A local pig breed multi-node block chain collaborative breeding method

Abstract

The application relates to the field of animal genetic breeding technology, and discloses a local pig breed multi-node block chain collaborative breeding method. According to the method, key data abstracts, evaluation results, model versions and breeding decisions are chained and left marks in a window period through a permissioned consortium chain and a consensus verification mechanism, the source is clear, tamper-proof and auditable, the problem that data standards are different, difficult to trace and difficult to trust in traditional multi-subject collaboration is solved, in addition, an AI multi-factor breeding evaluation system can simultaneously process multiple traits such as meat quality, growth, reproduction, health resistance and environmental management factors, can depict the nonlinear correlation between complex traits, can improve the evaluation precision and stability of the comprehensive genetic value of local pig breeds, can reduce the risk of decision-making errors caused by single-point data or single-algorithm deviation, can realize cross-node alignment and joint use of dispersed data through a rolling window and a labeled archive, and can make up for the short board of insufficient samples of a single breeding unit.

Description

Technical Field

[0001] This invention relates to the field of animal genetics and breeding technology, specifically to a multi-node blockchain collaborative breeding method for local pig breeds. Background Technology

[0002] Currently, large-scale pig breeding in China mainly relies on linear genetic evaluation models such as BLUP (Best Linear Unbiased Prediction) to estimate finite traits such as meat yield and growth rate. The BLUP method requires a large amount of phenotypic data from related individuals, often making it difficult to achieve sufficiently high evaluation accuracy in small-scale local pig breeds. Furthermore, breeding objectives involve balanced improvement of multiple traits, such as increasing growth rate and fertility while maintaining meat quality and disease resistance. Such multi-objective selection is difficult to fully achieve using a single index or linear model. Traditional breeding practices rely more on experience and single-trait selection, which may lead to neglecting some aspects and weakening the valuable quality advantages of local breeds.

[0003] In recent years, blockchain technology has been increasingly applied to data management and traceability in the agricultural sector. For example, research has constructed a smart agriculture system combining the Internet of Things (IoT), blockchain, and AI. This system uses sensor networks to collect production data in real time, blockchain to ensure data reliability, and AI for decision support, significantly improving agricultural production efficiency and accuracy in actual tests. In the livestock industry, blockchain has been used to record livestock identity and pedigree, and to prevent counterfeiting and trace meat quality. However, its application in genetic breeding is still in the exploratory stage. Traditional blockchain breeding schemes often directly copy competitive breeding models, resulting in a rigid mechanism. For example, the basic breeding population and each generation of breeding populations are simply labeled as V0, V1…Vn generations, and each generation selects a fixed number (e.g., the top 500) of the best-performing individuals as candidate breeding pigs. This mechanism is not suitable for local breeds with limited numbers of pigs: on the one hand, fixed selection quotas may miss individuals with special superior traits, and it is difficult to gather such a large number of candidates in a small population; on the other hand, the centralized selection model by generation lacks continuity and cannot reflect new data from individuals at different growth stages in a timely manner. Existing large-scale breeding data platforms mostly serve imported breeds or large-scale commercial strains. Their genetic evaluation parameters and breeding programs may not be suitable for the production characteristics of local pig breeds, and local breeding units lack effective tools to share data and collaboratively improve breeds. This has led to slow progress of local germplasm in the modern breeding system, and there is an urgent need for a new technical solution to integrate resources from multiple parties and improve breeding efficiency. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a multi-node blockchain collaborative breeding method for local pig breeds. This method has advantages such as solving problems like data isolation, limited evaluation methods, low improvement efficiency, and platform incompatibility in existing local pig breed breeding processes, thus resolving the aforementioned technical issues.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multi-node blockchain collaborative breeding method for local pig breeds, comprising the following steps: S1: Construct a consortium blockchain network among multiple breeding unit nodes; S2: Collect multi-trait measurement data of local pig breeds and environmental factors corresponding to each breeding unit node, clean and standardize them, generate corresponding tags and data summaries, and write the tags and data summaries into the consortium blockchain network. S3: Deploy an AI multi-factor breeding evaluation system at each breeding unit node, and conduct a comprehensive evaluation of breeding pigs based on the local data of each breeding unit node, output the individual comprehensive breeding value, and input the individual comprehensive breeding value into the consortium blockchain network; S4: Set up sampling windows in the AI ​​multi-factor breeding evaluation system, and trigger dynamic evaluation block generation and verification at the end of each sampling window; S5: After each sampling window ends, the newly added data within the sampling window is packaged into a dynamic evaluation block and submitted to the consortium blockchain network; S6: Perform consensus verification and validation on the newly added dynamic evaluation block, and write the newly added dynamic evaluation block into the blockchain after the majority of nodes have approved it. S7: Each breeding unit node collaboratively trains the AI ​​multi-factor breeding evaluation model based on local data and obtains model parameter updates. Without uploading the original data, the parameter update results are shared and a global model update is formed. The global model update results are written into the consortium blockchain network and broadcast to all breeding unit nodes. S8: When any breeding unit node raises an objection to the decision result, an appeal transaction is submitted to trigger a review and an additional correction record is added.

[0006] As a preferred embodiment of the present invention, the sampling window is specifically a rolling time window or a generation interval period.

[0007] As a preferred technical solution of the present invention, the AI ​​multi-factor breeding evaluation system in S3 adopts a GA-ANN model that combines genetic algorithm feature selection with a deep learning model. The input of the GA-ANN model includes at least daily weight gain, feed conversion ratio, backfat thickness, litter size, survival rate, morbidity rate and environmental factors. The model output is the individual comprehensive breeding value. The model training uses the evaluation results of historical measurement data as a reference signal and uses mean squared error regression loss for parameter optimization.

[0008] As a preferred technical solution of the present invention, the consensus verification and validation in S6 adopts the PBFT process. Specifically, each breeding unit node automatically validates the transactions in the newly generated dynamic evaluation area. The automatic validation includes at least the validation of data digest and digital signature, the validation of timestamp and window identifier, and the validation of the consistency of the recalculated evaluation results under the same model version. When the deviation between the evaluation result recalculated by the breeding unit node and the recorded result in the dynamic evaluation area is less than the preset consistency threshold, the evaluation result is determined to be consistent. After automatic verification, the breeding unit nodes conduct a second vote to confirm acceptance. If the majority threshold is reached, the dynamic evaluation area will be written into the blockchain.

[0009] As a preferred technical solution of the present invention, in S7, each breeding unit node performs collaborative training on the AI ​​multi-factor breeding evaluation model based on local data, specifically using a federated learning mechanism to update parameters. The parameter updates are transmitted using a secure aggregation method, including at least one of homomorphic encryption or secure multi-party computation.

[0010] As a preferred technical solution of the present invention, the appeal transaction includes at least the identifier of the appealed record, the identifier of the appellant node, the appeal reason, the evidence summary and the appeal timestamp. When the appeal transaction is accepted by a majority of nodes within the preset review period, the review process is triggered to re-verify the disputed data and recalculate the comprehensive breeding value of the associated individuals. The correction method is to add a correction transaction in the new block and associate it with the original record.

[0011] Compared with existing technologies, this invention provides a multi-node blockchain collaborative breeding method for local pig breeds, which has the following beneficial effects: 1. This invention, through a permissioned consortium blockchain and consensus verification mechanism, records key data summaries, evaluation results, model versions, and breeding decisions on the blockchain according to window periods, ensuring clear sources, immutability, and auditability. This solves the problems of inconsistent data standards, difficulty in traceability, and difficulty in mutual trust in traditional multi-entity collaboration. In addition, the AI ​​multi-factor breeding evaluation system can simultaneously process multiple traits such as meat quality, growth, reproduction, health and stress resistance, as well as environmental management factors. It can characterize the nonlinear correlation between complex traits, improve the accuracy and stability of the evaluation of the comprehensive genetic value of local pig breeds, and reduce decision-making errors caused by single-point data or single algorithm bias.

[0012] 2. This invention retains the original breeding and measurement data locally, only writing the summary and tags into the chain, and improves the model through federated learning collaborative training. This avoids the privacy leakage and commercial sensitivity risks caused by directly aggregating original data across different fields. It achieves cross-node alignment and joint utilization of scattered data through scrolling windows and tagged archives, making up for the shortcomings of insufficient samples from individual breeding units. It also supports the target expression of local varietal advantageous traits such as meat flavor and stress resistance, promoting continuous improvement and application while preserving the characteristics. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0014] 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.

[0015] Please see Figure 1 A multi-node blockchain collaborative breeding method for local pig breeds includes the following steps: S1: Construct a consortium blockchain network among multiple breeding unit nodes, and construct a permissioned consortium blockchain network among multiple breeding unit nodes to record breeding data summaries, model versions and breeding decision results, and provide a trusted distributed ledger foundation for multi-node collaborative evaluation and decision-making; A permissioned consortium blockchain network is established with multiple breeding units as participating nodes. Node access and identity authentication are completed, blockchain nodes and smart contracts are deployed, and data structures, write permissions, model version management rules, and window triggering rules are defined to form a trusted distributed ledger environment that can be used for multi-node collaboration. The construction of the consortium blockchain network includes: the consortium administrator deploying member management and certificate issuing authorities to register the identities of each breeding unit node and issue digital certificates; deploying blockchain node software on each node server and establishing peer-to-peer communication connections; deploying block production / sorting services to collect transactions and generate blocks; defining data structures, window triggering rules, and write permissions through smart contracts; and preferably using Practical Byzantine Fault Tolerance (PBFT) consensus algorithm to reach consensus on the order and validity of new blocks. S2: Breeding Data Collection and Tagging: Each node collects multi-trait measurement data of local pig breeds and environmental factors, cleans and standardizes the raw data, and generates tags and data summaries (including individual ID, measurement time, measurement environment, trait category, etc.) for each data point, and writes the tags and data summaries into the consortium blockchain; S3: Deploy an AI multi-factor breeding evaluation system at each breeding unit node, and conduct a comprehensive evaluation of breeding pigs based on the local data of each breeding unit node, output the individual comprehensive breeding value, and input the individual comprehensive breeding value into the consortium blockchain network to form a traceable evaluation record; The AI ​​multi-factor breeding evaluation system employs a GA-ANN model that combines genetic algorithm feature selection with a deep learning model. The model input includes at least daily weight gain, feed conversion ratio, backfat thickness, litter size, survival rate, morbidity, and environmental factors. The model output is an individual comprehensive breeding value or comprehensive score. Model training uses historical measurement data or traditional genetic evaluation results as reference signals, employs regression loss methods such as mean squared error for parameter optimization, and normalizes or standardizes the model output to ensure cross-node comparability. The genetic algorithm uses validation set prediction performance as a fitness index to search for the optimal feature subset or model parameters. S4: Set up sampling windows in the AI ​​multi-factor breeding evaluation system, and trigger dynamic evaluation block generation and verification at the end of each sampling window; The rolling time window or generation interval period is set in the system parameter configuration module. The window is defined as a fixed time length or a fixed number of generations. The end of each window triggers the generation of a dynamic evaluation block. The block header must include at least the window identifier, start and end times, etc. The block body must include at least the breeding pig file tag and its data summary, the AI ​​multi-factor breeding evaluation model version number and parameter summary, model performance indicators, and breeding decision results. The original measurement data is stored in the local database of each node or in encrypted storage. S5: After each sampling window ends, the newly added data within the sampling window is packaged into a dynamic evaluation block and submitted to the consortium blockchain network; S6: Perform consensus verification and validation on the newly added dynamic evaluation block, and write the newly added dynamic evaluation block into the blockchain after the majority of nodes have approved it. Consensus verification adopts the PBFT process, including pre-preparation, preparation, and submission phases. Each node automatically verifies the transactions in the new block. The automatic verification includes at least the verification of data digests and digital signatures, the verification of timestamps and window identifiers, and the verification of the consistency of recalculated evaluation results under the same model version. When the deviation between the evaluation result recalculated by the node and the result recorded in the block is less than the preset consistency threshold, the evaluation result is determined to be consistent. Based on the automatic verification, each breeding unit node votes or signs on the chain to confirm whether to accept the new block / whether to adopt the breeding decision result of this window. After reaching the preset majority threshold (preferably no less than two-thirds of the nodes agree), the block is written to the ledger. S7: Each breeding unit node collaboratively trains the AI ​​multi-factor breeding evaluation model based on local data and obtains model parameter updates. Without uploading the original data, the parameter update results are shared and a global model update is formed. The global model update results are written into the consortium blockchain network and broadcast to all breeding unit nodes. Multi-node collaborative training employs a federated learning mechanism: each node performs several rounds of iterative training on local data to obtain model parameter updates or gradient updates, and submits the update results to the consortium blockchain after digitally signing them; after consensus is reached, the aggregation node uses a federated average weighted aggregation strategy to summarize the updates from each node, obtain the globally updated model parameters, and broadcast them to each node; parameter updates are transmitted in encrypted form using a secure aggregation method, which includes at least one of homomorphic encryption or secure multi-party computation to ensure that only the aggregation result can be recovered and the original data of a single node cannot be deduced. S8: When any breeding unit node raises an objection to the decision result, an appeal transaction is submitted to trigger a review and an additional correction record is added; An appeal transaction must include at least the identifier of the appealed record, the identifier of the appealing node, the grounds for the appeal, a summary of evidence, and an appeal timestamp. When an appeal transaction is accepted by a majority of nodes within the preset review period, a review process is triggered to re-verify the disputed data and recalculate the comprehensive breeding value of the associated individuals. The correction method is to add a correction transaction to the new block and associate it with the original record, without rolling back or deleting the existing block.

[0016] Example 1: Application process of artificial intelligence multi-factor evaluation model: In this embodiment, a local pig breed breeding alliance network is constructed, comprising 10 nodes including several core breeding farms, pig breeding cooperatives, and technology extension stations. These nodes are interconnected via a consortium blockchain, sharing breeding data and model parameters. First, experts determine the input-output structure of the multi-factor evaluation model based on the local pig breed breeding objectives. Inputs include: individual growth traits (such as 6-month-old body weight, daily weight gain, feed conversion ratio), meat quality traits (backfat thickness measured by live ultrasound), reproductive traits (number of piglets born to multiparous sows, piglet survival rate), health and stress resistance traits (morbidity rate, deworming interval, etc.), and pigsty environmental data (altitude, temperature, humidity, feeding methods), totaling more than twenty characteristic indicators. The output is a comprehensive breeding value score (a relative score with 100 as the baseline). Then, a deep learning model (a feedforward neural network model containing two hidden layers) is selected as the evaluation core. The consortium blockchain is responsible for task allocation and parameter aggregation for model training: after initializing the model parameters, the model is sent to each node. Each node uses historical data from its current operation (covering at least three years of production measurement records) to train the model locally for several rounds (e.g., 100 iterations per round), calculating the gradient to update the model parameters. After completing the predetermined number of rounds, the gradient information submitted by all nodes is collected, and a secure multi-party computation is used to weighted average the gradients from all nodes to obtain the new parameters of the global model. Then, the updated model parameters are passed to each node for the next round of training. This process is repeated approximately 20 rounds until the network loss function converges, resulting in the final comprehensive evaluation model. Throughout the entire training process, the original data from each node remains in the database; machine learning is achieved only through parameter exchange, ensuring data privacy and security.

[0017] After training, the system enters the model application phase. The latest block of the consortium blockchain stores the model version number and parameters for traceability. Each node begins inputting recent (e.g., within the last year as set in the rolling window) individual data of local pig breeds into the model to calculate the comprehensive breeding value. To ensure the reliability of the evaluation results, the comprehensive score for each pig is a predicted value provided by the model, while considering a certain confidence interval; for individuals with significant data gaps or low model prediction confidence, the system marks them to assist breeders in making comprehensive judgments. In this embodiment, the model evaluation results show that compared to the traditional method of sorting solely by daily weight gain, 20% of the individuals in the comprehensive breeding value ranking have changed—some breeding pigs that are not particularly outstanding based on growth rate have gained higher rankings due to meat quality and reproductive advantages, while some individuals with faster growth but poorer meat quality have actually seen their comprehensive scores decrease. This fully demonstrates the role of AI multi-factor evaluation in balanced selection. Finally, each node shares the comprehensive breeding values ​​and rankings of all candidate breeding pigs in this round through the consortium blockchain and locks the evaluation results data on the chain to prevent any unilateral tampering.

[0018] Example 2: Dynamic Block Generation and Rolling Update of Breeding Decisions: This embodiment illustrates the blockchain update process under the rolling window mechanism. In a local pig breed breeding alliance, the rolling window size is pre-set to a block update cycle of 6 months (equivalent to updating twice a year; this frequency can be adjusted according to breeding progress). Assuming that we are currently at the end of the Nth window cycle, each node has completed the calculation of the comprehensive breeding value of all breeding pigs under testing in the current period according to Embodiment 1, and a new batch of reserve breeding pigs is about to enter the selection stage.

[0019] Once the cycle's end signal is triggered, the consortium blockchain begins constructing new blocks. First, each node packages the new data collected during the cycle, including: registration information for newly born local pig breeds, new records of various performance measurements, and major events that occurred during the cycle (such as disease outbreaks and their handling). Next, the system automatically invokes an AI evaluation model to update the comprehensive breeding value assessment of all candidate breeding pigs (including those from the previous cycle and those newly reaching breeding age in this cycle), and adds corresponding tags to the assessment results, such as "Evaluation Time = 2025Q4" and "Evaluation Model Version = V3.2". Then, an evaluation report is compiled, which includes a list of candidate breeding pigs sorted from highest to lowest comprehensive breeding value and a summary of key performance indicators.

[0020] Based on the assessment report and the pre-set breeding objectives, the breeding decision plan for this cycle was generated accordingly: for example, the top 5% of boars and sows in the comprehensive score were selected to form candidates for the new generation core group, with the specific number calculated based on the actual group size (if there are currently 200 candidate boars in stock, then approximately 10 of the top 5% would be selected); at the same time, individuals with a medium ranking but outstanding performance in a certain single trait (such as boars in the top 1% of meat quality score) can be retained by breeding experts as a means of maintaining diversity, and these individuals will be specially marked in the plan. In addition, the decision plan provides recommended mating combinations (boar and sow pairings are based on comprehensive breeding value and the principle of avoiding inbreeding), as well as a culling list (older breeding pigs that rank low in several consecutive assessments will be culled). The entire draft plan was submitted to the consortium blockchain by the initiating node to enter the consensus process.

[0021] Subsequently, the consortium blockchain initiates the PBFT consensus algorithm, with each node verifying and voting on the proposed new data block (containing data records, evaluation reports, breeding decision plans, etc.). To improve efficiency, this embodiment employs a two-level consensus: first, a pre-defined set of ledger nodes (e.g., the three with the highest reputation out of 10 nodes) verify the proposal, and then broadcast it to all nodes for voting. Upon verification, if all nodes unanimously agree that the data records are consistent with their own copies, the AI ​​evaluation results are within acceptable limits, and the new plan conforms to conventional breeding principles, then consensus is reached. Once the agreement rate reaches a predetermined threshold (e.g., ≥2 / 3 of the nodes agree), the new block officially takes effect on the blockchain. The block records detailed important breeding data for this cycle, a comprehensive evaluation summary, and selection and culling lists, etc. Relevant tags (such as pig ID, data category, and time) make it very convenient to query the complete file of a specific pig.

[0022] Once a new block is confirmed, the system will automatically execute the breeding plan: each relevant node receives instructions and carries out actual operations according to the plan (such as allocating selected boars to core pens and arranging mating schedules). Simultaneously, data archived in the previous window period and no longer requiring frequent access (such as breeding records from older generations) is marked as historical archives in the ledger. Nodes can selectively offload this data from primary storage to cold storage to save on-chain space. This rolling update mechanism ensures that the latest breeding decisions are based on real-time data while avoiding infinitely expanding data storage.

[0023] In this embodiment, the design of rolling out a block every six months allows the system to adapt to the breeding rhythm of local pig breeds. On average, each sow produces two litters of piglets per year, and the growth information of each litter can be evaluated and selected successively in the subsequent two cycles. Through continuous rolling updates, a generation of breeding pigs can be completed in about two years, which is more continuous and efficient than the traditional method relying on annual measurements and centralized selection. When the improvement of the overall herd index is observed to slow down, the rolling window length can be appropriately shortened (e.g., changed to quarterly updates) to accelerate the improvement pace. Therefore, the combination of dynamic blocks and rolling windows transforms breeding into a continuous improvement process, accumulating small progress to ultimately achieve a significant improvement in the performance of local pig breeds.

[0024] Example 3: Multi-node consensus and dissent appeal mechanism: This embodiment describes a specific scenario of a consortium blockchain in the areas of consensus verification and dispute resolution for breeding data. The breeding consortium blockchain comprises multiple nodes from different regions, each acting as both a data provider and a supervisor, jointly ensuring the reliability of on-chain information. Assume that during the generation of the N+1th block, the following situation occurs: Node A uploads daily weight gain data for a boar from its farm, claiming that the boar's average daily weight gain during the 6-7 month age period reaches 1200 grams, far exceeding the normal growth level of local pig breeds. Once this data is on the blockchain, it will significantly increase the pig's overall breeding value. During the consensus review of the new block, another node B strongly questions the authenticity of this data and initiates a dispute appeal through a smart contract.

[0025] After receiving the appeal transaction from Node B, the consortium blockchain automatically initiated the objection handling process. First, the new block consensus process was paused, and the system notified all nodes to review the disputed data. Node B attached the weight gain records of other pigs in the same litter and the city's average growth curve for the same period to its appeal, pointing out that the pig's data significantly deviated from the norm. The nodes discussed this and compared the data using relevant on-chain data tags: including the pig's historical growth records at each growth stage, the growth performance of its littermates, and the average daily weight gain under similar feeding conditions in neighboring areas. The results showed that the pig's previous weight gain had been average, and its littermates showed no abnormalities; the reported value was indeed abnormally high. Furthermore, some nodes pointed out that there were no special fattening experiments or nutritional improvement projects in Node A's location during that period, making such a significant growth leap unlikely. On the other hand, Node A explained that the data originated from manual records and might contain typos or upload errors, and agreed to provide the original records for verification.

[0026] After discussion, most nodes tended to believe the data was unreliable or at least required further verification. According to the consortium blockchain arbitration rules, the system entered a voting phase: 9 out of 10 nodes voted in favor of the appeal (believing the data was inaccurate and should be removed), and 1 abstained. Since a majority of more than two-thirds was reached, the appeal was approved. Therefore, the consortium blockchain processed the disputed data according to its predetermined strategy: a void mark was added to the pig's "daily weight gain at 6-7 months of age" record in the new block, with the explanation "This data is invalid after consortium arbitration to avoid affecting the evaluation"; simultaneously, this data will not be included in the calculation of the comprehensive breeding value for this cycle. Node A, having made its first data misreport and actively cooperated, was required to rectify its internal processes but was not further penalized. The consortium blockchain fully recorded the process and result of this dispute resolution on the chain, forming a transparent reputation event log. The new block then continued the consensus process, successfully reaching an agreement after removing the abnormal data and confirming it on the blockchain.

[0027] This embodiment demonstrates that the multi-node consensus and objection appeal mechanism of the present invention effectively safeguards the credibility of data and decisions. On one hand, multi-node witnessing ensures that any attempt by a single node to upload distorted data cannot escape external supervision, institutionally preventing fraud and falsification of results, and encouraging participants to provide truthful and reliable information. On the other hand, even if unintentional errors or data anomalies occur due to special circumstances, the objection mechanism can correct them promptly, preventing erroneous information from negatively impacting breeding selection. For projects like local pig breeds that require the collective efforts of multiple parties for preservation and breeding, this mechanism establishes a foundation of trust and cooperation among all participants, effectively enhancing the credibility and success rate of breeding improvement work.

[0028] In summary, through the above embodiments, it is evident that the multi-node blockchain collaborative breeding method for local pig breeds provided by this invention can achieve secure data sharing and dynamic updates, combine artificial intelligence to achieve comprehensive evaluation of multiple indicators, and ensure decision-making fairness and error correction capabilities through multi-node consensus. It is particularly suitable for small-group collaborative breeding of local pig breeds. The specific scenarios and parameters described in each embodiment can be adjusted and changed according to actual needs without affecting the implementation of the core solution of this invention. With the help of this invention, the breeding efficiency and genetic progress of local pig breeds will be significantly improved, possessing great value for widespread application.

[0029] 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 multi-node blockchain collaborative breeding method for local pig breeds, characterized in that: Includes the following steps: S1: Construct a consortium blockchain network among multiple breeding unit nodes; S2: Collect multi-trait measurement data of local pig breeds and environmental factors corresponding to each breeding unit node, clean and standardize them, generate corresponding tags and data summaries, and write the tags and data summaries into the consortium blockchain network. S3: Deploy an AI multi-factor breeding evaluation system at each breeding unit node, and conduct a comprehensive evaluation of breeding pigs based on the local data of each breeding unit node, output the individual comprehensive breeding value, and input the individual comprehensive breeding value into the consortium blockchain network; S4: Set up sampling windows in the AI ​​multi-factor breeding evaluation system, and trigger dynamic evaluation block generation and verification at the end of each sampling window; S5: After each sampling window ends, the newly added data within the sampling window is packaged into a dynamic evaluation block and submitted to the consortium blockchain network; S6: Perform consensus verification and validation on the newly added dynamic evaluation block, and write the newly added dynamic evaluation block into the blockchain after the majority of nodes have approved it. S7: Each breeding unit node collaboratively trains the AI ​​multi-factor breeding evaluation model based on local data and obtains model parameter updates. Without uploading the original data, the parameter update results are shared and a global model update is formed. The global model update results are written into the consortium blockchain network and broadcast to all breeding unit nodes. S8: When any breeding unit node raises an objection to the decision result, an appeal transaction is submitted to trigger a review and an additional correction record is added.

2. The multi-node blockchain collaborative breeding method for local pig breeds according to claim 1, characterized in that: The sampling window is specifically a rolling time window or a generation interval period.

3. The multi-node blockchain collaborative breeding method for local pig breeds according to claim 1, characterized in that: The AI ​​multi-factor breeding evaluation system in S3 adopts a GA-ANN model that combines genetic algorithm feature selection with a deep learning model. The input of the GA-ANN model includes at least daily weight gain, feed conversion ratio, backfat thickness, litter size, survival rate, morbidity rate, and environmental factors. The model output is the individual comprehensive breeding value. The model training uses historical measurement data evaluation results as a reference signal and uses mean squared error regression loss for parameter optimization.

4. The multi-node blockchain collaborative breeding method for local pig breeds according to claim 1, characterized in that: The consensus verification and validation in S6 adopts the PBFT process. Specifically, each breeding unit node automatically validates the transactions in the newly generated dynamic evaluation area. The automatic validation includes at least the validation of data digest and digital signature, the validation of timestamp and window identifier, and the validation of the consistency of the recalculated evaluation results under the same model version. When the deviation between the evaluation result recalculated by the breeding unit node and the recorded result in the dynamic evaluation area is less than the preset consistency threshold, the evaluation result is determined to be consistent. After automatic verification, the breeding unit nodes conduct a second vote to confirm acceptance. If the majority threshold is reached, the dynamic evaluation area will be written into the blockchain.

5. The multi-node blockchain collaborative breeding method for local pig breeds according to claim 4, characterized in that: In S7, each breeding unit node collaboratively trains the AI ​​multi-factor breeding evaluation model based on local data, specifically using a federated learning mechanism to update parameters. The parameter updates are transmitted using a secure aggregation method, including at least one of homomorphic encryption or secure multi-party computation.

6. The multi-node blockchain collaborative breeding method for local pig breeds according to claim 5, characterized in that: The appeal transaction includes at least the identifier of the appealed record, the identifier of the appellant node, the grounds for the appeal, a summary of evidence, and an appeal timestamp. When the appeal transaction is accepted by a majority of nodes within the preset review period, the review process is triggered to re-verify the disputed data and recalculate the comprehensive breeding value of the associated individuals. The correction method is to add a correction transaction to the new block and associate it with the original record.

Images