Knowledge graph-based algorithm model operation driving engine construction method

Abstract

The invention discloses a knowledge graph-based algorithm model operation driving engine construction method, which comprises the following steps of: respectively configuring a data source, an algorithm and a model to form a data source library, an algorithm library and a model library, selecting the required data source, algorithm and model from the libraries, and arranging according to a model operation rule to obtain an algorithm model operation rule graph. And analyzing the algorithm model operation rule map into triple text data, converting the triple text data into formalized language text, verifying by using multi-model interaction behaviors, and returning error prompt information through an interaction interface if the rule is unqualified. And if the rule text is qualified, inputting the rule text into a task engine generator, and generating an algorithm model driving engine by the generator according to the rule text. According to the method, the problems of difficulty in arrangement of algorithm and model operation modes, tedious management of operation rules, low execution efficiency and the like in a manufacturing environment are solved, so that the production efficiency and the product quality are improved.

Description

Technical Field

[0001] The present invention relates to intelligent algorithms and models, and specifically to a method for building an algorithm model operation driving engine based on a knowledge graph, and belongs to the field of artificial intelligence application technology. Background Art

[0002] With the popularization and in-depth development of artificial intelligence technology, the intelligent applications formed by artificial intelligence algorithms and models have become a key force in promoting the transformation and upgrading of various industries. In this context, various manufacturing companies are increasingly paying attention to and widely applying intelligent algorithms and models to enable their production and management. Algorithms and models have gradually become an important part of the core competitiveness of enterprises as a kind of enterprise soft assets. At the same time, the application scenarios of algorithms and models have become increasingly complex. In the digital environment of enterprises, a large number of algorithms and models are scattered in various scenarios. These application scenarios not only cover a variety of operating processes, but also involve strict data processing standards and complex model operation rules. This makes it difficult to manage and maintain algorithms, models and application scenarios efficiently, and their management and maintenance require excessive technical and management costs. Therefore, how to scientifically manage and effectively organize algorithms, models and their scenario applications has become an important challenge facing enterprises. However, most companies currently have the following limitations in managing and organizing these complex algorithms, models and their application rule systems.

[0003] Complex algorithm and model management: As intelligent application scenarios become more complex, enterprises use a variety of algorithms and models in different application scenarios. These algorithms and models are not only diverse, but also differ in implementation mechanisms, input and output requirements, parameter settings, etc., leading to management complexity.

[0004] Difficulty in arranging the operation mode of algorithms and models: Existing technologies mostly use text or tables to display the operation mode of algorithms and models, lacking an intuitive visual interface. This makes it difficult for technicians to quickly arrange the operation rule system of algorithms and models, and since there is no corresponding operation mode verification mechanism, there is a risk of errors in the application.

[0005] The implementation cycle of scenario applications is long and the technical cost is high: Traditional scenario application implementation methods require high technical barriers and cannot quickly adapt to changes in the production environment. When it is necessary to modify or add new algorithms or model operation methods, it is usually necessary to rewrite the code or perform complex system configuration, resulting in slow response speed and affecting production efficiency.

[0006] In summary, various functional theme applications in intelligent environments, such as fault diagnosis and prediction of various equipment, process parameter optimization, quality prediction, dynamic production scheduling, machine vision, etc., all require effective integration of data, algorithms and models to meet the application needs of actual algorithms and models. However, different application themes have different integration methods, which brings complexity to integration. At the same time, running multiple algorithms and models separately will consume a lot of computing resources and may cause program deadlock. In addition, frequent manual intervention in traditional methods also increases the difficulty of application implementation. Summary of the invention

[0007] In view of the fact that a large number of algorithms and models in the intelligent manufacturing environment are cumbersome to execute and difficult to flexibly cope with the resource-limited and ever-changing production environment, the present invention proposes a method for constructing an algorithm model operation rule-driven engine based on a knowledge graph. The present invention solves the problems of difficulty in arranging the operation mode of algorithms and models in the manufacturing environment, cumbersome management of operation rules, and low execution efficiency, thereby improving production efficiency and product quality.

[0008] The technical solution of the present invention is achieved in this way: The algorithm model operation driving engine construction method based on the knowledge graph includes the following steps: 1) Configure data, algorithms and models; manage and configure data sources, algorithms and models respectively through the configuration management module to form a data source library, algorithm library and model library; 2) Constructing an algorithm model operation rule map; selecting the required data source, algorithm and model from the data source library, algorithm library and model library, and arranging them according to the proposed model operation rules to obtain an algorithm model operation rule map; the model operation rules are set through a graphical visual operation interface; 3) Analyze the operation rules of the algorithm and model; parse the algorithm model operation rule map constructed in step 2) into triple text data; 4) Check the consistency of rules; convert the triple text data into a formal language text of description logic, which constitutes the rule text; use the multi-model interactive behavior verification based on timed automata to check the operation logic of the algorithm model represented by the rule text. If there is a logical error or deadlock problem in the rule text, it is unqualified and an error prompt message is returned through the interactive interface; otherwise, the rule text is judged to be qualified and proceed to the next step; 5) Task engine generation: Input qualified rule text into the task engine generator, and the task engine generator generates an algorithm model driving engine based on the rule text.

[0009] Furthermore, in step 1), when configuring the data source, the data interface is first configured, and the configuration content includes the use of the protocol and protocol-related parameters; after completing the data interface configuration, the data source of the data interface is configured, including the variable name, data type and physical quantity formula; after completing the configuration, the corresponding data source data is connected, and the collected data values ​​will be converted by the configured physical quantity formula to form a data source library.

[0010] Furthermore, in step 1), when configuring the algorithm, the algorithm in the algorithm library is added by uploading the algorithm file, and the algorithm name, input parameters, output parameters and hyperparameter information need to be configured during uploading; When configuring a model, you can add a new model to the model library by uploading the model file. When uploading, you need to configure the model name, input parameters, output parameters, and hyperparameter information.

[0011] Furthermore, in step 2), the algorithm model operation rule map is constructed as follows: 2.1) First, set the trigger conditions; 2.2) Next, configure the model to be run after triggering and the data source of the model input; the model input dimension is determined by the input parameter attributes of the model. The configured data source and model are displayed as nodes in the algorithm model operation rule map and connected with the "input" edge relationship; 2.3) According to the output parameter attributes of the model configuration, the output source node is automatically generated and connected to the model with an "output" edge relationship; 2.4) After the configuration is completed, set the end node. When the running task is executed to the end node, the current driving task is ended and waits for the driving task to be triggered again.

[0012] Furthermore, the trigger condition is set as a time trigger or an event trigger; the time trigger condition is set as an interval setting time trigger, and the event trigger condition is set as a certain variable exceeding a set threshold to trigger.

[0013] Furthermore, in step 5), the task engine generator generates an algorithm model driving engine as follows: 5.1) Storing the qualified rule text in a relational database as a record of the set algorithm and model operation rules; 5.2) Store the configured algorithm and model operation rule graph in the graph database for subsequent reference and modification; 5.3) The task engine generator will construct corresponding tables in the relational database based on the input and output of the algorithm and model in the rule text to form a task space for storing the data generated after the algorithm and model are run; 5.4) The task engine generator generates an algorithm model driving engine based on the rule text, monitors the trigger conditions in real time, and drives the algorithm and model with a sequential structure to execute the corresponding tasks when the trigger response is triggered.

[0014] The present invention is based on an algorithm library and a model library combined with a knowledge graph. By parsing graph relationship data, the operating modes of algorithms and models in the system are automatically arranged and executed, thereby realizing the effective organization and optimized operation of a large number of algorithms and models in an intelligent manufacturing environment based on subject rules. It solves the problems of difficult arrangement of the operating modes of algorithms and models in a manufacturing environment, cumbersome management of operating rules, and low execution efficiency, thereby improving production efficiency and product quality.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1) Configurable management of algorithms and models: Provide algorithm management and model management modules to achieve configurable management of algorithms and models. Through configuration information, clarify the input parameters, output parameters, and setting parameters of the algorithm and model. This modularizes the management process of algorithms and models, so that each algorithm and model has clear and unified management specifications, thereby enhancing the integration and collaboration between different algorithms and models, improving overall management efficiency, and reducing management complexity. At the same time, through clear configuration information, users can quickly trace back to specific algorithms or models, reducing errors and confusion caused by management complexity.

[0016] 2) Visual arrangement of algorithm and model operation mode: Provides a visual interface for algorithm and model operation mode, allowing users to add, delete or modify rules through simple operations such as drag and drop and click, which improves the convenience of algorithm and model operation mode arrangement. At the same time, the visualization function of the knowledge graph can provide a global perspective, helping technical personnel to quickly understand and process the operation logic of intelligent models and algorithms, and improve the scientificity and accuracy of rule making.

[0017] 3) Automated algorithm and model operation mode verification: When the algorithm and model operation mode of the application scenario are choreographed, the system will automatically call the built-in rule checker to verify whether there are logical errors or deadlocks in its operation mode. If problems are found, the corresponding error information will be returned to prompt the user to make modifications. Only when the choreographed application scenario passes the verification can the next step of application generation be entered, thereby reducing the risk of application errors.

[0018] 4) Scenario application function generation: When building a new scenario application or modifying an existing scenario application, you only need to configure it through the front-end interactive page, without writing back-end code. When the scenario application passes the inspection, the system will call the task engine generator to build the application's task space and generate the corresponding task engine, automatically drive the algorithms and models in the application, and run them according to the arrangement. This ensures that users can quickly configure the operation mode of algorithms and models, so that functional applications can adapt to changes in the production environment.

[0019] The present invention introduces knowledge graph technology, based on the algorithm library and model library combined with knowledge graph, to achieve automatic management and operation of algorithm models in a digital environment. Knowledge graph is a structured form of knowledge representation that can effectively organize and manage complex data and information, and support efficient query and reasoning. By constructing a knowledge graph for a specific topic, it is possible to achieve automatic management and execution of various algorithms and model operation modes in the functional application process, thereby improving overall production efficiency and product quality. The present invention can provide a more intelligent, efficient and flexible algorithm and model operation driving method for intelligent applications in various fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 - Schematic diagram of the construction process of the driving engine of the present invention.

[0021] Figure 2 - Flowchart of the scene setting of the example of the present invention.

[0022] Figure 3 -Flowchart of data source configuration of the present invention.

[0023] Figure 4 -Flowchart of algorithm input of the present invention.

[0024] Figure 5 -Model input flow chart of the present invention.

[0025] Figure 6 -Configuration flow chart of the algorithm and model operation mode of the present invention.

[0026] Figure 7 -The algorithm and model of the present invention run the driving engine to generate a flow chart.

[0027] Figure 8 - Flowchart of application scenarios of the present invention. DETAILED DESCRIPTION

[0028] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0029] The algorithm model of the present invention is based on the knowledge graph to run the driving engine construction method, and see Figure 1The schematic diagram of the drive engine construction process shown in the figure is as follows: 1) Configurable management of data, algorithms, and models The data source, algorithm, model and other elements are managed and configured respectively by the configuration management module, thereby forming the corresponding data source library, algorithm library and model library, which constitute the resource library required for the generation of the driving engine of the present invention. When configuring the data source, the data interface needs to be configured first, and the configuration content includes the use of the protocol and protocol-related parameters. After completing the data interface configuration, the data source of the data interface can be configured, including information such as variable name, data type and physical quantity formula. After completing the configuration, the corresponding data source data is accessed, and the collected data value will be converted by the configured physical quantity formula to form a data source library. When configuring the algorithm, the algorithm of the algorithm library can be added by uploading the algorithm file, and the algorithm name, input parameters, output parameters, hyperparameters and other information need to be configured when uploading. The configuration of the model also adopts a similar method to the algorithm configuration, and the model library is added by uploading the model file, and the model name, input parameters, output parameters, hyperparameters and other information need to be configured when uploading.

[0030] 2) Constructing algorithm model operation rule map Select the required data source, algorithm and model from the data source library, algorithm library and model library, and arrange them according to the proposed model operation rules to obtain the algorithm model operation rule map; the model operation rules are set through a graphical operation interface. The specific method is: 2.1) First, set the trigger conditions. The trigger conditions can be set as time trigger or event trigger. For example, you can set the time trigger condition to trigger at 5-minute intervals, or set the event trigger condition to trigger when a variable exceeds the set threshold.

[0031] 2.2) Next, configure the model to be run after triggering and the data source of the model input. The model input dimension is determined by the input parameter attributes of the model. The configured input source and model are displayed as nodes in the graph and connected with the "input" edge relationship.

[0032] 2.3) When the configuration is completed, the output source node is automatically generated according to the output parameter attributes of the model configuration and connected to the model with the edge relationship of "output". At this point, the configuration of the model is initially completed. Later, different configurations can be made based on the model output, such as configuring the output of the model as the input of other models, or parsing and judging the output source to perform different task operations.

[0033] 2.4) After the configuration is completed, the end node needs to be set. When the running task is executed to the end node, the current driving task is ended and waits for the driving task to be triggered again.

[0034] 3) Analytical algorithms and model operation rules The operation rule knowledge graph of the algorithm and model obtained through graphical configuration belongs to graph structure data. In order to perform subsequent algorithm and model execution arrangement, the graph structure data needs to be parsed into sequential structure text data. Therefore, in this method, the operation rule graph structure is parsed into triple text data for subsequent operations.

[0035] 4) Check the consistency of rules A rule checker is set in this method to verify the configured algorithm and model operation mode. Different model interaction mechanisms are the basis for realizing information sharing between models, and this interaction can be effectively described by logical behavior. Therefore, the present invention converts the triple text data into a formal language text describing the logic for model verification. The algorithm model operation logic is verified using multi-model interactive behavior verification based on timed automata. If the rules have logical errors or deadlocks, the checker will judge them as unqualified and return an error prompt message through the interactive interface; otherwise, the rules are judged to be qualified.

[0036] 5) Task engine generation Qualified rule texts will be input into the task engine generator, which will generate an algorithm model driving engine based on the rule text. Four main operations will be performed in the generator. First, the rule text will be stored in the relational database as a record of the set algorithm and model operation rules. Secondly, the algorithm and model operation graph configured this time will be stored in the graph database for subsequent reference and modification. In the third step, the generator will build corresponding tables in the relational database based on the input and output of the algorithm and model in the rule text to form a task space for storing the data generated after the algorithm and model are run. Finally, the generator generates an algorithm model driving engine based on the rule text, monitors the trigger conditions in real time, and drives the algorithm and model with a sequential structure when the trigger response is triggered to perform the corresponding tasks.

[0037] The following specific embodiments are given to further understand the implementation process and logic of the present invention.

[0038] This embodiment adopts the B / S (Browser / Server) architecture. The front-end uses the layui framework to implement the user interface, and the back-end uses the Spring Boot framework to process business logic. In terms of data storage, the relational database uses MySQL and the graph database uses Neo4j to support the management and execution of algorithm model operation rules. This embodiment takes the configuration of the fault diagnosis theme function as an example. The flow chart is as follows Figure 2 shown.

[0039] 1. Configure the data source In the data interface management function, users can add data interfaces to access MQTT data by entering information such as IP address, port, user name, password, and topic. After the configuration is completed, a data interface record will be formed. Then, users can use the data source management function to further configure the data source information of the interface. According to the data source name, data type, and physical quantity conversion formula, set up each data source information in detail. After the configuration is completed, each data source in the data interface will be connected to the system with data converted by the physical quantity conversion formula. The data source configuration process is as follows: Figure 3 shown.

[0040] 2. Algorithm input In the algorithm management module, users can add and configure data preprocessing algorithms and feature value extraction algorithms. Through the "Add" function in algorithm management, users can upload algorithm files and configure the algorithm name, input parameters, output parameters, and setting parameters. After the configuration is completed, the system will generate a new algorithm record in the algorithm library. Based on the configuration information, the input form, output form, and adjustable parameter items of the algorithm will be clearly determined. The algorithm entry process is as follows Figure 4 shown.

[0041] 3. Model input In the model management module, users can add and configure fault diagnosis models. Through the "Add" function in model management, users can upload model files and configure the model name, input parameters, output parameters, and model setting parameters. After the configuration is completed, the system will generate a new model record in the model library. Based on the configuration information, the model's input form, output form, and adjustable parameters will be clearly determined. The model entry process is as follows: Figure 5 shown.

[0042] 4. Operation mode configuration In the intelligent application scenario configuration function, the algorithm and model operation mode of the fault diagnosis application theme are configured.

[0043] First, set the trigger condition of the operation mode in the application scenario map, and set the condition to trigger once every 5 minutes. After adding successfully, a "trigger condition" node will be generated in the map, and its "condition attribute" will be set to "trigger every five minutes".

[0044] Next, add the data preprocessing algorithm to the scene. When the algorithm is added, a "Data Preprocessing Algorithm" node is generated in the graph, and the input node and "Output Source 1" node to be filled in are automatically generated based on the configuration information of the algorithm. The algorithm parameter setting is completed by filling in the attribute values ​​of the data preprocessing algorithm node. Configure the input source node and select the data source "Vibration Signal 1" in the data source library as the content of the node. After the configuration is completed, the input node becomes the "Vibration Signal 1" data source node.

[0045] After signal preprocessing, the feature extraction algorithm needs to be used to extract features from the data. The next step is to add the feature extraction algorithm in the algorithm library to the graph. After successful addition, a "Feature Extraction Algorithm" node will be generated in the graph, and the corresponding input node and "Output Source 2" node will be generated. In order to use the preprocessed data as a feature extraction object, the input source node generated by the feature extraction algorithm needs to be configured as the "Output Source 1" node.

[0046] Then, the features extracted by the feature extraction algorithm (i.e., "Output Source 2") are input into the fault diagnosis model. Select the fault diagnosis model and add it to the graph. The "Fault Diagnosis Model" node will be generated in the graph, and the corresponding input nodes and output nodes will be generated. Configure the input node as the "Output Source 2" node to input the feature extraction results into the fault diagnosis model. Change the name attribute of the output node to "Diagnosis Result".

[0047] Since different diagnosis results need to be processed in the fault diagnosis scenario, a judgment node named "Judgment Condition" is added to the graph, the "Diagnosis Result" node is configured as the input of the judgment condition, and the judgment expression is set as the node attribute. The corresponding branch edge relationship is generated according to the judgment expression content, including "faulty" and "no fault" branches.

[0048] Next, configure the corresponding task execution nodes. Under the "Faulty" branch, add "Task Execution Node A" and connect it to the "Judgment Condition" node through the "Faulty" edge relationship as the execution task after it is judged to be faulty. The execution content can select "Store Records" and "Push Alarms" from the execution library and configure them in the node as two attributes of the node. Under the "No Fault" branch, add "Task Execution Node B" and connect it to the "Judgment Condition" node through the "No Fault" edge relationship as the execution task after it is judged to be faulty. The execution content selects "Store Records" from the execution library and configures it in the node as a node attribute.

[0049] At this point, the logical settings for fault diagnosis are basically completed, and an end node needs to be added to the graph to end the operation of this application function. After adding the end node to the graph, connect "Task Execution Node A" and "Task Execution Node B" to it, and automatically form an "end" edge relationship. When the application completes the task content of "Task Execution Node A" or "Task Execution Node B", it is considered to have completed an application of the scenario function. The scenario application model and algorithm operation mode configuration process are as follows Figure 6 shown.

[0050] 5. Engine Generation When the operation mode configuration is completed, the configuration content will be passed to the rule checker for verification. If there is a logical contradiction or deadlock between the configured algorithm and the model operation mode, it will fail the test. The system will return the error reason on the front-end page, prompting the user that the theme function application setting failed and needs to be modified. If it passes the test, the task space is generated according to the content of the task execution node.

[0051] In this fault diagnosis scenario, the fault diagnosis results will be stored and an alarm will be triggered when a fault occurs. Therefore, the task generator will create two new tables in the MySQL database, one for storing fault diagnosis results and the other for storing alarm records. After generating the task space, the task generator will build an algorithm model to run the driving engine for the scenario function. The engine content generates the trigger condition monitoring method and the sequential structure task execution method based on the parsed formal language text. At this point, the scenario application configuration is complete. The algorithm and model running driving engine generation process is as follows: Figure 7 shown.

[0052] 6. Scenario Application When the scenario application is configured successfully, the algorithm model running driving engine of the scenario function will start running in the background. Before the functional task is triggered, the engine only executes the trigger condition monitoring method. The trigger condition configured in this fault diagnosis scenario is triggered once every 5 minutes, so the trigger condition monitoring method of the engine will implement the timer function. When it counts to 5 minutes, the timer will be cleared and enter the sequential structure task execution method.

[0053] In this method, the engine will collect data from the data source "Vibration Signal 1", and the collection length is consistent with the input parameter length of the "Data Preprocessing Algorithm". The processing results will be input into the "Feature Extraction Algorithm" for feature extraction. The extracted features are then input into the "Fault Diagnosis Model", and its output content will be conditionally judged: if there is a fault, the record of this diagnosis will be stored, and an alarm message will be pushed in the system platform, and the alarm information will be recorded; if there is no fault, only the record of this diagnosis will be stored.

[0054] After executing any branch, the engine sequence structure task execution method ends and re-enters the trigger condition monitoring method to wait for the next task to be triggered. Users can view the records generated by each scene task space in the scene application management. Figure 8 shown.

[0055] The present invention aims to flexibly manage, effectively organize and operate a large number of algorithms and models in digital environments such as smart manufacturing workshops and digital twin systems, thereby realizing configurable and automated execution of algorithms and models in the smart manufacturing process. By applying knowledge graph technology to complex digital environments, the present invention can effectively integrate various data sources, build subject-oriented algorithm and model operation rules, and automatically verify and execute their operation modes, thereby enabling artificial intelligence algorithms and models to empower various fields.

[0056] Finally, it should be noted that the above examples of the present invention are merely examples for illustrating the present invention, and are not intended to limit the embodiments of the present invention. Although the applicant has described the present invention in detail with reference to the preferred embodiments, for those of ordinary skill in the art, other different forms of changes and modifications can be made based on the above description. It is impossible to list all the embodiments here. Any obvious changes or modifications derived from the technical solution of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for constructing an algorithm model operation driving engine based on a knowledge graph, characterized in that: The following steps are involved: 1) Configure data, algorithms and models; manage and configure data sources, algorithms and models respectively through the configuration management module to form a data source library, algorithm library and model library; 2) Constructing an algorithm model operation rule map; selecting the required data source, algorithm and model from the data source library, algorithm library and model library, and arranging them according to the proposed model operation rules to obtain an algorithm model operation rule map; the model operation rules are set through a graphical visual operation interface; 3) Analyze the operation rules of the algorithm and model; parse the algorithm model operation rule map constructed in step 2) into triple text data; 4) Check the consistency of rules; convert the triple text data into a formal language text of description logic, which constitutes the rule text; use the multi-model interactive behavior verification based on timed automata to check the operation logic of the algorithm model represented by the rule text. If there is a logical error or deadlock problem in the rule text, it will fail and return an error prompt message through the interactive interface; Otherwise, the rule text is judged as qualified and proceed to the next step; 5) Task engine generation: Input qualified rule text into the task engine generator, and the task engine generator generates an algorithm model driving engine based on the rule text.

2. According to the method for constructing an algorithm model operation driving engine based on knowledge graph according to claim 1, it is characterized in that: In step 1), when configuring the data source, first configure the data interface, including the use of the protocol and protocol-related parameters; after completing the data interface configuration, configure the data source of the data interface, including the variable name, data type and physical quantity formula; after completing the configuration, the corresponding data source data is connected, and the collected data values ​​will be converted by the configured physical quantity formula to form a data source library.

3. The method for constructing an algorithm model operation driving engine based on a knowledge graph according to claim 1, characterized in that: In step 1), when configuring the algorithm, add the algorithm to the algorithm library by uploading the algorithm file. When uploading, you need to configure the algorithm name, input parameters, output parameters and hyperparameter information; When configuring a model, you can add a new model to the model library by uploading the model file. When uploading, you need to configure the model name, input parameters, output parameters, and hyperparameter information.

4. The method for constructing an algorithm model operation driving engine based on a knowledge graph according to claim 3 is characterized in that: In step 2), the algorithm model operation rule map is constructed as follows: 2.1) First, set the trigger conditions; 2.2) Next, configure the model to be run after triggering and the data source of the model input; the model input dimension is determined by the input parameter attributes of the model. The configured data source and model are displayed as nodes in the algorithm model operation rule map and connected with the "input" edge relationship; 2.3) According to the output parameter attributes of the model configuration, the output source node is automatically generated and connected to the model with an "output" edge relationship; 2.4) After the configuration is completed, set the end node. When the running task is executed to the end node, the current driving task is ended and waits for the driving task to be triggered again.

5. According to the method for constructing an algorithm model operation driving engine based on knowledge graph according to claim 4, it is characterized in that: The trigger condition is set as time trigger or event trigger; the time trigger condition is set as interval setting time trigger, and the event trigger condition is set as a certain variable exceeding a set threshold value to trigger.

6. The method for constructing an algorithm model operation driving engine based on a knowledge graph according to claim 1, characterized in that: In step 5), the task engine generator generates the algorithm model driving engine as follows: 5.1) Storing the qualified rule text in a relational database as a record of the set algorithm and model operation rules; 5.2) Store the configured algorithm and model operation rule graph in the graph database for subsequent reference and modification; 5.3) The task engine generator will construct corresponding tables in the relational database based on the input and output of the algorithm and model in the rule text to form a task space for storing the data generated after the algorithm and model are run; 5.4) The task engine generator generates an algorithm model driving engine based on the rule text, monitors the trigger conditions in real time, and drives the algorithm and model with a sequential structure to execute the corresponding tasks when the trigger response is triggered.

Images