Preserved egg production whole process tracing method and system based on block chain

Abstract

The invention relates to the technical field of product data management, in particular to a preserved duck egg production whole-process traceability method and system based on a block chain, and the method comprises the steps: registering raw material eggs in batches, gathering and capturing key parameters such as cellaring temperature and humidity, feed liquid alkalinity and the like in production, and generating a process witness which cannot be tampered; a double-rail treatment normal form composed of an execution rail and a learning rail is used for processing witness, the execution rail conducts real-time intervention according to the operation standard, and the learning rail conducts iterative optimization on the predictive model; and finally, associating the process data with a subsequent supply chain transfer event, and aggregating to form a supply chain decision sand table which can be used for simulation and deduction. According to the invention, refined data management and closed-loop control in the production process are combined, and isolated production management is improved to an integrated supply chain collaborative decision-making level through data association and sand table deduction penetrating through the whole chain, so that the quality control precision and management decision-making level of the non-standard product supply chain are systematically improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of product data management, in particular to a method and system for tracing the whole process of producing preserved eggs based on blockchain. BACKGROUND

[0002] In the field of modern enterprise production and operation management, especially in the field of product data management technology, using information technology to realize data tracking of the whole life cycle of products has become the key to improving management efficiency and data transparency. In the prior art, a distributed data recording framework is provided by a traceability system based on blockchain, which creates tamper-proof records for products and their circulation links, supports cross-department and cross-enterprise supply chain data collaboration and monitoring. At the same time, combined with automatic collection equipment and radio frequency identification technology, standardized data collection and fine tracking management of materials and finished products are realized, providing basic data support for quality control and logistics scheduling.

[0003] However, the existing management system exposes its core technical problem when applied to the production of non-standard products with special processes, i.e., it lacks the ability to dynamically and credibly manage structured data of key process parameters within the production process. For example, the core quality of preserved eggs depends on special process links such as pickling and curing, but existing systems usually stop at batch management of raw materials and finished products, and cannot effectively collect and associate data such as material liquid ratio, storage environment, and pickling time, which determine the key control points of the final product quality. The lack of such data management results in a blind area of quality control in the production process, making it difficult to accurately define production responsibilities, and also causing a lack of data support for process optimization and management decisions.

[0004] Therefore, the present application provides a method and system for tracing the whole process of producing preserved eggs based on blockchain. SUMMARY

[0005] The present application aims to provide a method and system for tracing the whole process of producing preserved eggs based on blockchain, which deeply contextualizes the binding of production batches and process environment, generates tamper-proof encrypted witnesses for key dynamic parameters in the production process, and realizes real-time intervention and forward-looking optimization of the production process using a dual-track governance paradigm, thereby solving the problem that traditional traceability systems cannot penetrate the internal process flow, resulting in a blind area of quality control in the production process, and causing a lack of data support for process optimization and management decisions.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions: A method for tracing the whole process of producing preserved eggs based on blockchain, comprising: registering a unique production instruction for a raw egg batch, generating an initial job record containing an initial material profile, and recording on a distributed operating ledger; associating the initial job record with a cellar unit and a process recipe as production elements, generating a work-in-process job profile associated with the production elements by performing strategic anchoring of resources; aggregating the work-in-process job profile with captured operating parameters, including liquor alkalinity, cellar temperature and humidity, and aging duration, by generating an immutable witness through an asymmetric encryption algorithm; submitting the immutable witness to a dual-track governance paradigm for processing, the dual-track governance paradigm including an enforcement track and a learning track; the enforcement track: a digital agent generates and dispatches immediate corrective instructions for real-time intervention on the current batch based on the comparison result of the witness and the current operating benchmark; the learning track: inputting the witness data into a predictive operating model for real-time inference and prediction of process deviation, and iteratively optimizing the model parameters based on the witness data and subsequent intervention and quality results; based on the immediate corrective instructions and the optimized model parameters, associating the flow events of subsequent supply chain nodes, constructing a product data management profile, and finally aggregating to form a derivable supply chain decision sandbox.

[0007] Preferably, the initial material profile specifically includes: creating a root data object as the starting point of product data management by recording an initial transaction on the distributed operating ledger, the data structure of the root data object including: raw egg purchase order number associated with the procurement process, enterprise digital credential identifying the identity of the supplier, geographic location data of the raw material origin, purchase quantity and unit of measure of batch size, and hash value of the electronic quarantine certificate issued by the third-party quarantine agency, which is used for subsequent certificate file verification.

[0008] Preferably, the production element association specifically includes: creating a relationship binding transaction on the distributed operating ledger, and injecting production context information into the batch product data management record; the input field of the relationship binding transaction references the hash value of the transaction where the initial job record is located, and the data payload field records the physical asset number of the assigned cellar unit, the version identifier of the adopted process recipe, the employee digital identity credential of the on-duty production operator, and the planned cellar start time; the output field of the transaction generates a unique identifier of the work-in-process job profile referenced in subsequent production links.

[0009] Preferably, the tamper-proof witness specifically comprises: periodically performing data snapshot operations based on a preset time frequency, continuously appending dynamic process data; the data snapshot collects the pH value of the material liquid, the temperature value of the storage environment and the humidity value of the storage environment at the current time point through the Internet of Things sensor, and obtains the cumulative duration from the storage starting point through the timer; the four parameter values collected are combined with the workpiece operation archive unique identifier to form a JSON format data block; the hash value of the data block is calculated, and the hash value is digitally signed using the private key bound to the storage unit to form the tamper-proof witness.

[0010] Preferably, the digital agent specifically comprises: after receiving the tamper-proof witness, analyzing the material liquid alkalinity and the storage temperature and humidity parameter values; comparing the analyzed parameter values with the multi-dimensional threshold values stored in the operation benchmark parameter library one by one, the multi-dimensional threshold values defining the normal fluctuation range of each parameter at different aging stages; when it is detected that any parameter value exceeds the normal fluctuation range of its corresponding stage, a structured immediate correction instruction is generated immediately, the data structure of the immediate correction instruction containing the network address of the target environment control device, the parameter item to be adjusted and the target value of the adjustment.

[0011] Preferably, the predictive operation model specifically comprises: extracting the full-cycle historical witness data of completed production batches from the distributed operation ledger to form a data set containing multiple time series; taking the finished product good rate recorded by the quality detection department as the label of supervised learning, and associating it with the corresponding time series data set; using the labeled data set to train the long short-term memory network containing attention mechanism; deploying the trained model, receiving the witness data sequence generated by the current production batch, and outputting the probability prediction value of the process deviation occurring within the next 24 hours in real time.

[0012] Preferably, the dual-track governance paradigm specifically comprises: a data feedback loop connecting the execution track and the learning track; the loop is configured to: after the execution track issues the immediate correction instruction, continuously capture the operation parameter change sequence within a period of time after the execution of the instruction; package the issued instruction content, the parameter change sequence after the execution and the final quality inspection result of the batch into a structured experience data set; input the experience data set as a new training sample into the predictive operation model of the learning track to perform incremental training of the model.

[0013] Preferably, the association of the flow events of the subsequent supply chain nodes specifically comprises: when the batch product completes the production link, an uplink operation is performed at each right and responsibility transfer point of the subsequent cellar, cleaning, grading, packaging, warehousing, delivery and logistics transportation; the uplink operation creates a new flow transaction, and the event type, the digital identity of the operation party, the geographic location data of the operation location, the transferred product quantity and the real-time temperature of the temperature-controlled logistics compartment are recorded in the transaction in the form of key-value pairs; the new transaction contains a hash pointer pointing to the previous link transaction, and a logistics and right change chain is constructed on the account book.

[0014] Preferably, the supply chain decision sandbox specifically comprises: performing simulation deduction comprising the following steps: calling a data aggregation function to extract the cost and time data of the historical batch in the production, logistics and warehousing links from the distributed operation account book, and constructing a parameterized supply chain digital twin model based on the same; receiving an external input simulation scenario through a command line interface, the simulation scenario being a set of adjustments to multiple parameters in the model; running a Monte Carlo simulation calculation in the digital twin model to quantize the probability distribution and expected value of the total cost, delivery cycle and expected profit rate of the batch product under the simulation scenario.

[0015] A blockchain-based production whole-process traceability system for preserved eggs comprises: A registration record module is configured to register a unique production instruction for a raw egg batch, generate an initial job record containing an initial material file, and record the same on a distributed operation account book; A production factor association module is configured to associate the initial job record with a cellar unit, a process recipe and an operator identity, and generate a work-in-process job file associated with the production factors; A witness generation module is configured to aggregate the work-in-process job file with captured material liquid alkalinity, cellar temperature and humidity, and curing duration, and generate an unforgeable witness through an asymmetric encryption algorithm; A dual-track governance module is configured to perform a track: a digital agent generates and issues an immediate correction instruction for real-time intervention of the current batch according to the comparison result of the witness and the current operation benchmark; and a learning track: the data of the witness is input into a predictive operation model to perform real-time inference and prediction of process deviation, and the model parameters are iteratively optimized based on the witness data and subsequent intervention and quality results; A product data management module is configured to associate the flow events of the subsequent supply chain nodes according to the immediate correction instruction and the optimized model parameters, and form a deducible supply chain decision sandbox through data aggregation, to jointly complete the product data management of the batch.

[0016] Compared with the prior art, the present application has the following advantages: 1、The present application builds a continuous and high-fidelity dynamic process data set for each production batch through the synergy of contextual binding and tamper-proof witness. On this data basis, the execution track in the dual-track governance paradigm is not a simple threshold alarm, but constitutes a closed-loop control system of real-time feedback and intervention. When facing batch-to-batch differences in raw material physical properties or fluctuations in external environmental factors, instead of relying on solidified static process instructions, the tamper-proof witness accurately captures the deviation of actual process parameters, and the execution track generates targeted correction instructions, enabling the production system to actively suppress the impact of various disturbances on the uniformity of the final product quality, demonstrating higher process control robustness.

[0017] 2、The present application combines tamper-proof witness with the learning track in the dual-track governance paradigm. The tamper-proof witness provides a high-fidelity process data source for machine learning, and the learning track establishes a quantitative mapping model from multi-dimensional process parameters to final product quality indicators based on these longitudinal historical data. This model not only optimizes static process parameter benchmarks based on historical data, but more importantly, uses the intervention measures of the execution track and the resulting process response data as key training samples. This enables the model to learn and solidify effective coping strategies for unexpected deviations, thereby dynamically adjusting the process path recommendations. This adaptive optimization capability enables the process model to continuously evolve to adapt to long-term changes in raw materials, equipment, and environment.

[0018] 3、The present application continuously associates the tamper-proof witness of the production process with subsequent supply chain flow events, and builds a supply chain decision sandbox at the top level, converting bottom-level technical data into top-level insights. Decision makers can dynamically value-layer and differentiate price products based on the fine-grained process quality indicators recorded in the witness data to meet the needs of different markets. In terms of risk management, the decision sandbox, with the efficiency and quality constraints of the real production process embedded in the model, can perform higher-precision supply chain risk stress testing and strategy simulation when facing external shocks such as demand fluctuations or logistics disruptions, outputting more valuable cost and delivery cycle impact assessments than traditional systems, thereby supporting enterprises to develop more flexible response strategies. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The overall method flowchart of a blockchain-based whole-process tracing method for preserved egg production is proposed for the present application; Figure 2 The structure schematic diagram of a blockchain-based whole-process tracing system for preserved egg production is proposed for the present application; Figure 3 The specific method flowchart of the supply chain decision sandbox in the embodiment of the present application. DETAILED DESCRIPTION

[0020] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0021] The present application provides a kind of based on the whole process traceability method of preserved egg production based on blockchain, technical scheme is as follows: A kind of based on the whole process traceability method of preserved egg production based on blockchain, the specific implementation steps of the method proposed in the present application include: S1, the unique production instruction of raw material egg batch to be produced is registered, including initial material file, record the source of raw material, supplier, batch number and other static information;Then, the production instruction and initial material file are packaged into initial job record, and broadcast to distributed operation account book for chain storage.

[0022] S2, the initial job record that has been chained is anchored with the physical storage unit occupied by this production, the process formula version number used and the digital identity of operator;By creating a binding transaction on the distributed operation account book, an in-process job archive associated with production factors is generated.

[0023] S3, in the storage and curing process, key operation parameters such as liquor alkalinity, storage environment temperature and humidity and cumulative curing time are captured periodically by Internet of Things sensors;Dynamic parameters are aggregated with in-process job archive and signed by asymmetric encryption algorithm to generate tamper-proof process witness and submit to distributed operation account book.

[0024] S4, the generated process witness is submitted to the dual-track governance paradigm for processing, in the execution track, the digital agent compares the real-time parameters in the witness with the preset operation benchmark, and if a deviation is found, it automatically generates and issues immediate correction instructions to intervene in the current production environment;In the learning track, the witness data sequence is input into the predictive operation model in real time for reasoning to predict process deviation, and the model parameters are iteratively optimized based on the witness data and subsequent intervention and quality results.

[0025] S5, based on the immediate correction instructions recorded in the execution track and the optimized model parameters in the learning track, the subsequent supply chain node events such as batch product out of pit, grading, packaging, logistics, etc. are continuously data associated and chained;The whole process data from production to sales are aggregated to form a deducible supply chain decision sand table for strategy simulation.

[0026] Embodiment one: This embodiment provides a specific application of a blockchain-based traceability method for the entire production process of preserved eggs. A company utilizes a distributed ledger environment based on the Hyperledger Fabric consortium blockchain and equips each storage unit in the storage workshop with industrial-grade pH and temperature / humidity sensors to perform fully digital and intelligent management of each batch of premium preserved eggs, from raw material acceptance to finished product delivery. For details, please refer to... Figure 1 .

[0027] Furthermore, a unique production instruction is registered for each batch of raw egg material to be produced, including an initial material file that records static information such as the source of the raw material, supplier, and batch number. Subsequently, the production instruction and the initial material file are packaged into an initial operation record and broadcast to the distributed operational ledger for on-chain storage and notarization. Corresponding to step S1 above, the specific process is as follows: The warehouse manager executes a production instruction creation process for a newly arrived batch of duck eggs. During this process, a root data object is generated as the initial material file. This object includes the purchase order number, the supplier's digital certificate ("Luyuan Ecological Farm"), the geographical location data of the raw material's origin, the purchase quantity, and a PDF quarantine certificate issued by the local quarantine department. The SHA-256 hash value of this certificate is then calculated and recorded. The initial transaction is submitted to the blockchain network, obtaining a unique transaction ID, which serves as the starting point for batch product data management.

[0028] By specifically defining the content of the initial material file, a complete and tamper-proof digital identity starting point is established for batch products, and the basic attributes of raw materials and vouchers are reliably solidified, providing a unique and reliable traceability root for data association in all subsequent stages.

[0029] Furthermore, the initial work records already on the blockchain are strategically anchored to the physical storage units occupied by this production, the process formula version number used, and the digital identity of the operator; by creating binding transactions on the distributed operations ledger, work-in-process operation files associated with production factors are generated. Corresponding to step S2 above, the specific process is as follows: The production workshop supervisor initiates a relationship binding transaction by binding batches to production resources. The input field references the ID of the initial transaction, and its data payload records the storage unit number "A-07", the process formula version number "Premium-V3.2", the on-duty operator's digital identity credential, and the planned storage start time. After the transaction is successfully recorded on the blockchain, a unique work-in-process record identifier is generated for that batch.

[0030] Through the relationship binding transaction, the abstract batch material is strongly associated with the concrete production resource and the responsibility subject, a production situation with a clear process context is created, and the subsequent collected process data can be accurately attributed to a specific batch, equipment and personnel, solving the problem of difficult to define production responsibility in traditional management.

[0031] Further, during the cellar aging process, key operating parameters such as liquor alkalinity, cellar environment temperature and humidity, and cumulative aging time are periodically captured by Internet of Things sensors; dynamic parameters are aggregated with the work-in-process job archive and signed by an asymmetric encryption algorithm to generate an unforgeable process witness, and submitted to a distributed operating ledger, corresponding to step S3, the specific process is: After the start of production, the edge computing gateway deployed in cellar unit A-07 automatically collects data from the pH sensor and temperature and humidity sensor every preset time period. In this invention, the monitoring of "liquor alkalinity" is realized by collecting the pH value of the liquor through the pH sensor. At a certain time point during the aging process, the "pH value 8.9, temperature 21.5℃, humidity 78%" is collected, and the cumulative time value is obtained through the timer. The gateway combines the four parameter values with the unique identification of the work-in-process job archive into a JSON data block, calculates its hash value, and calls the private key in the hardware encryption module bound to the cellar unit for signature, forming an unforgeable witness, and finally submitting to the blockchain through the API interface.

[0032] By specifically limiting the content and generation method of the unforgeable witness, the invisible production process is converted into a series of trusted, continuous and high-frequency digital snapshots, providing objective and irrefutable data basis for quality control and process deviation analysis.

[0033] A production execution closed-loop adaptive control paradigm, by situational binding of production batches and physical units, and based on periodically generated unforgeable witnesses, drives digital agents to autonomously generate and issue immediate correction instructions.

[0034] An operating benchmark parameter library containing the normal fluctuation range of key parameters at each process stage is constructed; a digital agent is configured to continuously monitor and analyze newly generated unforgeable witnesses; when the witness data deviates from the benchmark, the digital agent calculates the precise adjustment amount according to the preset rules or algorithm model, and generates a structured instruction containing the target device address, parameter item and target value, and sends it directly to the corresponding environment control executor. The operating benchmark parameter library can be dynamically updated by the learning track to adapt to external condition changes.

[0035] The paradigm changes the production process from passive monitoring to active, real-time adaptive adjustment, shortens the delay from problem discovery to solution, effectively suppresses the accumulation of small fluctuations in the production process into large quality defects, and significantly improves the stability of non-standard product production process and the consistency of finished product quality.

[0036] Further, the generated process witness is submitted to the dual-track governance paradigm for processing. In the execution track, the digital agent compares the real-time parameters in the witness with the preset operation benchmarks. If a deviation is found, it automatically generates and issues immediate correction instructions to intervene in the current production environment. In the learning track, the witness data sequence is input into the predictive operation model for inference to predict process deviations. Based on the witness data and subsequent intervention and quality results, the model parameters are iteratively optimized. Corresponding to step S4, the specific process is as follows: Execution track operation: In a subsequent data snapshot, a witness data shows that the temperature value exceeds the normal range specified in the operation benchmark library for the corresponding curing stage. The digital agent of the execution track is immediately triggered to automatically generate immediate correction instructions containing the target device address "A-07-TempController", the parameter to be adjusted "target temperature", and the adjustment value "21°C", and dispatch them to the environment control executor.

[0037] Through specific definition of the digital agent workflow, automatic, precise and real-time intervention of the production process is achieved, turning passive quality alarms into active closed-loop control, avoiding the expansion of quality problems caused by manual intervention delay or mistakes.

[0038] Learning track operation: The predictive operation model of the learning track receives this temperature over-standard witness data and immediately appends it to the data sequence of this batch. It performs real-time forward inference calculation without updating the model weights. Based on the current weights of the model, it outputs an updated prediction report in real time, quantifying the probability prediction values of multiple key process deviations ("alkalinity over-rising", "temperature continuously high", "humidity below threshold") occurring within the next 24 hours.

[0039] Through specific definition of the predictive operation model, prospective prediction of future production risks is achieved, improving the forward-looking nature of decision-making.

[0040] Dual-track collaborative operation: The correction instructions generated by the above-mentioned execution track, the parameter sequence captured after the execution of the temperature recovery normal within a period of time, and the quality inspection result of the batch being rated as "excellent product" are collectively packaged into a new structured experience data set and fed back to the learning track as training samples for incremental training of the model.

[0041] By specifically defining the data feedback loop in the dual-track governance paradigm, a self-optimizing mechanism that can learn from real intervention events is created, which not only performs control but also learns from the results of control, achieving continuous accumulation and self-evolution of process knowledge.

[0042] A process model incremental evolution method based on historical witness and intervention data, by structurally associating "process-intervention-result" data, performs continuous incremental training on the predictive operation model.

[0043] Extract the full-cycle witness data sequence of historical batches, the correction instruction sequence of track allocation, and the final quality inspection results from the distributed ledger; package these three types of data into a structured experience dataset, with intervention instructions as key feature variables; input this dataset as a new training sample into the predictive operation model, update the model weights through retraining or online learning algorithms, so that it can not only predict deviations but also evaluate the effectiveness of different intervention measures in a specific context.

[0044] This method builds a dynamic process knowledge base that can self-improve and continuously evolve, speeds up the iteration cycle of process optimization, and realizes the solidification and inheritance of effective intervention strategies.

[0045] Further, based on the immediate correction instruction records of track allocation and the model parameters after learning track optimization, the subsequent supply chain nodes such as batch product out-of-kiln, grading, packaging, logistics, etc. are continuously data-associated and chained; aggregate the full-process data from production to sales to form a deducible supply chain decision sand table for strategy simulation, corresponding to the above step S5, the specific process is: Supply chain associated operation: when the batch of products is out of the cellar, all the real-time correction instruction records generated during the production process of the batch and the model parameters output after the optimization of the learning track are called. The operator scans the cellar unit identification and the work-in-process job file identification through the mobile terminal, creates a transfer transaction of the type of out of cellar, and in the data structure of the transaction, in addition to recording the actual out of cellar product quantity, the operator digital identity and the geographic location data of the operation, a field is specially included to store the hash value of the above-mentioned correction instruction record and the optimized model parameter, so as to bind the production process data and the transfer event. Then, the products are transported to the cleaning station, and after the cleaning operation is completed, the cleaning transaction is created again by scanning the code to record the cleaning batch number and the process water quality parameter used. In the grading link, the quality inspector sorts the products according to the standard, and creates a grading transaction, which records the number of special grade products, superior products and qualified products corresponding to the batch in the form of key-value pairs, and associates the digital identity of the quality inspector. In the packaging link, the worker pastes a two-dimensional code containing batch information on each box of products, and performs a scan-in operation, which creates a transfer transaction containing "operation type: packaging and warehousing", "location: finished product A warehouse", "quantity: 500 boxes". Then, the products are out of the warehouse and handed over to "Shunjie Logistics" for transportation, and the transaction record is created again, which includes the digital identity of the carrier and the real-time temperature of the temperature-controlled logistics compartment, and points to the previous step of the warehousing transaction through the hash pointer.

[0046] By specific limitation of continuous associated operation, the boundary of product data management is seamlessly extended from the production domain to the entire supply chain, building an uninterrupted, clear responsibility end-to-end data chain, providing a unified and reliable data foundation for cross-enterprise logistics tracking, inventory management and responsibility definition.

[0047] Decision sand table operation: finally, a supply chain manager performs a decision sand table deduction, inputting the simulated scenario of "if the demand in the southern region increases by 30% next month, how should the production plan be adjusted". The sand table calls the data aggregation function, extracts historical data including the batch, builds a digital twin model, and gives the quantitative suggestion of "suggested to start two cellar units in advance, and to order an additional 20% of packaging materials" through Monte Carlo simulation calculation.

[0048] By specific limitation of the supply chain decision sand table workflow, the aggregated whole-process data is converted into forward-looking decision-making ability, enabling enterprises to make scientific risk assessment and resource planning based on real and complete operation data, improving the scientificity and accuracy of decision-making.

[0049] A supply chain risk deduction sand table method integrating high-fidelity production data, which quantitatively simulates and evaluates the impact of external disturbances on the supply chain by building a digital twin model based on real process data.

[0050] Extract and integrate full-chain data from raw materials to logistics from the distributed ledger through data aggregation functions; second, build a parameterized supply chain digital twin model, the core of which is that the model parameters of the production link (production cycle, probability distribution of good rate) are directly obtained by statistical or fitting of historical witnessed data, rather than static setting; finally, receive external input simulation scenarios (raw material price rise, transportation interruption), and through Monte Carlo simulation and other calculation methods, quantize the impact of the scenario on key indicators such as total cost and delivery cycle.

[0051] By introducing real data from the bottom of the production process, the accuracy of supply chain simulation and decision-making is greatly improved, enabling managers to stress test various complex risk scenarios and develop more forward-looking and robust strategies, thereby improving the resilience of the entire supply chain.

[0052] This embodiment converts a traditional preserved egg production process into an intelligent production process with transparent data, real-time controllable process, and continuously optimized process. The method not only solves the problem of traditional traceability black box through penetrating process data collection, but also builds an active quality control and process evolution closed loop through a dual-track governance paradigm, and finally converts technical data into decision-making capability, solving the core problems of quality control blind spots and insufficient decision-making data support for non-standard product production in the background technology.

[0053] Embodiment two: This embodiment provides a preserved egg production full-process traceability system based on a blockchain. The system is composed of hardware devices and software modules deployed in the production and management environment, and specific reference is made to Figure 2 .

[0054] The hardware part of the system includes: a desktop terminal with a barcode scanner configured in the raw material warehouse; a mobile industrial tablet equipped for the workshop director and operators in the production workshop; an industrial-grade pH sensor and a high-precision temperature and humidity sensor deployed inside each storage unit. These sensors are connected to an edge computing gateway responsible for data collection, local calculation and communication through a Modbus-RTU industrial bus. The gateway has a hardware security module built-in according to the TPM 2.0 standard, which is used to securely generate and store the encrypted private key of the storage unit; and a central server for carrying out core business logic.

[0055] The software modules of the system are deployed on the above hardware, specifically including: The registration record module runs in the form of a software application on the desktop terminal in the warehouse, provides a graphical user interface, and allows the warehouse administrator to input initial material profile information and call the blockchain client SDK to encapsulate the initial job record into a transaction and submit it to the chain.

[0056] Production factor association module: deployed in the form of a mobile application on the industrial tablet of the plant manager, this application is able to call the scanning function through the camera, read the identification of the batch and the cellar unit, and provide an interface to select the process recipe and confirm the operator identity, finally generate and submit the relationship binding transaction.

[0057] Witness generation module: embedded in the software of the edge computing gateway. This software communicates with the sensors through the industrial bus protocol, performs periodic data collection, JSON formatting, hash calculation and digital signature operations, and sends the generated tamper-proof witness to the blockchain node of the central server through a secure network protocol.

[0058] Dual-track governance module: deployed on the central server. Among them, the execution track unit is a continuously running background service that listens to new witnesses on the blockchain and compares them with a SQL database (as an operating benchmark parameter library), generates correction instructions through a preset logic and sends them to the environmental control executor through the Internet of Things protocol MQTT. The learning track unit is an application integrated with the machine learning framework TensorFlow, which builds a long short-term memory network LSTM model containing attention mechanism through the Keras front end, responsible for pulling samples from the blockchain historical data, performing offline training and online inference of the model. In this embodiment, the specific implementation architecture of the long short-term memory network model containing attention mechanism is as follows: Input layer: receives a fixed-length (past 24 time steps) time series data batch, with a data dimension of (batch number, 24, 4), where the 4 features correspond to the pH value of the liquid, the temperature value of the cellar, the humidity value of the cellar and the cumulative aging time.

[0059] LSTM layer: contains two layers of stacked LSTM units, used to extract long-term dependency features from time series.

[0060] Attention layer: applies a time attention mechanism on the output of the LSTM layer, which calculates the weight score of the hidden state of each time step, so that the model can dynamically focus on the historical moments that have a greater impact on future deviation prediction.

[0061] Output layer: input the weighted average feature vector into the fully connected layer, and finally output a multi-element prediction vector through the Sigmoid activation function, which contains the probability prediction values of "alkalinity deviation", "temperature deviation" and "humidity deviation" occurring within the next 24 hours (every 4 hours as a step).

[0062] Product data management module: deployed on the central server, and provides a web-based data analysis and visualization platform for managers. The platform is responsible for parsing and displaying the whole-process data on the blockchain, and has built-in simulation calculation engine of supply chain decision sand table. The engine can realize Monte Carlo simulation based on NumPy and SimPy libraries of Python, and managers can input simulation scenarios and obtain quantitative analysis results through the web interface.

[0063] Through the collaborative work of each module, based on high-fidelity process witness and dual-track governance paradigm, the system builds a robust production system that can actively suppress production disturbances and achieve closed-loop adaptive control. Through the continuous evolution of the learning track, a dynamic process knowledge base that can be self-optimized is established. Finally, the system seamlessly associates the underlying fine-grained production data with the upper-layer supply chain events, and converts technical data into precise insights through the decision sand table, thereby systematically improving the quality control precision, risk resistance ability and management decision level of the non-standard product supply chain.

[0064] Embodiment three: This embodiment further illustrates the dual-track governance paradigm. When an unalterable witness from cellar unit A-07 is submitted to the distributed ledger, the two units of the dual-track governance module are triggered simultaneously.

[0065] In the execution track, the workflow first parses the witness to extract real-time parameters (pH 8.9, temperature 21.5°C) and the corresponding in-process job archive identifier. Then, the identifier is used to query the operating benchmark parameter library to obtain the parameter threshold value of the current production stage. Next, a comparison operation is performed, and if the parameters deviate, a structured immediate correction instruction is generated according to the preset rules, which specifies the target device, the parameter to be adjusted, and the target value. Finally, the instruction is issued to the corresponding environmental control executor (temperature control relay or ventilation system controller) through the MQTT Internet of Things protocol to complete the closed-loop intervention.

[0066] In the learning track, its workflow is divided into two independent stages: 1. Real-time inference: When the witness data point is received, it is appended to the corresponding time series dataset and immediately input to the deployed LSTM model for inference. The output layer of the model is designed as a multi-element prediction vector, which can simultaneously output the probability prediction values of the three "key process deviations" of "alkalinity deviation", "temperature deviation" and "humidity deviation" occurring within the next 24 hours (every 4 hours as a step). 2. Incremental training: The data feedback loop is activated after the batch production is completed. During the production process, the learning track receives witness data in real time, as well as the corrective instructions issued by the execution track and the parameter sequence after the execution of the instructions. After the batch production is completed, the final quality inspection results (good rate label) are obtained. At this time, all these data are automatically packaged into a structured experience dataset as new training samples, which are input into the predictive operation model for incremental training and iterative optimization, so as to realize the continuous self-optimization of the model.

[0067] Embodiment Four This embodiment further details the generation process of the tamper-proof witness. The process is autonomously executed on the edge computing gateway of each cellar unit. The process starts with a timer task triggered, for example, every 30 minutes, and the gateway reads the raw data from the sensors through the industrial bus protocol. Then, the gateway obtains the cumulative duration since the start of the cellar through the internal timer, and encapsulates the four parameters (pH value, temperature value, humidity value, cumulative duration) with the in-process job archive unique identifier into a structured JSON object. The specific structure of the JSON object is: {“batch_id”:“ZZP-A07-001”,“timestamp”:1678886400,“ph”:8.9,“temp”:21.5,“humidity”:78,“duration_hours”:120.5}. The gateway performs SHA-256 hash algorithm on this JSON object to generate a unique digital fingerprint, and calls its built-in hardware security module to digitally sign the hash digest using the device private key stored therein to ensure the authenticity and integrity of the data. Finally, the original JSON data block and the generated digital signature are jointly constructed into the "data payload" of a transaction and broadcast to the blockchain network. In addition, the edge computing gateway is also configured with a network interruption response mechanism. When the gateway is disconnected from the blockchain node network of the central server, the gateway will start the local cache mode and continue to generate and sign the witness according to the preset frequency, and store it in the local non-volatile storage queue. Once the network connection is restored, the gateway will immediately push all the witness data packets in the cache queue to the distributed operation ledger in chronological order, ensuring that the production process data remains complete, continuous and time-correct even in the case of network fluctuations.

[0068] Example Five: This example further details the reasoning process of the supply chain decision sandbox, with reference to Figure 3 . The sandbox provides simulation capabilities for multiple scenarios: Scenario One: Cost Fluctuation Impact Assessment. Take the example of a supply chain manager who wants to assess the impact on annual profit if the main raw material supplier increases the supply price by 15%. The reasoning process is as follows. The reasoning starts with the manager initiating a new simulation on the decision sandbox interface. The data aggregation function of the sandbox is triggered, which automatically queries all the full-process data of all batches in the past year from the distributed ledger. Specifically, this function associates the raw material procurement cost through the "purchase order number" in the initial job record to the financial system, associates the good rate data through the "hierarchical" transaction, counts the actual aging time (funds occupation time) through the "unforgeable witness", and counts the transportation time and cost through the logistics "flow transaction". Based on these real historical data, the system builds a supply chain digital twin model containing multiple random parameters. The parameters of the production link ("aging period" is fitted as a normal distribution with a mean of 480 hours and a standard deviation of 20 hours, and "good rate" is fitted as a Beta distribution) are directly derived from real process witnesses, rather than static settings. Then, the manager inputs the scenario parameters through the command line interface to set the raw material cost increase coefficient (+15%). The simulation engine is then started, and tens of thousands of Monte Carlo simulation runs are performed. In each run, the engine will sample from the probability distributions of the digital twin model and substitute it into the cost and profit calculation formula with the scenario parameters. After all simulation runs are completed, the sandbox performs statistical analysis on the tens of thousands of profit results, and finally presents the expected impact and risk interval of this scenario on the annual profit to the manager in the form of probability distribution graph and key indicators, providing data support for subsequent decision-making (whether to raise prices or replace suppliers).

[0069] Scenario 2: Logistics Disruption Stress Test A supply chain manager wants to evaluate the impact of a 48-hour transportation disruption of the primary cold-chain logistics provider due to weather on the on-time delivery rate of finished goods and inventory cost. 1. Scenario input: The manager selects the "Logistics Disruption" simulation scenario in the sand table interface and inputs the parameters "Disruption Node: South China Logistics Center" and "Disruption Duration: 48 hours". 2. Data aggregation and model invocation: The sand table aggregation function invokes historical data and focuses on analyzing the timestamps of "Outbound" and "Transportation" transactions to build a baseline model of logistics turnover time. 3. Simulation calculation: The simulation engine forcibly adds a 48-hour delay variable to the logistics model. Subsequently, the engine simulates the order flow for the next quarter and deduces the "domino effect" of the delay on the delivery cycle of all subsequent batches, while calculating the increase in in-transit inventory (work-in-process cost) and warehouse backlog inventory due to delayed delivery. 4. Result output: The sand table finally outputs a stress test report: quantitatively predicts that the "on-time delivery rate" will drop from 98% to 75%, and the "average inventory cost" will increase by 22%. At the same time, the report will automatically recommend coping strategies based on the results, such as "immediately activate backup logistics provider B" or "suggest temporarily increase regional safety inventory by 30%". This deduction enables managers to develop contingency plans based on data, enhancing the resilience of the supply chain.

[0070] While embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based method for tracing the entire production process of preserved eggs, characterized in that, include: Register a unique production instruction for each batch of raw egg, generate an initial job record containing the initial material file, and record it on the distributed operations ledger; The initial work records are associated with the storage unit and process formula as production factors. By strategically anchoring the execution resources, work-in-process operation files associated with the production factors are generated. The process involves aggregating the work-in-process records with captured operational parameters and generating an immutable witness using an asymmetric encryption algorithm. The operational parameters include the alkalinity of the liquid material, the temperature and humidity of the storage cellar, and the maturation time. The immutable witness is submitted to the dual-track governance paradigm for processing, which includes an execution track and a learning track. Execution Track: Based on the comparison results between the witness and the current operational baseline, the digital agent generates and dispatches an immediate corrective instruction to intervene in the current batch in real time; study Track: Input the witnessed data into the predictive operation model to perform real-time inference to predict process deviations, and iteratively optimize the model parameters based on the witnessed data and subsequent intervention and quality results; Based on real-time correction instructions and optimized model parameters, the flow events of subsequent supply chain nodes are correlated to build product data management archives, and finally aggregated to form a predictable supply chain decision-making sandbox.

2. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The initial material file specifically includes: By recording initial transactions on a distributed operational ledger, a root data object is created as the starting point for product data management. The data structure of the root data object includes: raw egg purchase order number associated with the procurement process, enterprise digital certificate identifying the supplier, geographical location data of the raw material origin, purchase quantity and unit of measurement of the batch size, and hash value of electronic quarantine certificate issued by a third-party quarantine agency. The hash value is used for subsequent certificate file verification.

3. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The aforementioned production factor linkages specifically include: A relationship binding transaction is created on the distributed operations ledger to inject production context information into the batch's product data management record. The input field of the relationship binding transaction references the hash value of the transaction in which the initial job record is located, while the data payload field records the physical asset number of the assigned storage unit, the version identifier of the process formula used, the employee digital identity credential of the production operator on duty, and the planned storage start time. The output field of the transaction generates a unique identifier for the work-in-process operation file that is referenced in subsequent production stages.

4. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The immutable witness specifically includes: Based on a preset time frequency, data snapshot operations are periodically performed to continuously add dynamic process data. The data snapshots collect the pH value of the liquid material, the temperature value of the storage environment, and the humidity value of the storage environment at the current time point through IoT sensors, and obtain the cumulative duration from the start time of storage through a timer. The four parameter values ​​collected are combined with the unique identifier of the work-in-process file to form a data block in JSON format. The hash value of the data block is calculated, and the hash value is digitally signed using the private key bound to the storage unit to form the tamper-proof witness.

5. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The digital agent specifically includes: After receiving the tamper-proof witness, the alkalinity of the slurry and the temperature and humidity of the storage cellar are parsed out. The parsed parameter values ​​are compared one by one with the multidimensional thresholds stored in the operation benchmark parameter library. The multidimensional thresholds define the normal fluctuation range of each parameter at different maturation stages. When any parameter value is detected to exceed the normal fluctuation range of its corresponding stage, a structured real-time correction instruction is immediately generated. The data structure of the real-time correction instruction includes the network address of the target environmental control device, the parameter item to be adjusted, and the target value to be adjusted.

6. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The predictive operation model specifically includes: The distributed operations ledger extracts full-cycle historical witness data of completed production batches to form a dataset containing multiple time series. The finished product yield rate recorded by the quality inspection department for each batch is used as a label for supervised learning and associated with the corresponding time series dataset. The labeled dataset is used to train a long short-term memory network with an attention mechanism. The trained model is deployed to receive the witness data sequence generated for the current production batch and output the probability prediction value of process deviations in the next 24 hours in real time.

7. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The dual-track governance paradigm specifically includes: A data feedback loop connecting the execution track and the learning track is configured to: continuously capture the sequence of operational parameter changes over a period of time after the execution track issues an immediate correction instruction; package the issued instruction content, the sequence of parameter changes after execution, and the final quality inspection results of the batch into a structured experience dataset; and input the experience dataset as a new training sample into the predictive operation model of the learning track to perform incremental training of the model.

8. The blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The association of subsequent supply chain node transfer events specifically includes: After a batch of products completes its production process, an on-chain operation is performed at each point of responsibility handover during subsequent processes such as unloading from the cellar, cleaning, grading, packaging, warehousing, unloading, and logistics transportation. The on-chain operation creates a new transaction, which records the event type, the digital identity of the operator, the geographical location of the operation, the quantity of products handed over, and the real-time temperature of the temperature-controlled logistics vehicle in the form of key-value pairs. The new transaction contains a hash pointer to its preceding transaction, thus constructing a logistics and ownership change chain on the ledger.

9. A blockchain-based method for tracing the entire production process of preserved eggs according to claim 1, characterized in that, The supply chain decision-making sandbox specifically includes: The simulation process includes the following steps: calling a data aggregation function to extract historical batch cost and time data from the distributed operations ledger for production, logistics, and warehousing, and constructing a parameterized supply chain digital twin model based on this data; receiving externally input simulation scenarios through a command-line interface, where each simulation scenario is a set of adjustments to multiple parameters in the model; and running Monte Carlo simulation calculations in the digital twin model to quantitatively output the probability distribution and expected value of the simulation scenario for the total cost, delivery cycle, and expected profit margin of the batch products.

10. A blockchain-based traceability system for the entire production process of preserved eggs, characterized in that, include: Registration record module: Used to register a unique production instruction for each batch of raw egg, generate an initial operation record containing the initial material file, and record it on the distributed operations ledger; Production factor association module: used to associate the initial operation record with the storage unit, process formula and operator identity to generate work-in-process operation files associated with the production factors; Witness generation module: aggregates the work-in-process operation file with the captured material liquid alkalinity, cellar temperature and humidity, and maturation time; generates an immutable witness through an asymmetric encryption algorithm; Dual-track governance module: Execution track: The digital agent generates and dispatches immediate corrective instructions for real-time intervention in the current batch based on the comparison results between the witness and the current operating benchmark. study Track: Input the witnessed data into the predictive operation model to perform real-time inference to predict process deviations, and iteratively optimize the model parameters based on the witnessed data and subsequent intervention and quality results; Product Data Management Module: Based on real-time correction instructions and optimized model parameters, it associates the flow events of subsequent supply chain nodes and forms a predictable supply chain decision-making sandbox through data aggregation to jointly complete the batch product data management.

Images