deploying a variety of sensors and equipment in various links of the supply chain, it realizes the efficient collection and preprocessing of raw data. In the digital twin model construction link, it uses advanced technologies such as industrial Internet of Things edge intelligent collaboration, digital twin and geographic information system integration, and creates an intelligent model with real-time self-learning and optimization, and active intervention capabilities, realizing a comprehensive digital mapping of the physical entities of the supply chain. It also uses a collaborative algorithm based on multi-agent reinforcement learning to ensure the automation and intelligence of information interaction between models in each link. At the same time, in terms of blockchain network construction, the layered consensus mechanism and the design of targeted smart contracts ensure the consistency and reliability of data and the automation of business processes. Finally, the data interaction and traceability process provides users with personalized, multi-terminal traceability services, realizing transparent and traceable management of the entire supply chain process.

[0118] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A full-process supply chain traceability system based on digital twins and blockchain, characterized by: The system includes a data acquisition layer, a twin model construction layer, a blockchain network layer, a collaborative optimization module, and a data traceability layer, wherein: The data collection layer, based on sensors and data collection devices deployed at each link of the supply chain, acquires raw data from each link in real time and pre-processes the raw data using edge computing devices; The twin model construction layer is used to construct digital twin models corresponding to each link of the supply chain based on the preprocessed data. Each digital twin model is updated based on real-time data to dynamically reflect the operating status of the physical entity in the corresponding link, and generates link-level optimization parameters through simulation analysis; The blockchain network layer is configured as a consortium chain network consisting of core supply chain enterprises and suppliers as primary nodes, and logistics companies and sellers as secondary nodes. It uses a layered consensus mechanism to verify and store shared data and key traceability information between digital twin models, and associates cross-link data through a globally unique identifier. The layered consensus mechanism includes a Byzantine fault-tolerant consensus algorithm used between primary nodes and an improved practical Byzantine fault-tolerant algorithm used between secondary nodes and primary nodes. The collaborative optimization module is coupled between the twin model construction layer and the blockchain network layer. Based on the multi-agent reinforcement learning algorithm, it drives each digital twin model to interact with the blockchain network through state information and optimization parameters, and dynamically adjusts the cross-link collaborative strategy according to the consensus results on the chain; The data traceability layer is used to respond to external traceability query requests, retrieve cross-link data associated with the target globally unique identifier in the blockchain network, and generate a full-link visual traceability map in combination with each digital twin model; The improved practical Byzantine fault tolerance algorithm includes the following steps: Select a subset of primary nodes as the verification node subset, and set the fast verification threshold K and synchronization period S; When a secondary node generates data, it generates a message and submits the message to a subset of validating nodes in the primary node for preliminary verification; If within a short period of time, K nodes in the verification node subset successfully verify the message, a quick confirmation message is sent to the secondary node; After receiving the quick confirmation message, the secondary node directly writes the data into the local ledger; When the synchronization period S arrives, the secondary node synchronizes the ledger with the primary node and repairs the differences; The data tracing method includes the following steps: Receive a traceability query request sent by a user, wherein the request includes a globally unique identifier; Retrieving corresponding historical transaction data in the blockchain network according to the globally unique identifier, and distributing the historical transaction data to the corresponding digital twin model; Using each digital twin model to reverse simulate the historical transaction data and generate a visual map of the entire supply chain process; Dynamically filter data based on user roles to display customized traceability information.

2. A supply chain traceability system based on digital twins and blockchain according to claim 1, characterized in that: The various links of the supply chain include production link and transportation link.

3. A supply chain traceability system based on digital twins and blockchain according to claim 2, characterized in that: The twin model construction layer includes: A virtual production equipment model is constructed based on the initialization data of the production process, used to simulate the operating status of the production equipment, and autonomously optimize the production parameters using a reinforcement learning algorithm, and pass the production parameters to the collaborative optimization module; A twin model of the transportation process, integrating geographic information and real-time transportation data, is used to simulate vehicle positions and transportation routes, and to generate alternative routes based on road condition predictions and upload them to the collaborative optimization module; The model collaboration interface realizes real-time data exchange between production and transportation models based on the collaborative optimization module, and shares optimization parameters and strategy execution results through the blockchain network.

4. A supply chain traceability system based on digital twins and blockchain according to claim 3, characterized in that: The collaborative optimization module includes: A state perception unit, configured to receive historical optimization parameters, real-time operating status, and simulation analysis results of the virtual production equipment model and the transportation process twin model; The strategy learning unit, based on the learning update rules in the reinforcement learning mechanism, treats each digital twin model as an intelligent agent and calculates the optimal behavior strategy for each intelligent agent based on the received data; The on-chain feedback unit is used to encapsulate the optimal behavior strategy into a blockchain transaction, which is then written into the on-chain ledger after verification by a hierarchical consensus mechanism.

5. The full-process supply chain traceability system based on digital twins and blockchain according to claim 2 is characterized in that: The blockchain network layer includes: Smart contract module, including transaction settlement contracts and goods delivery contracts, is used to automatically execute contract terms when supply chain operations meet preset conditions; The identifier management module is used to generate a globally unique identifier and bind the globally unique identifier to the first blockchain transaction to form a data association anchor.

6. A supply chain traceability system based on digital twins and blockchain according to claim 5, characterized in that: The smart contract module also includes: The cross-model collaborative response unit is used to automatically call the target model interface to generate a collaborative processing solution when any digital twin model detects an abnormal event, and write the event processing record into the blockchain network.

7. The supply chain traceability system based on digital twin and blockchain according to claim 2 is characterized in that: The key traceability information includes product completion time, production batches and quality inspection results in the production process, as well as the real-time location and vehicle transportation route of the transportation process.

Images