Scientific and technological achievement full life cycle traceability tracking management system

Abstract

The invention discloses a scientific and technological achievement full-life-cycle traceability tracking management system, and relates to the technical field of scientific research management, and the system comprises a data fusion collection module which is used for collecting data of all stages of a scientific and technological achievement full-life cycle, and carrying out the format conversion and protocol adaptation of the data; the safe and credible storage module is used for storing the collected data by adopting a mixed framework combining distributed storage and blockchain storage; the intelligent analysis and decision module is used for analyzing the stored data based on an artificial intelligence model, mining data values and generating decision suggestions; and the dynamic interaction application module provides visual achievement traceability query, real-time state monitoring, intelligent early warning and decision support functions for different user roles. A semantic mapping rule is dynamically learned by using a reinforcement learning algorithm through the adaptive protocol translation unit, a heterogeneous protocol is automatically adapted, and seamless joint of multi-source data is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of scientific research management, and in particular to a traceability and tracking management system for scientific and technological achievements throughout their life cycle. Background Art

[0002] Scientific and technological achievement management encompasses the entire process of scientific and technological achievements, from R&D project establishment, experimental verification, and transformation to industrial application. It utilizes technical means to achieve systematic data recording, process monitoring, risk prevention and control, and value mining. With the rapid development of scientific and technological innovation, scientific and technological achievements have seen a surge in number, diverse types, and interdisciplinary integration. Traditional management methods face numerous challenges. A thorough search revealed that existing scientific and technological achievement management technologies often suffer from data fragmentation, insufficient storage security, and a lack of intelligent analysis and decision-making.

[0003] For example, at the data collection level, data generated by different R&D equipment and experimental platforms are difficult to effectively integrate due to different protocol standards; in the storage link, centralized storage is prone to data tampering risks, and simple blockchain storage cannot meet the efficient reading and writing needs of massive data; in the results transformation stage, due to the lack of intelligent analysis tools, it is difficult to quickly match market demand and technological advantages, resulting in a low results conversion rate. In addition, cross-institutional data sharing and collaborative management have long-term privacy leakage risks and lack of incentives, which seriously restrict the optimal allocation of scientific and technological resources. Summary of the Invention

[0004] (1) Technical problems solved

[0005] In response to the shortcomings of the existing technology, the present invention provides a traceability and tracking management system for the entire life cycle of scientific and technological achievements, which solves the problems of "data fragmentation, insufficient storage security, and lack of intelligent analysis and decision-making" in the above-mentioned background technology.

[0006] (2) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a scientific and technological achievement full life cycle traceability and tracking management system, including:

[0008] Data fusion and acquisition module, used to collect data from all stages of the entire life cycle of scientific and technological achievements, and perform format conversion and protocol adaptation on the data;

[0009] The secure and reliable storage module uses a hybrid architecture combining distributed storage and blockchain evidence storage to store collected data;

[0010] Intelligent analysis and decision-making module, which analyzes stored data based on artificial intelligence models, mines data value and generates decision-making recommendations;

[0011] Dynamic interactive application modules provide visual traceability query, real-time status monitoring, intelligent early warning and decision support functions for different user roles;

[0012] A cross-institutional collaboration module that uses federated learning technology to enable secure sharing and collaborative analysis of data between different institutions while protecting data privacy;

[0013] The data fusion acquisition module, the secure and reliable storage module, the intelligent analysis and decision-making module, the dynamic interactive application module and the cross-institutional collaboration module are connected through network communication.

[0014] Preferably, the data fusion acquisition module includes an edge computing and differential privacy fusion unit, an adaptive protocol translation unit, and a cross-modal data preprocessing unit. The edge computing and differential privacy fusion unit deploys a lightweight data processing module at the IoT acquisition edge node. When the network is abnormal, the lightweight data processing module uses a blockchain-based edge cache mechanism to write the data hash value into the local private chain in real time. After the network is restored, the data is lostless through hash comparison and erasure code technology. The adaptive protocol translation unit uses a reinforcement learning algorithm to dynamically learn the semantic mapping rules of different protocols to achieve automatic protocol adaptation. Specifically, the following steps are included:

[0015] Step 1: Initialize the Q table and build the state space and action space

[0016] Step 2: Receive the protocol state s to be adapted and select action a from the action space A according to the ε-greedy strategy;

[0017] Step 3: Execute the protocol conversion operation a, determine the protocol adaptation result, and grant positive rewards if successful , failure will give negative rewards ;

[0018] Step 4: According to the formula Update the Q table, where is the learning rate, is the discount factor, For the next state;

[0019] Step 5: Determine whether to terminate the training. If not, return to step 2. If it is terminated, output the optimal strategy.

[0020] In step three, the result after executing the protocol conversion operation a is judged by using the preset protocol adaptation success judgment rule.

[0021] The cross-modal data preprocessing unit constructs a cross-modal comparative learning model and combines it with the scientific research field knowledge graph to perform semantic enhancement processing on unstructured data including but not limited to images and logs.

[0022] Preferably, the secure and trusted storage module includes a dynamic data twin storage unit and a cross-chain data exchange unit. The dynamic data twin storage unit dynamically generates a lightweight data twin based on data activity. When the data needs to be stored, the twin hash value is written into the blockchain, and the consistency between the original data and the twin is verified through zero-knowledge proof technology. The cross-chain data exchange unit adopts an asynchronous consensus algorithm based on DAG, allowing blockchain nodes of different institutions to exchange data hashes asynchronously.

[0023] Preferably, the intelligent analysis and decision-making module includes a hierarchical knowledge reasoning network unit and an adaptive threshold optimization unit. In the hierarchical knowledge reasoning network unit, the bottom layer adopts the Transformer model to fuse multi-stage data, the middle layer builds a rule engine based on the domain knowledge graph, and the top layer generates decision recommendations through a causal reasoning algorithm. The adaptive threshold optimization unit adopts an adaptive Bayesian optimization framework to dynamically adjust the weights of risk assessment indicators according to historical warning accuracy and industry standards.

[0024] Preferably, the cross-institutional collaboration module includes a privacy-enhancing computing unit and a decentralized incentive unit. The privacy-enhancing computing unit adopts multi-party secure computing and homomorphic encryption nesting technology to encrypt the gradient data in the federated learning parameter update phase, and realizes secure aggregation in the encrypted state through a multi-party secure computing protocol. The decentralized incentive unit constructs a blockchain collaboration platform based on the token economy, and designs a dual-token mechanism to quantify data contribution and reward model training participants.

[0025] Preferably, when learning semantic mapping rules of different protocols, the adaptive protocol translation unit sets a reward function to give positive rewards to operations that successfully adapt to the protocol and negative rewards to operations that fail to adapt.

[0026] Preferably, the cross-modal data preprocessing unit adopts a domain adaptive adversarial network to dynamically adjust the unstructured data preprocessing strategy. The specific steps are as follows:

[0027] Step 1: Enter source domain data , using the generator G to convert it into target domain style data ;

[0028] Step 2: Source domain data and generate data Input the discriminator D, which determines the source of the data;

[0029] Step 3: Calculate the discriminator loss , and update the discriminator D parameters;

[0030] Step 4: Calculate the generator loss ,in is the weight coefficient, To achieve semantic consistency loss, update the parameters of the generator G;

[0031] Step 5: Repeat the training until the discriminator D cannot distinguish whether the data is real data or generated data, and outputs the preprocessed cross-modal data ;

[0032] The privacy-enhancing computing unit also includes a model poisoning attack detection module, which uses the isolation forest algorithm to analyze the gradient distribution. When an anomaly is found, the dynamic weight decay mechanism is triggered, forcing the update of the model weights of malicious participants and recording the violations in the blockchain for evidence.

[0033] Preferably, the decentralized incentive unit automatically executes task allocation and revenue settlement through smart contracts, and all operation records are stored on the chain.

[0034] Preferably, the hierarchical knowledge reasoning network unit generates decision suggestions by presenting them in a visual causal relationship diagram.

[0035] (3) Beneficial effects

[0036] The present invention provides a traceability and tracking management system for the entire life cycle of scientific and technological achievements. It has the following beneficial effects:

[0037] (1) When the scientific and technological achievements full life cycle traceability and tracking management system is in use, the adaptive protocol translation unit uses the reinforcement learning algorithm to dynamically learn the semantic mapping rules, automatically adapt to heterogeneous protocols, and realize seamless connection of multi-source data; at the same time, the cross-modal data preprocessing unit combines the scientific research field knowledge graph with the domain adaptive adversarial network to perform semantic enhancement on unstructured data including but not limited to images and logs, effectively solving the problem of data fragmentation and providing a complete and standardized data foundation for full process management.

[0038] (2) When the scientific and technological achievements full life cycle traceability and tracking management system is in use, a dynamic data twin storage unit is used to generate a lightweight twin based on data activity, thereby reducing storage costs, and verifying data consistency through zero-knowledge proof technology to ensure the credibility of the evidence. The cross-chain data exchange unit uses a DAG-based asynchronous consensus algorithm to realize the asynchronous exchange of data hashes between blockchain nodes of different institutions, which not only ensures data security but also improves the efficiency of reading and writing massive data and cross-chain interaction.

[0039] (2) When in use, the traceability and tracking management system for the entire life cycle of scientific and technological achievements integrates the Transformer model and causal reasoning algorithm through hierarchical knowledge reasoning network units, deeply mines the value of multi-stage data, generates visual decision-making suggestions, and assists in accurate decision-making. It also quantifies data contributions and automatically settles profits through the dual-token mechanism and smart contracts of decentralized incentive units, drives efficient collaboration between industry, academia, and research, and significantly improves the conversion rate of scientific and technological achievements. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the complete framework of the system in the present invention;

[0041] Figure 2 Schematic diagram of the detailed framework structure of the data fusion acquisition module of the present invention;

[0042] Figure 3 This is a schematic diagram of the detailed framework structure of the secure and reliable storage module of the present invention;

[0043] Figure 4 This is a detailed schematic diagram of the framework structure of the intelligent analysis and decision-making module of the present invention;

[0044] Figure 5 Schematic diagram of the detailed framework structure of the cross-institutional collaboration module of the present invention.

[0045] In the figure: 1. Data fusion and acquisition module; 2. Secure and trusted storage module; 3. Intelligent analysis and decision-making module; 4. Dynamic interactive application module; 5. Cross-institutional collaboration module; 101. Edge computing and differential privacy fusion unit; 102. Adaptive protocol translation unit; 103. Cross-modal data preprocessing unit; 201. Dynamic data twin storage unit; 202. Cross-chain data exchange unit; 301. Hierarchical knowledge reasoning network unit; 302. Adaptive threshold optimization unit; 501. Privacy-enhancing computing unit; 502. Decentralized incentive unit. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] See also Figure 1-Figure 5The present invention provides a traceability and tracking management system for the entire life cycle of scientific and technological achievements, including: a data fusion and acquisition module 1, a secure and trusted storage module 2, an intelligent analysis and decision-making module 3, a dynamic interactive application module 4, and a cross-institutional collaboration module 5. The data fusion and acquisition module 1 is used to collect data from various stages of the entire life cycle of scientific and technological achievements, and perform format conversion and protocol adaptation on the data. Specifically, the data fusion and acquisition module 1 includes an edge computing and differential privacy fusion unit 101, an adaptive protocol translation unit 102, and a cross-modal data preprocessing unit 103. The edge computing and differential privacy fusion unit 101 deploys a lightweight data processing module at the edge node of the Internet of Things collection. When the network is abnormal, the lightweight data processing module adopts a blockchain-based edge cache mechanism to write the data hash value into the local private chain in real time. After the network is restored, the data is lostless through hash comparison and erasure code technology. The adaptive protocol translation unit 102 uses a reinforcement learning algorithm to dynamically learn the semantic mapping rules of different protocols to achieve automatic protocol adaptation. Specifically, the following steps are included:

[0048] Step 1: Initialize the Q table and build the state space and action space

[0049] Step 2: Receive the protocol state s to be adapted and select action a from the action space A according to the ε-greedy strategy;

[0050] Step 3: Execute the protocol conversion operation a, determine the protocol adaptation result, and grant positive rewards if successful , failure will give negative rewards ;

[0051] Step 4: According to the formula Update the Q table, where is the learning rate, is the discount factor, For the next state;

[0052] Step 5: Determine whether to terminate the training. If not, return to step 2. If it is terminated, output the optimal strategy.

[0053] In step three, the result after executing the protocol conversion operation a is judged by the preset protocol adaptation success judgment rule. In addition, when the adaptive protocol translation unit 102 learns the semantic mapping rules of different protocols, it sets a reward function to give positive rewards to operations that successfully adapt to the protocol and negative rewards to operations that fail to adapt.

[0054] The cross-modal data preprocessing unit 103 constructs a cross-modal comparative learning model and combines it with the scientific research field knowledge graph to perform semantic enhancement processing on unstructured data including but not limited to images and logs. Specifically, the cross-modal data preprocessing unit 103 adopts a domain adaptive adversarial network to dynamically adjust the unstructured data preprocessing strategy. The specific steps are as follows:

[0055] Step 1: Enter source domain data , using the generator G to convert it into target domain style data ;

[0056] Step 2: Source domain data and generate data Input the discriminator D, which determines the source of the data;

[0057] Step 3: Calculate the discriminator loss , and update the discriminator D parameters;

[0058] Step 4: Calculate the generator loss ,in is the weight coefficient, is the semantic consistency loss, updates the generator G parameters, and The calculation is constrained by combining the knowledge graph in the scientific research field;

[0059] Step 5: Repeat the training until the discriminator D cannot distinguish whether the data is real data or generated data, and outputs the preprocessed cross-modal data ;

[0060] The privacy-enhancing computing unit 501 also includes a model poisoning attack detection module, which uses the isolation forest algorithm to analyze the gradient distribution. When an anomaly is found, the dynamic weight decay mechanism is triggered, forcing the update of the model weights of the malicious participants and recording the violations in the blockchain for evidence.

[0061] The secure and trusted storage module 2 adopts a hybrid architecture that combines distributed storage and blockchain evidence to store collected data. Specifically, the secure and trusted storage module 2 includes a dynamic data twin storage unit 201 and a cross-chain data exchange unit 202. The dynamic data twin storage unit 201 dynamically generates lightweight data twins based on data activity, reducing storage costs while ensuring data evidence requirements; the consistency between the original data and the twin is verified through zero-knowledge proof technology to ensure the credibility of data evidence. The cross-chain data exchange unit 202 adopts an asynchronous consensus algorithm based on DAG to realize the asynchronous exchange of data hashes between blockchain nodes of different institutions, thereby improving the efficiency and security of cross-institutional data exchange, and forming a unique innovation in data storage and exchange technology.

[0062] The intelligent analysis and decision-making module 3 analyzes the stored data based on the artificial intelligence model, mines the data value and generates decision suggestions. Specifically, the intelligent analysis and decision-making module 3 includes a hierarchical knowledge reasoning network unit 301 and an adaptive threshold optimization unit 302. In the hierarchical knowledge reasoning network unit 301, the Transformer model is used to fuse multi-stage data. The middle layer builds a rule engine based on the domain knowledge graph. The top layer generates decision suggestions through a causal reasoning algorithm and presents them as a visual causal relationship diagram, forming a complete analysis chain from data feature extraction, logical reasoning to decision suggestion output. In addition, when generating decision suggestions, the hierarchical knowledge reasoning network unit 301 presents them as a visual causal relationship diagram. The adaptive threshold optimization unit 302 adopts an adaptive Bayesian optimization framework to dynamically adjust the risk assessment indicator weights according to historical warning accuracy and industry standards, thereby improving the accuracy of analysis and decision-making and providing scientific decision support means for scientific and technological achievement management.

[0063] The dynamic interactive application module 4 provides visual traceability query, real-time status monitoring, intelligent early warning and decision support functions for different user roles. The cross-institutional collaboration module 5 uses federated learning technology to achieve secure sharing and collaborative analysis of data between different institutions, while protecting data privacy. Specifically, the cross-institutional collaboration module 5 includes a privacy-enhancing computing unit 501 and a decentralized incentive unit 502. The privacy-enhancing computing unit 501 uses multi-party secure computing and homomorphic encryption nesting technology to encrypt the gradient data in the federated learning parameter update phase, and realizes secure aggregation in the encrypted state through a multi-party secure computing protocol. The decentralized incentive unit 502 builds a blockchain collaboration platform based on the token economy, designs a dual-token mechanism to quantify data contribution and reward model training participants, among which the decentralized incentive unit 502 automatically executes task allocation and income settlement through smart contracts, and all operation records are stored on the chain;

[0064] The data fusion acquisition module 1, the secure and reliable storage module 2, the intelligent analysis and decision-making module 3, the dynamic interactive application module 4 and the cross-institutional collaboration module 5 are connected through network communication.

[0065] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. The traceability and tracking management system for the entire life cycle of scientific and technological achievements is characterized by: include: Data fusion acquisition module (1), used to collect data from all stages of the entire life cycle of scientific and technological achievements, and to convert the data into different formats and adapt the data to different protocols; The secure and trusted storage module (2) uses a hybrid architecture combining distributed storage and blockchain evidence storage to store collected data; Intelligent analysis and decision-making module (3), which analyzes the stored data based on artificial intelligence models, mines the data value and generates decision-making suggestions; Dynamic interactive application module (4) provides visual traceability query, real-time status monitoring, intelligent early warning and decision support functions for different user roles; Cross-institutional collaboration module (5), which uses federated learning technology to achieve secure sharing and collaborative analysis of data between different institutions while protecting data privacy; The data fusion acquisition module (1), the secure and reliable storage module (2), the intelligent analysis and decision-making module (3), the dynamic interactive application module (4), and the cross-institutional collaboration module (5) are connected through network communication.

2. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 1 is characterized by: The data fusion acquisition module (1) includes an edge computing and differential privacy fusion unit (101), an adaptive protocol translation unit (102) and a cross-modal data preprocessing unit (103). The edge computing and differential privacy fusion unit (101) deploys a lightweight data processing module at the edge node of the Internet of Things acquisition. When the network is abnormal, the lightweight data processing module uses an edge cache mechanism based on blockchain to write the data hash value into the local private chain in real time. After the network is restored, the data is lostless retransmission is achieved through hash comparison and erasure code technology. The adaptive protocol translation unit (102) uses a reinforcement learning algorithm to dynamically learn the semantic mapping rules of different protocols to achieve automatic protocol adaptation. Specifically, the following steps are included: Step 1: Initialize the Q table and build the state space and action space 3. Step 2: Receive the protocol state s to be adapted and select action a from the action space A according to the ε-greedy strategy; Step 3: Execute the protocol conversion operation a, determine the protocol adaptation result, and grant positive rewards if successful , failure will give negative rewards ; Step 4: According to the formula Update the Q table, where is the learning rate, is the discount factor, For the next state; Step 5: Determine whether to terminate the training. If not, return to step 2. If it is terminated, output the optimal strategy. In step three, the result after executing the protocol conversion operation a is judged by using the preset protocol adaptation success judgment rule.

4. The cross-modal data preprocessing unit (103) constructs a cross-modal contrastive learning model and combines it with the scientific research field knowledge graph to perform semantic enhancement processing on structured data including but not limited to images and logs.

5. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 1 is characterized by: The secure and trusted storage module (2) includes a dynamic data twin storage unit (201) and a cross-chain data exchange unit (202). The dynamic data twin storage unit (201) dynamically generates a lightweight data twin based on data activity. When data needs to be stored, the twin hash value is written into the blockchain, and the consistency between the original data and the twin is verified by zero-knowledge proof technology. The cross-chain data exchange unit (202) adopts an asynchronous consensus algorithm based on DAG, allowing blockchain nodes of different institutions to exchange data hashes asynchronously.

6. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 1 is characterized by: The intelligent analysis and decision module (3) includes a hierarchical knowledge reasoning network unit (301) and an adaptive threshold optimization unit (302). In the hierarchical knowledge reasoning network unit (301), the bottom layer adopts a Transformer model to fuse multi-stage data, the middle layer builds a rule engine based on the domain knowledge graph, and the top layer generates decision suggestions through a causal reasoning algorithm. The adaptive threshold optimization unit (302) adopts an adaptive Bayesian optimization framework to dynamically adjust the weight of risk assessment indicators according to historical warning accuracy and industry standards.

7. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 1 is characterized by: The cross-institutional collaboration module (5) includes a privacy-enhancing computing unit (501) and a decentralized incentive unit (502). The privacy-enhancing computing unit (501) uses multi-party secure computing and homomorphic encryption nesting technology to encrypt the gradient data in the federated learning parameter update phase and realizes secure aggregation in the encrypted state through a multi-party secure computing protocol. The decentralized incentive unit (502) constructs a blockchain collaboration platform based on the token economy and designs a dual-token mechanism to quantify data contribution and reward model training participants.

8. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 2 is characterized by: When learning semantic mapping rules of different protocols, the adaptive protocol translation unit (102) sets a reward function to give positive rewards to operations that successfully adapt to the protocol and negative rewards to operations that fail to adapt.

9. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 5 is characterized by: The cross-modal data preprocessing unit (103) adopts a domain adaptive adversarial network to dynamically adjust the unstructured data preprocessing strategy. The specific steps are as follows: Step 1: Enter source domain data , using the generator G to convert it into target domain style data ; Step 2: Source domain data and generate data Input the discriminator D, which determines the source of the data; Step 3: Calculate the discriminator loss , and update the discriminator D parameters; Step 4: Calculate the generator loss ,in is the weight coefficient, To achieve semantic consistency loss, update the parameters of the generator G; Step 5: Repeat the training until the discriminator D cannot distinguish whether the data is real data or generated data, and outputs the preprocessed cross-modal data ; The privacy-enhancing computing unit (501) further includes a model poisoning attack detection module, which uses an isolation forest algorithm to analyze gradient distribution. When an anomaly is found, a dynamic weight decay mechanism is triggered to force the update of the model weight of the malicious participant, and the violation is recorded in the blockchain for evidence storage.

10. The scientific and technological achievements full life cycle traceability and tracking management system according to claim 5, characterized in that: The decentralized incentive unit (502) automatically executes task allocation and revenue settlement through smart contracts, and all operation records are stored on the chain.

11. The scientific and technological achievements full life cycle traceability management system according to claim 4, characterized in that: When generating decision suggestions, the hierarchical knowledge reasoning network unit (301) presents them in a visual causal relationship diagram.

Images