Blockchain-based quality intelligent traceability system for green plum raw material supply chain

Abstract

This invention relates to the field of supply chain management technology, specifically a blockchain-based intelligent traceability system for the raw plum supply chain. The system comprises four modules: an access decision module, a cycle management module, a risk prediction module, and an automated execution module. By verifying and evaluating initial delivery information and external benchmark data, an inventory access decision is generated; based on the inventory access decision, the initial delivery information is analyzed to establish a full lifecycle management file; based on the correlation between the initial delivery information and external benchmark data, evolutionary trend deduction is performed, and the operational risk level is quantified and output; based on the changes in business nodes and the operational risk level in the full lifecycle management file, a control plan is matched, and the control plan dynamically optimizes preset business parameters. This invention achieves a transformation from passive traceability to proactive and intelligent management through risk prediction and closed-loop automatic control.

Description

Technical Field

[0001] This invention relates to the field of supply chain management technology, specifically to a blockchain-based intelligent traceability system for the quality of raw plums in the supply chain. Background Technology

[0002] In modern agricultural product supply chain management, including plums, accurate, continuous, and synchronized supply chain data is fundamental for refined production planning, efficient resource allocation, and reliable process control. Therefore, establishing a management model that supports real-time collaboration among multiple parties is crucial for optimizing business processes both within and across enterprises.

[0003] However, existing technologies have some problems. First, traditional management methods rely on post-event traceability. When quality issues such as damaged goods occur, although the system can record and query the data, the loss has already been incurred. Furthermore, when management strategies need to be adjusted according to changing circumstances, existing technologies typically rely on manual coordination, failing to achieve efficient closed-loop control and dynamic response, thus preventing timely and effective risk control. Additionally, in the raw material delivery stage, there is a lack of comprehensive quantitative assessment, affecting the quality of incoming materials.

[0004] In summary, existing technologies, when dealing with complex supply chains, suffer from the inability to provide proactive early warnings and optimize closed-loop processes accordingly. To address this, a blockchain-based intelligent traceability system for the raw plum supply chain is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a blockchain-based intelligent traceability system for the raw material supply chain of plums, used for supply chain management. To address the problems of existing technologies, this invention first acquires initial delivery information and external benchmark data. Through cross-validation and rationality assessment of the initial delivery information and external benchmark data, an access assessment score is generated. This score is then compared with a preset score threshold to determine inventory access decisions. Next, based on the inventory access decisions, the initial delivery information is analyzed to establish a full lifecycle management archive containing status and process data, and a record index is generated based on the process data. Second, based on preset key business handover points, transaction requests containing status data and record indexes are transmitted to the blockchain network. The correlation between process and status data is identified, and a supply chain decision model is constructed based on this correlation. This model is used to extrapolate the evolution trends of process and status data, quantifying and generating operational risk levels. Finally, based on the changes in business nodes and operational risk levels in the full lifecycle management archive, a control plan is matched, and the control plan dynamically optimizes preset business parameters. This invention transforms quality traceability from a reactive, after-the-fact mode to a proactive, intelligent management mode through risk prediction and closed-loop automatic control, thereby improving the efficiency of multi-party collaboration in the supply chain and the reliability of raw material quality.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The access decision module acquires initial delivery information and external benchmark data. By cross-validating and evaluating the rationality of the initial delivery information and external benchmark data, it generates an access assessment score and compares the access assessment score with a preset score threshold to determine the inventory access decision.

[0008] The cycle management module analyzes initial delivery information based on inventory access decisions, establishes a full lifecycle management file containing status data and process data, and generates a record index based on the process data; based on preset key business handover points, it transmits transaction requests containing status data and record indexes to the blockchain network.

[0009] The risk prediction module identifies the correlation between process data and status data, and builds a supply chain decision model based on the correlation. It then uses the supply chain decision model to extrapolate the evolution trend of process data and status data, and quantifies and generates operational risk levels.

[0010] The automated execution module matches control plans with changes in business nodes and operational risk levels based on the full lifecycle management archives. These control plans dynamically optimize preset business parameters.

[0011] Preferably, the initial delivery information includes quality results, key operational parameters, logistics tracking records, and warehouse management logs; the external benchmark data includes environmental data and supply chain planning benchmarks.

[0012] Preferably, the step of cross-validating and assessing the reasonableness of initial delivery information and external benchmark data to generate an access assessment score, and comparing the access assessment score with a preset scoring threshold to determine the inventory access decision, includes: generating a time-series score based on a time-series comparison of logistics tracking records and the supply chain plan benchmark; evaluating the quality results and preset quality standards to generate a parameter score; mapping warehouse management logs to a preset logical rule base, and performing logical deduction based on environmental data to generate a reasonableness score; weighted summing of the time-series score, parameter score, and reasonableness score to generate an access assessment score; and determining an inventory access decision based on the access assessment score and the preset scoring threshold, wherein the inventory access decision includes permission, rejection, and pending review.

[0013] Preferably, the cycle management module, based on the inventory access decision parsing of initial delivery information, establishes a full lifecycle management archive containing status data and process data, and generates a record index based on the process data, including: extracting the initial delivery information permitted by the inventory access decision; extracting the original string based on the initial delivery information; using a preset cryptographic algorithm to calculate the business code of the initial delivery information from the original string; encapsulating the business code, inventory access decision, and quality results into the status data of the full lifecycle management archive; using logistics tracking records, key operation parameters, and continuously generated process data as the process data of the full lifecycle management archive; establishing a batch processing archive of logistics and operation events based on the process data; transmitting the batch processing archive to off-chain distributed storage; obtaining the storage address of the batch processing archive; generating an integrity verification value for the process data using a preset algorithm; and generating a record index by combining the integrity verification value and the storage address.

[0014] Preferably, the cycle management module includes: responding to a traceability request initiated by a management terminal, the traceability request including a product identification code; mapping the product identification code to a business code, locating a full lifecycle management file containing the business code, retrieving status data and process data from the full lifecycle management file; integrating the status data and process data to generate a supply chain traceability report including visual charts, and sending the supply chain traceability report to the management terminal.

[0015] Preferably, the step of transmitting a transaction request containing status data and a record index to the blockchain network based on a preset key business handover point includes: constructing a logistics status change confirmation credential containing status data and a record index based on the preset key business handover point; completing business confirmation by digitally signing the logistics status change confirmation credential; combining the logistics status change confirmation credential and the digital signature to generate a standardized transaction request; the transaction request calling a preset smart contract function; and broadcasting the transaction request to the blockchain network for archiving and solidification.

[0016] Preferably, the risk prediction module identifies the correlation between process data and status data, and constructs a supply chain decision model based on the correlation, including: acquiring historical supply chain data containing historical process data and historical status data, using key operational parameters in the historical process data as independent variables and quality results in the historical status data as dependent variables; using multiple regression analysis to fit the historical supply chain data to generate the correlation between the quality results in the process data and the status data; the supply chain decision model is constructed based on a gradient boosting decision tree model and trained using cross-validation.

[0017] Preferably, the step of using a supply chain decision model to extrapolate the evolution trend of process data and status data and quantify and generate an operational risk level includes: using process data and status data as input to the supply chain decision model, performing calculations and extrapolations through the supply chain decision model, and outputting an estimated loss rate; and quantifying and determining the operational risk level by comparing the estimated loss rate with a preset risk threshold range.

[0018] Preferably, the automated execution module matches control plans based on changes in business nodes and operational risk levels in the full lifecycle management archive, including: matching the operational risk level with a hierarchical management protocol library based on changes in business nodes in the full lifecycle management archive, wherein the hierarchical management protocol library contains a list of control plans with execution priority sorting; scheduling and activating the control plan corresponding to the operational risk level, wherein the control plan is used to dynamically adjust preset business parameters, wherein the business parameters include quality inspection level and logistics control standards.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0020] 1. This invention establishes a multi-dimensional access decision-making mechanism to comprehensively evaluate the timeliness, quality results, and operational rationality of raw material delivery, generating a quantified comprehensive evaluation score. This mechanism upgrades the traditional receiving verification process, which relies on manual experience, to an objective and comprehensive automated evaluation process. By identifying and intercepting hidden quality risks caused by improper operations, it improves the overall quality foundation of incoming raw materials.

[0021] 2. This invention is based on a hybrid architecture combining blockchain and off-chain storage. It solidifies confirmation credentials for key business nodes on the blockchain, leveraging the immutability and traceability of blockchain technology to solve the problem of low collaborative efficiency caused by data silos and lack of trust in traditional supply chains. This solution provides a unified and trustworthy data-sharing fact repository, ensuring the economic efficiency of storing massive amounts of process data while guaranteeing transparency and clear definition of responsibilities in end-to-end supply chain processes, simplifying cross-entity auditing and collaborative workflows.

[0022] 3. This invention utilizes a supply chain decision-making model to establish the intrinsic correlation between key operational parameters and final quality results throughout the entire supply chain. The supply chain decision-making model takes real-time collected process data as input, performs evolutionary trend prediction, outputs a forward-looking operational risk level, and links with the business system through an automated execution module to form a closed-loop automatic control of "prediction-decision-execution". This approach transforms the management model from reactive post-event traceability to proactive pre-event intervention, reducing potential raw material losses, achieving forward-looking intelligent quality management, and improving the supply chain's resilience and economic efficiency. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the intelligent traceability system for the raw plum supply chain quality based on blockchain, as presented in this invention.

[0024] Figure 2 This is a flowchart of the access decision module in an embodiment of the present invention;

[0025] Figure 3 This is a flowchart illustrating the automated closed-loop control process in an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figures 1 to 3 This invention provides a blockchain-based intelligent traceability system for the raw plum supply chain, specifically including:

[0028] The access decision module acquires initial delivery information and external benchmark data. By cross-validating and evaluating the rationality of the initial delivery information and external benchmark data, it generates an access assessment score and compares the access assessment score with a preset score threshold to determine the inventory access decision.

[0029] The cycle management module analyzes initial delivery information based on inventory access decisions, establishes a full lifecycle management file containing status data and process data, and generates a record index based on the process data; based on preset key business handover points, it transmits transaction requests containing status data and record indexes to the blockchain network.

[0030] The risk prediction module identifies the correlation between process data and status data, and builds a supply chain decision model based on the correlation. It then uses the supply chain decision model to extrapolate the evolution trend of process data and status data, and quantifies and generates operational risk levels.

[0031] The automated execution module matches control plans with changes in business nodes and operational risk levels based on the full lifecycle management archives. These control plans dynamically optimize preset business parameters.

[0032] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0033] Example 1

[0034] This embodiment provides a specific application of a blockchain-based intelligent traceability system for the quality of plum raw material supply chain. Its typical application scenario is that "Fruit Company A" purchases plums from "Farm B" to produce high-end plum wine. In order to ensure the quality of plum raw materials, a blockchain-based intelligent traceability system for the quality of plum raw material supply chain is introduced.

[0035] See Figure 1 The system proposed in this invention specifically includes:

[0036] The access decision module acquires initial delivery information and external benchmark data. By cross-validating and evaluating the rationality of the initial delivery information and external benchmark data, it generates an access assessment score and compares the access assessment score with a preset score threshold to determine the inventory access decision.

[0037] The cycle management module analyzes initial delivery information based on inventory access decisions, establishes a full lifecycle management file containing status data and process data, and generates a record index based on the process data; based on preset key business handover points, it transmits transaction requests containing status data and record indexes to the blockchain network.

[0038] The risk prediction module identifies the correlation between process data and status data, builds a supply chain decision model based on the correlation, and uses the supply chain decision model to extrapolate the evolution trend of process data and status data, and quantifies and generates operational risk levels.

[0039] The automated execution module matches control plans with changes in business nodes and operational risk levels based on the full lifecycle management archives. These control plans dynamically optimize preset business parameters.

[0040] Furthermore, initial delivery information and external benchmark data are acquired. Through cross-validation and reasonableness assessment of the initial delivery information and external benchmark data, an access assessment score is generated. The access assessment score is compared with a preset score threshold to determine the inventory access decision. This corresponds to the system's access decision module, which specifically includes:

[0041] When the plum raw material with batch number "A01" is shipped from "Farm B", the system obtains initial delivery information and external benchmark data through the terminal. The initial delivery information includes quality results, key operational parameters, logistics tracking records, and warehouse management logs. Specifically, the quality results include the sugar content, firmness, and damaged fruit rate of the plums. For example, the average Brix content of the plums in batch "A01" is 11.5°, the fruit firmness is 85N, and there are no obvious pests or diseases. Key operational parameters include the picking time, the resting time before loading, and the cold chain temperature settings during transportation. For example, the picking time is 6:00 on the 15th, and the plums are loaded at 9:00 after pre-cooling treatment. The refrigeration setting of the transport vehicle is 5°C. The logistics tracking records are obtained based on the positioning module and include driving trajectory and timestamp data. For example, the transport vehicle leaves the farm at 9:05 and is expected to arrive at 17:00. The warehouse management log includes temporary storage environment records before the plum raw material is put into storage. External benchmark data includes environmental data and supply chain planning benchmarks. The system obtains environmental data through public meteorological APIs, such as temperatures along the transportation route ranging from 28°C to 32°C. The supply chain planning benchmarks include the delivery times and preset quality standards agreed upon in the purchase orders of "Fruit Company A". By integrating multi-dimensional information, the system achieves panoramic monitoring of the raw material delivery process and comprehensive assessment of potential hidden quality risks, thereby improving the accuracy and comprehensiveness of source quality assessment.

[0042] A time-series score is calculated by comparing the delivery time in the supply chain planning baseline with the actual arrival time recorded in the logistics tracking. The calculation method is as follows: a maximum score of 100 points is set, and within a preset tolerance time, the system calculates the score according to a linear deduction rule. This linear deduction rule can be 2.5 points deducted for every 5 minutes of delay. If the delay exceeds the tolerance time, the time-series score is 0. For example, if the scheduled time is 15:00 and the actual arrival time is 15:10, within the allowable half-hour error range, the time-series score can be 95 points. A parameter score is generated by comparing the quality results with preset quality standards. The parameter score generation method is as follows: first, a single score from 0 to 100 is generated for each independent quality result based on its degree of conformity to the standard; then, preset weighting coefficients are assigned to the single scores based on the importance of the quality results to the final product. The final parameter score is the weighted sum of the single scores. For example, for a batch with a quality result of "12% sugar content and 1% damage rate" (the preset quality standard is "sugar content greater than 11% and damage rate less than 2%), the system generates individual scores for the two independent quality results. The individual score for sugar content can be rated as 97.5 points, and the individual score for damage rate can be rated as 100 points. Subsequently, the system calculates a parameter score of 98 points by weighting the individual scores according to the preset weights that reflect the quality requirements of plum wine (the weight of sugar content can be set to 0.8, and the weight of damage rate can be set to 0.2).

[0043] The system maps warehouse management logs to a pre-defined logical rule base. This pre-defined logical rule base is a set of conditional rules based on historical supply chain data, established through statistical analysis and incorporating combinations of environmental and operational factors. The specific establishment process includes: First, collecting and integrating historical data from 500 procurement batches, including harvesting temperature, transportation time, storage humidity, and final damage rate. Second, the system uses multiple regression analysis to fit the historical data, identifying parameter combinations highly correlated with the final damage rate. For example, failure to pre-cool in time after harvesting under high-temperature conditions. The system outputs the quantitative impact coefficient of this parameter combination on the final damage rate. The multiple regression analysis method can be stepwise multiple regression, with the selection criterion being that the statistical significance level of the independent variable is less than 0.05. Based on the quantitative impact coefficient, the system converts risk into specific reasonableness deduction values ​​to form rules. The specific conversion process is as follows: Rank the parameters in descending order based on their quantified influence coefficients. Assign the highest pre-set deduction value to the highest-ranking parameter combination; for example, the highest deduction value could be 20 points. The deduction values ​​for the remaining parameter combinations are calculated linearly based on the ratio of the quantified influence coefficient to the highest influence coefficient. For example, if the quantified influence coefficient of a parameter combination is 50% of the highest quantified influence coefficient, then the deduction value is 50% of the highest deduction value. Finally, the specific rules generated through this conversion process are entered into the rule base. For example, Rule 1: If the ambient temperature on the day is greater than 30℃ and the post-harvest standing time is greater than 1 hour, then 20 points will be deducted from the reasonableness score. Rule 2: If the location record during transportation shows an unplanned stopover time of more than 30 minutes, then 15 points will be deducted from the reasonableness score.

[0044] The system uses 100 points as the initial reasonableness score and checks the rule base item by item. When the combination of warehouse management logs and environmental data matches the conditions of a rule, the corresponding deduction operation is executed, and the final reasonableness score is obtained. For example, if the system retrieves environmental data and finds that the temperature on that day was 32℃, while the warehouse log shows that the goods were loaded onto the truck 2 hours after harvesting without pre-cooling, violating rule one, 20 points will be deducted from the initial 100 points according to rule one, and the reasonableness score can be judged as 80 points. The time-series score, parameter score, and reasonableness score are weighted and summed. The weights can be dynamically adjusted according to the focus of the management stage. For example, when focusing on quality requirements, the time-series weight can be 0.2, the parameter weight can be 0.6, and the reasonableness weight can be 0.2, resulting in the admission assessment score. The specific method for adjusting the weights is as follows: The system periodically conducts retrospective analysis of historical data to identify the main factors that lead to quality risks; if the analysis shows that the loss caused by "Factor A" accounts for more than the preset management threshold in the total loss, for example, the loss caused by "Factor A" accounts for more than 30% in the total loss, the system automatically increases the weight of the corresponding score of "Factor A" and decreases the weight of other factors.

[0045] Furthermore, a dynamic weight adjustment submodule is introduced during cross-validation and rationality assessment. This submodule is a reinforcement learning model employing a deep Q-network model. The reinforcement learning model takes a vector containing the current batch time series, parameters, rationality score, and market demand forecast index as input. The market demand forecast index is obtained through the supply chain management system API. A preset weight adjustment strategy is used as the action space, such as "maintain the current weight" or "increase the weight of the parameter score by 10%". The model's reward function uses the expected profit of the batch raw materials as a positive reward and the additional warehousing costs and the opportunity cost of rejecting the batch raw materials as negative rewards, calculating the final reward value. By introducing the dynamic weight adjustment submodule, the weights of the admission decision can be adaptively adjusted according to the dynamic changes in market demand, improving the robustness and adaptability of the supply chain management system.

[0046] Based on expert experience and combined with historical data statistical analysis, a pre-set scoring threshold is established. Specifically, by analyzing the admission assessment scores of historical batches and the final actual quality results, a correlation chart between the admission assessment scores and the quality pass rate is drawn. Based on the correlation chart, a critical score balancing the risk of missed inspection and the risk of false rejection is selected as the scoring threshold. Inventory admission decisions are determined by comparing the admission assessment scores and the scoring threshold. These inventory admission decisions include approval, rejection, and pending review. For example, the "approval" threshold is 85 points, and the "pending review" threshold is 70 points. The inventory admission decision for batch "A01" plum raw materials is determined to be "approval". By standardizing the evaluation of the timing, parameters, and rationality of raw material delivery, an objective admission assessment score is quantified and generated as the basis for decision-making, achieving strict screening of incoming quality and improving the reliability and consistency of source quality control.

[0047] Furthermore, based on the initial delivery information analyzed by the inventory access decision, a full lifecycle management archive containing status data and process data is established, and a record index is generated based on the process data; based on preset key business handover points, transaction requests containing status data and record indexes are transmitted to the blockchain network, corresponding to the system's cycle management module, specifically including:

[0048] Based on the batch for which a "permission" decision has been obtained, the system extracts the raw string of initial delivery information, such as "A01," harvest time, and key quality results. This raw string is then combined according to a predefined key-value pairing rule to form a structured text string. Specifically, the text string contains field identifiers paired with numerical values, such as batch code, harvest timestamp, and quality result set. The system uses the SHA256 cryptographic algorithm to calculate the hash value of the text string, generating the business code for the current batch of raw materials in the supply chain. This business code, inventory access decision, and quality results are encapsulated as status data in a full lifecycle management archive. Logistics tracking records and key operational parameters are aggregated into process data for the archive. The system batch-processes and archives process data, such as minute-by-minute location and temperature readings. This batch-processed archive is then transmitted to the enterprise's off-chain distributed storage server, and the system obtains the storage address for the archive, such as IPFS. The system uses a predefined algorithm (e.g., SHA256) to generate an integrity check value for the process data. Finally, the system combines the storage address with the integrity check value to generate a record index. By establishing unique and trustworthy digital identities for raw materials and archiving process data off-chain, the completeness and traceability of end-to-end process data in the supply chain are ensured, improving the efficiency of full lifecycle management and the reliability of data traceability.

[0049] Based on the full lifecycle management archive and record index, when a management terminal initiates a traceability request containing a product identification code, the system maps the product identification code to the raw material's business code and retrieves the corresponding full lifecycle management archive. The system extracts status data from the full lifecycle management archive, such as raw material origin, warehousing decisions, and core quality results. Based on the record index in the full lifecycle management archive, it retrieves process data from distributed off-chain storage, such as harvesting time, transportation trajectory, and temperature changes. The system integrates and visualizes the status and process data to generate a supply chain traceability report, which is then sent to the terminal device. By mapping product identification codes to raw material business codes and integrating and displaying full lifecycle data, end-to-end transparency and credibility are achieved, providing data support for product quality and value.

[0050] Based on key business handover points, such as "confirmation of receipt," a logistics status change confirmation credential is constructed. This credential includes status data and a record index. A digital certificate is used to digitally sign the logistics status change confirmation credential. The system combines the credential and the digital signature into a standardized transaction request and broadcasts it to the blockchain network. For example, the blockchain may employ a consortium blockchain architecture. The transaction request is a data packet containing a specific purpose function call and parameters. This data packet defines the name of the target smart contract, the type of business operation to be performed, and the specific business data containing the logistics status change confirmation credential and the digital signature. Consensus nodes in the blockchain network verify the data packet by executing a pre-defined smart contract. Upon successful verification, the transaction request is archived and solidified. The smart contract includes a data structure recording batch status and a function with access control. The batch status includes a batch ID, owner address, operational risk level, and a status flag indicating whether the asset is locked. The function can be an "update risk status" function that can only be called by the system administrator. By using digital signature mechanisms at key business handover points and solidifying credentials on the blockchain, clear responsibilities are defined, providing a unified and trustworthy database for supply chain collaboration and improving the efficiency and transparency of supply chain coordination.

[0051] Furthermore, the correlation between process data and status data is identified, and a supply chain decision-making model is constructed based on this correlation. This model is then used to extrapolate the evolutionary trends of the process and status data, quantifying and generating operational risk levels. This corresponds to the system's risk prediction module, which specifically includes:

[0052] A supply chain decision-making model is constructed, specifically a gradient boosting decision tree model. The construction process involves: First, defining the model's input feature set, which consists of key operational parameters. These parameters include the ambient temperature during harvesting, the resting time before pre-cooling, the duration and temperature of pre-cooling, the average temperature during transportation, the cumulative duration of temperature exceeding a preset standard during transportation, and the average humidity during transportation. For example, the preset temperature standard could be 5°C. Second, defining the model's target output variable, which is the estimated spoilage rate used to quantify the final quality result. For example, the estimated spoilage rate can be defined as the percentage of damaged fruit (including rotten, broken, and diseased fruit) out of the total weight of the current batch after the raw materials are stored for 7 days in a standard cold storage facility with a temperature of 5°C and a fluctuation range not exceeding 0.5°C, and a humidity of 90% and a fluctuation range not exceeding 3%). Then, the model is trained using a structured dataset containing 1000 historical batches, covering the key operational parameters and the target output.

[0053] Based on historical supply chain data, a supply chain decision-making model is trained. First, the data is preprocessed, including normalizing continuous values, one-hot encoding categorical features, using the 3-sigma principle to identify and remove obvious outliers during data collection, and interpolating missing data using the mean of adjacent time points. Second, a five-fold cross-validation method is used to evaluate the performance of the supply chain decision-making model on different data subsets and prevent overfitting. After model training, performance is evaluated using an independent validation dataset. The supply chain decision-making model is deployed when the average absolute error between the estimated and actual loss rates is less than 0.5%. The gradient boosting decision tree model can be configured with a learning rate of 0.01, 500 decision trees, and a maximum depth of 5 layers per tree. The supply chain decision-making model includes an update mechanism; for example, after processing 200 new batches, the system automatically incorporates new data into the dataset and triggers a retraining process. By constructing and training a quantitative supply chain decision-making model, post-event traceability is transformed into proactive pre-event intervention, which enhances the scientific nature and foresight of supply chain quality control, ensures the reliability and accuracy of the decision-making model, and provides a solid and reliable foundation for subsequent risk warning.

[0054] Using the process and status data of the current batch of raw materials as input, a supply chain decision-making model is used for calculation and extrapolation, outputting a quantitative prediction of future quality status. For example, the model might predict that although the current batch of raw materials is of acceptable quality upon arrival at the warehouse, due to slight temperature fluctuations during transportation, the estimated loss rate of the current batch of raw materials will reach 2.5% within the next 7 days. The estimated loss rate output by the supply chain decision-making model is compared with a preset risk threshold range to quantitatively determine the operational risk level of the current batch. For example, the preset risk threshold range could be: an estimated loss rate below 1% is considered "low" risk; an estimated loss rate between 1% and 3% is defined as "medium" risk; and an estimated loss rate above 3% is considered "high" risk. The current batch's estimated loss rate of 2.5% results in a quantitative output of "medium" operational risk level. By extrapolating the evolutionary trends of process data and quantitatively outputting a forward-looking operational risk level, the model enables early warning and identification of potential loss risks, enhancing the company's ability to proactively avoid quality risks and reduce economic losses.

[0055] Furthermore, a comprehensive evaluation mechanism is introduced during the evolution trend projection process. By using environmental parameters, including transportation mileage and vehicle type, as new input variables, a carbon footprint score is calculated, outputting a multi-dimensional decision support vector containing the estimated cargo damage rate and carbon footprint score. Specifically, by connecting to the transportation management system and calling preset transportation plans, the system analyzes the transportation plans to obtain transportation mileage and vehicle type; it uses a carbon emission factor database to quantify and output the total carbon emission estimate of the transportation plan; it compares the total carbon emission estimate with the enterprise's internal carbon emission average to generate a deviation value, which is then linearly converted into a carbon footprint score. The system can compare the estimated cargo damage rate and carbon footprint score of the transportation plan and select a plan based on preset decision priorities, including low cargo damage rate and low carbon footprint score. By introducing a comprehensive evaluation mechanism, risk assessment is expanded from a single cargo damage rate dimension to a comprehensive dimension including both cargo damage rate and carbon footprint, improving the comprehensiveness and foresight of supply chain risk management.

[0056] Furthermore, based on the business node changes and operational risk levels matched with the full lifecycle management archives, a control plan is established. This control plan dynamically optimizes preset business parameters and corresponds to the system's automated execution module, specifically including:

[0057] The business node identifiers in the full lifecycle management file are processed through predefined standard operating procedures. These business nodes include pending warehousing, in stock, outbound, in transit, delivered and awaiting receipt, and order completed. Changes to node identifiers are recorded on the blockchain as new transaction requests. Based on changes in the nodes of the full lifecycle management file, for example, when the business node status of the current batch of raw materials changes from "pending warehousing" to "in stock," a preset business process is triggered. The system automatically matches the operational risk level with a preset hierarchical management protocol library, which defines the correspondence between operational risk levels and control plans.

[0058] The hierarchical management protocol library is a structured data table. Its establishment is based on cost-benefit assessment, and the specific steps are as follows: First, based on a pre-set cost database, the costs of the operational units constituting potential control plans are retrieved. This cost database pre-defines and quantifies the operational costs for specific operations and is dynamically updated. These operational costs are periodically entered based on financial data; for example, adding a comprehensive quality inspection or calling upon a Class A cold chain vehicle service. The costs include labor, material consumption, and equipment usage fees. The total execution cost of the potential control plan is obtained by summarizing these costs. Second, based on the process data of the current batch, the system calls upon the supply chain decision model to predict the future baseline estimated loss rate if no control plan is implemented. Next, the potential control plan is transformed into hypothetical operational parameters. These hypothetical operational parameters are used as input to the supply chain decision model to generate an optimized estimated loss rate for the potential control plan. The baseline estimated loss rate and the optimized estimated loss rate are subtracted to obtain the estimated loss rate reduction. Based on this reduction, the expected avoidable loss is calculated. The expected avoidable loss is compared with the total execution cost. When the expected avoidable loss is higher than the total execution cost, and the ratio of the expected avoidable loss to the total execution cost exceeds a preset adoption threshold, the potential control plan is entered into the hierarchical management agreement library. Simultaneously, a correlation is established between the potential control plan and the operational risk level. The preset adoption threshold is greater than 1, and its specific value can be preset according to the company's risk appetite and return on investment requirements. For example, if an investment of 1 unit of cost is required to recover at least 1.2 units of potential loss, the threshold can be set to 1.2.

[0059] When multiple control plans are effective, their execution priorities are set according to their cost-benefit ratio. When the system performs a matching action, it filters eligible control plans based on a hierarchical management protocol library, taking into account the overall operational risk level and the current raw material business node. The hierarchical management protocol library pre-stores a list of control plans sorted by execution priority for each operational risk level. When the system matches an operational risk level, it defaults to executing the control plan with the highest priority for that level. Simultaneously, the system recalculates the expected risk level after executing the highest-priority control plan. If the expected risk level does not decrease to "low," it continues to match and execute the next-highest priority control plans in the list until the operational risk level is met. For example, the record for a "medium" risk level might include: enhanced sampling and upgraded transfer procedures. The system parses the control plans into executable business parameter instructions. These business parameters are variable parameters set to achieve specific control objectives, and their specific form can be adjusted according to the business scenario. For example, business parameters can be quality inspection levels and logistics control standards. The system sends instructions to the warehouse management system API to adjust the quality inspection level of batch A01 from "standard sampling inspection" to "enhanced sampling inspection"; and sends instructions to the transportation management system API to "call high-standard vehicles with a temperature control accuracy of ±0.5℃ when batch A01 is internally transferred." By automatically matching operational risk levels with the hierarchical management protocol library and dynamically optimizing business parameters through scheduling and control plans, closed-loop automatic control is achieved, ensuring the timeliness and accuracy of control measures and reducing delays and oversights caused by manual intervention.

[0060] Furthermore, a deep integration mechanism with blockchain smart contracts is introduced during the dynamic optimization of business parameters. This deep integration mechanism is based on non-fungible tokens (NFTs) uniquely bound to physical batches. This binding is achieved when raw materials are received by writing business codes into the NFT metadata and physical QR codes. Specifically, the deep integration mechanism involves the smart contract automatically updating the on-chain status of the NFTs according to the operational risk level, and performing hierarchical management based on the updated on-chain status. The on-chain status includes high-risk pending review, medium-risk requiring supervision, and low-risk accepted. The hierarchical management includes transferring the NFTs to a temporary locked address for manual review and unlocking, sending early warning notifications to preset monitoring addresses, and increasing the collaboration priority points associated with suppliers. By binding dynamic risk prediction results with on-chain rules, an automated closed loop from "internal control" to "external collaboration" is achieved, improving the efficiency and reliability of supply chain collaboration.

[0061] This invention achieves a transformation from "passive traceability" to "proactive intelligent management" by constructing a closed-loop architecture of access decision-making, cycle management, risk prediction and automated execution, thereby improving the risk resistance and collaborative efficiency of the supply chain.

[0062] Example 2

[0063] In this embodiment, a blockchain-based intelligent traceability system for the raw plum supply chain is applied to the supply chain in Embodiment 1. Fruit Company A receives raw plums with batch number "A02". The initial delivery information shows that the average Brix content is 10.8° and the damaged fruit rate is 2.5%, both slightly lower than the preset quality standard.

[0064] See Figure 2 The system calculates an admission assessment score of 83.6 points by weighting and summing the scores for time sequence, parameters, and rationality. This score falls within a preset "pending review" range, which could be 70 to 84.9 points. The system then generates a "pending review" inventory admission decision. Based on this decision, a manual re-inspection process is triggered. Upon re-inspection, the current batch is identified as having 70% qualified products and 30% substandard products. Subsequently, the batch is digitally split: a full lifecycle management file with the business code "A02-PASS" is created for the 70% qualified products, with the business node identified as "in stock"; a file with the business code "A02-JUICE" is created for the 30% substandard products, with the business node identified as "pending disposal." The process data for the new file is archived to off-chain storage, such as the re-inspection report, and a corresponding record index is generated. Finally, based on the key business handover point of "manual re-inspection," the system constructs and broadcasts two digitally signed transaction requests to the blockchain network, archiving and solidifying the results of the current batch grading and differentiated disposal. By digitally splitting physical batches of mixed quality, assigning independent digital identities to materials of different grades for tracking, the system maximizes the value of raw materials and achieves differentiated disposal, improving the precision of inventory management and the flexibility of asset utilization.

[0065] Example 3

[0066] This embodiment applies a blockchain-based intelligent traceability system for the raw plum supply chain to the supply chain in Embodiment 1. The raw plums with batch number "A03" need to be transported from "Farm B" to "Cold Chain Company C" for temporary storage before being picked up by "Fruit Company A".

[0067] See Figure 3The system adds the "In Stock (Company C)" status to the business node identification flow path of the full lifecycle management archives. At key business handover points of "ownership transfer," such as when the "A03" batch of plums arrives at Company C and Company A picks up the goods from Company C, the relevant parties complete the business confirmation through digital signatures and broadcast the transaction request containing status data and record indexes to the blockchain network composed of parties A, B, and C. The system's risk prediction module obtains initial delivery information and continuously integrates process data generated during the storage period of "Company C." For example, if there is a cumulative temperature fluctuation in Company C's warehouse exceeding the standard by 1.5°C for 3 hours, the model will extrapolate based on the temperature fluctuation evolution trend, predicting that the estimated damage rate of this batch will increase from 0.8% before entering the warehouse to 2.2%, and update the operational risk level from "low" to "medium." Based on the operational risk level, the automated execution module matches the hierarchical management protocol library, automatically triggers the control plan, issues an "enhanced sampling inspection" instruction to Company A's quality inspection department, and can also notify Company C's settlement system via API that there are quality defects in this warehousing service, providing data support for subsequent processing. By continuously integrating process data, dynamic and cross-entity risk assessments are achieved, establishing a unified and reliable data foundation for multi-party collaboration and enhancing end-to-end transparency and accountability. At the same time, by automatically converting quantified risk assessment results into specific instructions, the fairness and efficiency of multi-party commercial settlements are improved.

[0068] After undergoing enhanced sampling inspection, batch "A03" was transferred to Cold Storage No. 1 of "Fruit Company A" for storage, and the business node identifier was updated to "In Stock". Based on the change in business node, the system assesses the raw materials in stock. For example, 7 days after the business node change, the system collects the process data associated with batch "A03" within those 7 days into the full lifecycle management file, such as the daily average temperature and humidity records and the number of times the cold storage door was opened and closed in Cold Storage No. 1. In the assessment on the 7th day, the system uses the process data from the previous 7 days as input, extrapolates through the supply chain decision model, and generates prediction results. For example, the model predicts that although the average temperature meets the standard, due to the persistently slightly high humidity, the estimated loss rate will rise to 2.8% within the original planned remaining storage period, which is in the threshold range of "medium" risk. The system dynamically adjusts the operational risk level of the current batch from "low" to "medium". After the assessment, the "medium" risk level triggers the automated execution module and matches the "proactive inventory optimization" control plan. Business parameters are optimized to adjust production plan priorities. The system automatically sends an API request to the manufacturing execution system (MES) of "Fruit Company A." This API request contains instructions, such as raising the outbound priority of batch "A03" to "high." Upon receiving the instructions, the MES will automatically prioritize the use of "A03" batch of plum raw materials in the next week's production schedule, ensuring its priority consumption in the next production cycle. By continuously analyzing process data, the system proactively predicts future quality deterioration risks, establishing a closed loop between quality management and production planning, reducing potential losses, and improving inventory turnover efficiency and asset preservation capabilities.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based intelligent traceability system for the raw plum supply chain, characterized in that: include: The access decision module acquires initial delivery information and external benchmark data. The initial delivery information includes quality results, key operational parameters, logistics tracking records, and warehouse management logs. The external benchmark data includes environmental data and supply chain planning benchmarks. By cross-validating and evaluating the initial delivery information and external benchmark data, an access assessment score is generated. The access assessment score is compared with a preset scoring threshold to determine the inventory access decision. This includes: generating a time-series score based on a time-series comparison of logistics tracking records and the supply chain planning benchmarks; and evaluating the quality results against preset quality standards to generate parameter scores. Specifically, the quality results include the quality of the plums... Sugar content, firmness, and damaged fruit rate; Warehouse management logs are mapped to a preset logical rule base, and logical deduction is performed in conjunction with environmental data to generate a reasonableness score; the preset logical rule base is a set of conditional rules containing environmental and operational combinations, established based on statistical analysis of historical supply chain data; the time-series score, parameter score, and reasonableness score are weighted and summed to generate an access assessment score; based on the access assessment score and a preset scoring threshold, an inventory access decision is determined, including permission, rejection, and pending review; by digitally decomposing physical batches of mixed quality, different grades of materials are assigned independent digital identities for tracking; The cycle management module analyzes initial delivery information based on inventory access decisions, establishes a full lifecycle management file containing status data and process data, and generates a record index based on the process data; based on preset key business handover points, it transmits transaction requests containing status data and record indexes to the blockchain network. The risk prediction module identifies the correlation between process data and status data, and builds a supply chain decision model based on the correlation. It uses the supply chain decision model to extrapolate the evolution trend of process data and status data, and quantifies and generates the operational risk level. In the process of extrapolating the evolution trend, it calculates the carbon footprint score by using environmental parameters including transportation mileage and transportation vehicle type as new input variables, and outputs a multi-dimensional decision support vector containing the estimated cargo loss rate and carbon footprint score. The automated execution module matches control plans with changes in business nodes and operational risk levels based on the full lifecycle management archives. These control plans dynamically optimize preset business parameters.

2. The intelligent traceability system for plum raw material supply chain quality based on blockchain as described in claim 1, characterized in that, The cycle management module, based on the inventory access decision parsing of initial delivery information, establishes a full lifecycle management archive containing status data and process data, and generates a record index based on the process data. This includes: extracting initial delivery information where the inventory access decision is approved; extracting the original string based on the initial delivery information; using a preset cryptographic algorithm to calculate the business code of the initial delivery information from the original string; encapsulating the business code, inventory access decision, and quality results into the status data of the full lifecycle management archive; using logistics tracking records, key operational parameters, and continuously generated process data as the process data of the full lifecycle management archive; establishing a batch processing archive of logistics and operational events based on the process data; transmitting the batch processing archive to off-chain distributed storage and obtaining the storage address of the batch processing archive; generating an integrity verification value for the process data using a preset algorithm; and generating a record index by combining the integrity verification value and the storage address.

3. The intelligent traceability system for the raw plum supply chain quality based on blockchain as described in claim 2, characterized in that, The cycle management module includes: responding to a traceability request initiated by the management terminal, the traceability request containing a product identification code; mapping the product identification code to a business code, locating the full lifecycle management file containing the business code, retrieving status data and process data from the full lifecycle management file; integrating the status data and process data to generate a supply chain traceability report containing visual charts, and sending the supply chain traceability report to the management terminal.

4. The intelligent traceability system for the raw plum supply chain quality based on blockchain as described in claim 1, characterized in that, The process of transmitting a transaction request containing status data and a record index to the blockchain network based on a preset key business handover point includes: constructing a logistics status change confirmation credential containing status data and a record index based on the preset key business handover point; completing business confirmation by digitally signing the logistics status change confirmation credential; combining the logistics status change confirmation credential and the digital signature to generate a standardized transaction request; the transaction request calling a preset smart contract function; and broadcasting the transaction request to the blockchain network for archiving and solidification.

5. The intelligent traceability system for the raw plum supply chain quality based on blockchain as described in claim 1, characterized in that, The risk prediction module identifies the correlation between process data and status data, and constructs a supply chain decision model based on the correlation. This includes: acquiring historical supply chain data containing historical process data and historical status data; using key operational parameters in the historical process data as independent variables and quality results in the historical status data as dependent variables; using multiple regression analysis to fit the historical supply chain data and generate the correlation between the quality results in the process data and the status data; and constructing the supply chain decision model based on a gradient boosting decision tree model and training it using cross-validation.

6. The intelligent traceability system for the raw plum supply chain quality based on blockchain as described in claim 1, characterized in that, The method of using a supply chain decision model to extrapolate the evolution trend of process data and status data and quantify and generate operational risk levels includes: using process data and status data as input to the supply chain decision model, performing calculations and extrapolations through the supply chain decision model, and outputting an estimated loss rate; and quantifying and determining the operational risk level by comparing the estimated loss rate with a preset risk threshold range.

7. The intelligent traceability system for the raw plum supply chain quality based on blockchain as described in claim 1, characterized in that, The automated execution module matches control plans based on changes in business nodes and operational risk levels in the full lifecycle management archive. This includes: matching the operational risk level with a hierarchical management protocol library based on changes in business nodes in the full lifecycle management archive, wherein the hierarchical management protocol library contains a list of control plans with execution priority sorting; scheduling and activating the control plan corresponding to the operational risk level, wherein the control plan is used to dynamically adjust preset business parameters, wherein the business parameters include quality inspection level and logistics control standards.

Images