Intelligent knowledge fusion platform for mechanical hydraulic product test of turboshaft turboprop engine

Abstract

The application discloses a kind of turbo-shaft turboprop engine mechanical hydraulic product test intelligent knowledge fusion platform.The platform includes distributed mass test data storage, data warehouse-text knowledge fusion and system plug-in function enhancement mechanism;Based on distributed mass data storage, flexible and scalable distributed storage scheme is provided for the mass data generated by test;On this basis, mass data is organized through data warehouse, based on data warehouse-text knowledge fusion, unstructured text data and structured mass data stored in data warehouse are linked through the form of URL, the fusion of unstructured and structured is realized;Through system plug-in function enhancement mechanism, the function of adding and updating decision algorithm at any time is realized.The application solves the technical problems of storage, management and query of product mass test data, provides data support for decision algorithm, realizes the extension and dynamic updating of system function.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computers, and particularly relates to an intelligent knowledge fusion platform for mechanical and hydraulic product testing of a turboshaft / turboprop engine. BACKGROUND

[0002] Currently, the following core problems exist in the testing of mechanical and hydraulic products of turboshaft / turboprop engines:

[0003] 1) The debugging, troubleshooting and long-term testing of products mainly rely on the experience of operators, and the data generated in these processes are not effectively utilized, resulting in long debugging cycles, inability to locate deep-seated and systematic faults of products, and inability to identify early warning signals;

[0004] 2) The testing data of mechanical and hydraulic products of turboshaft / turboprop engines are very large, for example, long-term testing data may include several hundred or even thousands of hours of information, and the data volume reaches GB or even TB levels, and the traditional database cannot meet the requirements in terms of capacity, scalability or query speed;

[0005] 3) As the data increases and the use of data deepens, new decision-making requirements will continuously arise, and new decision-making knowledge needs to be extracted, and the traditional platform system architecture is relatively fixed, and the function of adding and adjusting functions according to new requirements is weak. SUMMARY

[0006] The purpose of the present application is to provide an intelligent knowledge fusion platform for mechanical and hydraulic product testing of a turboshaft / turboprop engine. The present application solves the technical problems of storage, management and query of massive testing data of products, provides data support for decision-making algorithms, and realizes the expansion and dynamic updating of system functions.

[0007] The technical solution of the present application is an intelligent knowledge fusion platform for mechanical and hydraulic product testing of a turboshaft / turboprop engine, which comprises a distributed massive testing data storage, a data warehouse-text knowledge fusion and a system plug-in function enhancement mechanism. Based on the distributed massive data storage, a flexible and scalable distributed storage solution is provided for the massive data generated by testing. On this basis, the massive data is organized through a data warehouse, and the unstructured text data and the structured massive data stored in the data warehouse are linked in the form of URL based on the data warehouse-text knowledge fusion, realizing the fusion of unstructured and structured data and laying a foundation for subsequent knowledge extraction and data mining. Through the system plug-in function enhancement mechanism, the function of adding and updating decision-making algorithms at any time is realized.

[0008] In the foregoing intelligent knowledge fusion platform for mechanical and hydraulic product testing of a turboshaft / turboprop engine, the flexible and scalable distributed storage solution is to divide the massive data into multiple data blocks for storage.

[0009] The working process of the distributed mass data storage in the intelligent knowledge fusion platform for the turboshaft turboprop engine mechanical hydraulic product test comprises:

[0010] The NameNode module is used for managing metadata and is responsible for managing the file namespace of the entire distributed mass data storage.

[0011] The DataNode module is used as a working node of the distributed mass data storage and is responsible for storing the divided data blocks and providing read-write services of data for the client.

[0012] The writing process of the distributed mass data storage in the intelligent knowledge fusion platform for the turboshaft turboprop engine mechanical hydraulic product test is executed as follows:

[0013] First, the Client and the NameNode establish a connection, the NameNode checks metadata information and returns the information to the Client, then the Client receives the DataNode list returned by the NameNode and establishes a connection with the DataNode to prepare to receive data. The Client divides the file into blocks and sends the data blocks and the list information returned by the NameNode to the DataNode.

[0014] The reading process of the distributed mass data storage in the intelligent knowledge fusion platform for the turboshaft turboprop engine mechanical hydraulic product test is executed as follows:

[0015] The Client and the NameNode establish a connection, the NameNode obtains the data block mapping information of the required file and the DataNode list storing the data blocks by querying metadata information and sends them to the Client, and the Client obtains the information and finds the nearest available DataNode to retrieve the data block.

[0016] The working process of the distributed mass data storage in the intelligent knowledge fusion platform for the turboshaft turboprop engine mechanical hydraulic product test further comprises:

[0017] The Resource Manager module in the resource management and task scheduler accepts the distributed computing program submitted by the user and divides resources for the program; manages and monitors the resources on each Node Manager module to balance the load;

[0018] The Node Manager module in the resource management and task scheduler manages the computing resources of the machine, is responsible for accepting the tasks allocated by the Resource Manager module, and creates containers and recycles resources.

[0019] Under the support of NameNode module, DataNode module and resource management and task scheduler, the massive data is stored in the data warehouse.

[0020] In the aforementioned turbo-shaft turboprop engine mechanical hydraulic product test intelligent knowledge fusion platform, the working process of data warehouse-text knowledge fusion is as follows:

[0021] Based on the data warehouse-text knowledge fusion expression technology of the graph data model, the data warehouse and the existing text of the enterprise are defined as resources in the graph data model, and the data warehouse and the text are connected through a directed line segment; the text is assisted in knowledge extraction through search technology, and the data warehouse is assisted in knowledge extraction through data mining; at the same time, the graph traversal technology is used to realize the correlation knowledge extraction of the data warehouse and the text, and finally the deep mining and extraction of knowledge are realized.

[0022] In the aforementioned turbo-shaft turboprop engine mechanical hydraulic product test intelligent knowledge fusion platform, the architecture of the system plug-in function enhancement mechanism includes a user operation interface, a plug-in task pool and a plug-in; the user operation interface provides a plug-in management and plug-in arrangement function interface; the plug-in creates a corresponding plug-in task pool when used; the running mode of the system plug-in function enhancement mechanism is based on a workflow scheduling engine, and specifically as follows:

[0023] The user uploads the plug-in to the intelligent knowledge fusion platform, and the platform saves the plug-in to the local; then a process is created, the platform generates a zip package by arranging the plug-in and uploads it to the workflow scheduling engine, and the workflow scheduling engine returns a workflow Id; then when the user executes / schedules the plug-in, the platform submits a task to the resource management and task scheduler through the workflow engine, and finally operates the data warehouse to realize the function of the entire plug-in.

[0024] The advantages of the present application are: the present application proposes a turbo-shaft / turboprop engine mechanical hydraulic product test intelligent knowledge fusion platform based on big data, which uses big data technology to solve the problems of data storage, query and expansion; the stored product data is organized into different data warehouses according to the demand, providing system management and support for knowledge extraction; through the plug-in mechanism, the dynamic updating and expansion of system functions are realized.

[0025] Based on big data technology, the present application solves the storage capacity of massive data through a distributed data storage architecture, provides data support for decision algorithms; proposes a knowledge management mode of data warehouse and text knowledge fusion, realizes the fusion of existing knowledge and test data in the enterprise, lays a foundation for the extraction and mining of data knowledge; establishes a plug-in mechanism of the framework, which can expand the existing system function through the plug-in mode, realizes the expansion and dynamic updating of the system function.

[0026] The application provides a big data-based intelligent knowledge fusion platform decision system, which is based on the unique characteristics of a mechanical hydraulic product of a turboshaft / turboprop engine, optimizes the storage and query mode of massive data, provides data support for a decision algorithm, builds a bridge for existing text knowledge and test data of an enterprise through a data warehouse-text knowledge fusion, and lays a foundation for data knowledge extraction and mining, and realizes progressive updating and enhancement of platform functions through a plug-in mechanism of the product, so that the system can adapt to continuously updated decision requirements and function requirements.

[0027] In summary, the application provides systematic platform support for effective utilization of test data of a mechanical hydraulic product of a turboshaft / turboprop engine, knowledge extraction and test decision, and lays a foundation for intelligent debugging, troubleshooting and long test of the product. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 It is a data block concept;

[0029] Figure 2 It is a distributed file system architecture;

[0030] Figure 3 It is a distributed massive test data storage write flow;

[0031] Figure 4 It is a distributed massive test data storage read flow;

[0032] Figure 5 It is a three data model schematic diagram;

[0033] Figure 6 It is a data warehouse-text connection example;

[0034] Figure 7 It is a system plug-in function enhancement mechanism framework;

[0035] Figure 8 It is a system plug-in function enhancement mechanism timing diagram. DETAILED DESCRIPTION

[0036] The application will be further described below in combination with the drawings and examples, but it is not used as a basis for limiting the application.

[0037] Example 1. A turboshaft / turboprop engine mechanical hydraulic product test intelligent knowledge fusion platform, which is composed of Figures 1-8As shown, the platform includes distributed mass test data storage, data warehouse-text knowledge fusion and system plug-in function enhancement mechanism. The distributed mass data storage is the basis, providing a flexible and scalable solution for mass test data storage; on this basis, mass data is organized through a data warehouse, and the knowledge fusion method proposed in the application is used to link unstructured text data and structured data stored in the data warehouse through the form of URL, realizing the fusion of unstructured and structured data, laying the foundation for subsequent knowledge extraction and data mining; through plug-in design, the function of adding and updating decision algorithms at any time is realized. The system provides systematic platform support for the effective use of test data, knowledge extraction and test decision of turboshaft / turboprop engine mechanical hydraulic products, and lays the foundation for intelligent debugging, troubleshooting and long test of products. Specifically, the design of the distributed mass test data storage, data warehouse-text knowledge fusion and system plug-in function enhancement mechanism is as follows:

[0038] 1) Distributed mass data storage

[0039] When the turboshaft / turboprop engine mechanical hydraulic product test is carried out, a large amount of data will be generated during the test, so in order to realize the rapid storage of mass data generated during the test and realize the data scalability function, the application will establish a distributed mass test data storage architecture, and realize the distributed storage of mass data through a distributed file system. As shown in Figure 1 The concept of data block is abstracted on the distributed file system, which can be regarded as a special storage format of the distributed file system. When the system stores mass data, the distributed file system divides the data into multiple data blocks for storage, which not only simplifies the design of the storage system, but also is suitable for data backup, migration, expansion and other functions, improving the fault tolerance and scalability of the data. As shown in Figure 2 The distributed file system is mainly composed of NameNode and DataNode. NameNode mainly plays the role of metadata management service, responsible for managing the namespace of the entire file system, and DataNode as the working node of the distributed file system is responsible for storing data and providing data read / write services for the client. As shown in Figure 3 As shown in the implementation of the distributed file system write process, first, the Client and the NameNode establish a connection, view the metadata information through the NameNode and return the information to the Client, then the Client receives the DataNode list returned by the NameNode and establishes a connection with it to prepare to receive data. The Client processes the file in blocks, sends the data block and the list information returned by the NameNode to the DataNode. As shown in Figure 4As shown in the implementation of the distributed file system read process, the Client establishes a connection with the NameNode, the NameNode obtains the data block mapping information of the required file and the DataNode list storing the data block by querying the metadata information and sends them to the Client, and the Client obtains the information and finds the nearest available DataNode to retrieve the data block. The resource management and task scheduler include two core parts, namely the Resource Manager and the Node Manager. The Resource Manager accepts the distributed computing program submitted by the user and divides the resources for it; manages and monitors the resources on each Node Manager to balance the load. The Node Manager manages the computing resources of the machine, is responsible for accepting the tasks allocated by the Resource Manager, and creates containers and recycles resources. Finally, under the support of the distributed file system and the resource management and task scheduler, the massive data obtained in the experiment is stored in the distributed data warehouse.

[0040] 2) Data warehouse - text knowledge fusion

[0041] Relational information modeling and hierarchical information modeling are two common knowledge data description methods. The relational data model is to represent all data in the form of a table, the relationship in the table is defined by the column, and the relationship between tables is described by the primary key. The hierarchical data model is a modeling method for representing data through hierarchical structure, which expresses the relationship between data through parent-child structure, and XML is the most common hierarchical data model. As shown in Figure 5 , the application of these two data models is very mature, but the means of representing the relationship between data is relatively single, and lacks semantic information, and the data model structure established by these two methods is fixed and not easy to expand, and is not universal.

[0042] As shown in Figure 5 , the graph-based data model organizes data into the form of a directed graph, each node represents a data, which is also called a resource, and the concepts are connected by a directed line segment, which is called a property or a predicate. Unlike the relational data model, any two data in the graph-based data model can be connected through the property, and there is no hierarchical relationship, the expression form is flexible, and any new structure can be easily introduced into the existing data structure, avoiding the problem of not easy to expand of the relational data model and the hierarchical data model, and the data expression structure is universal.

[0043] The present application proposes a data warehouse-text knowledge fusion expression technology based on a graph data model, defines a data warehouse (data knowledge) and an existing text (text knowledge) of an enterprise as resources in a graph data model, connects the data warehouse and the text through attributes (directed line segments), assists knowledge extraction of the text through a search technology, assists knowledge extraction of the data warehouse through data mining, simultaneously extracts data warehouse-text correlation knowledge by using a graph traversal technology, and realizes deep mining and extraction of knowledge by combining the three technologies.

[0044] As shown in Figure 6 A data warehouse-text instance based on a graph data model is shown, three data warehouses and two texts exist here, and the data warehouse and the text are connected through a URL.

[0045] Use scenario 1: A user positions a text through searching, for example, queries a specific fault, the fault is connected to all related data warehouses through a URL, and the data in these data warehouses is all data related to the product fault, so we can provide the data of these data warehouses to a corresponding fault identification algorithm to train a fault decision model.

[0046] Use scenario 2: We find that data warehouse 2 is associated with text 1 and text 2 through graph correlation traversal, for example, long test fault data is associated with long test related text (for example, a long test outline) and fault related text (a long test fault report), so the two texts have correlation, and the system automatically associates the two texts. In this way, when we browse the long test outline, we can also associate all long test fault reports, thereby increasing the connection dimension of text knowledge.

[0047] 3) System plug-in function enhancement mechanism

[0048] The present application develops a system plug-in function enhancement mechanism, users can develop plug-ins of multiple types and multiple functions according to specifications, and arrange these plug-ins to form a task. Figure 7 As shown in the figure, the architecture of the system plug-in function enhancement mechanism mainly includes a user operation interface, a plug-in task pool, and a plug-in (built-in / self-defined). The platform user interface provides plug-in management and plug-in arrangement function interfaces. The plug-in can be built into the platform or uploaded to the platform by the user according to the self-defined plug-in conforming to the specification. The plug-in creates a corresponding plug-in task pool when it is used, and this mechanism can reduce the system overhead generated by frequent creation and recycling of task instances, and save storage resources. The system plug-in function enhancement mechanism is based on a workflow scheduling engine, and its running mode is as shown in Figure 8As shown, the user uploads a plug-in to the intelligent knowledge fusion platform, and the platform saves the plug-in to the local. Then a process is created, the platform generates a zip package by arranging the plug-in and uploads it to the workflow scheduling engine, and the workflow scheduling engine returns a workflow Id. After that, when the user executes / schedules the plug-in, the system submits a task to the resource management and task scheduler through the workflow engine, and finally the data warehouse is operated to realize the function of the entire plug-in.

[0049] Embodiment 2. An intelligent knowledge fusion platform for mechanical hydraulic product test of a turboshaft turboprop engine, which is composed as shown in Figures 1-8 As shown, the platform includes a distributed massive test data storage, a data warehouse-text knowledge fusion, and a system plug-in function enhancement mechanism; based on the distributed massive data storage, a flexible and scalable distributed storage solution is provided for massive data generated by tests; on this basis, the massive data is organized through the data warehouse, and based on the data warehouse-text knowledge fusion, the unstructured text data and the structured massive data stored in the data warehouse are linked through the form of URL to realize the fusion of unstructured and structured data, lay a foundation for subsequent knowledge extraction and data mining; through the system plug-in function enhancement mechanism, the function of adding and updating decision algorithms at any time is realized.

[0050] The flexible and scalable distributed storage solution is that the massive data is divided into multiple data blocks for storage.

[0051] The working process of the foregoing distributed massive data storage includes:

[0052] The NameNode module provides management services for metadata and is responsible for managing the file namespace of the entire distributed massive data storage;

[0053] The DataNode module is used as a working node of the distributed massive data storage and is responsible for storing the divided data blocks and providing read / write services for the client Client.

[0054] The write process of the foregoing distributed massive data storage is executed as follows:

[0055] First, the Client and the NameNode establish a connection, the NameNode checks the metadata information and returns it to the Client, then the Client receives the DataNode list returned by the NameNode and establishes a connection with it to prepare to receive data. The Client divides the file into blocks and sends the data blocks and the list information returned by the NameNode to the DataNode.

[0056] The read process of the foregoing distributed massive data storage is executed as follows:

[0057] The client establishes a connection with the NameNode, the NameNode obtains the data block mapping information of the required file and the DataNode list storing the data block by querying metadata information and sends them to the client, and the client obtains the information and finds the nearest available DataNode to retrieve the data block.

[0058] The foregoing working process of the distributed mass data storage further comprises:

[0059] The Resource Manager module in the resource management and task scheduler accepts the distributed computing program submitted by the user and divides resources for the program; and manages and monitors the resource situation on each Node Manager module to balance the load.

[0060] The Node Manager module in the resource management and task scheduler manages the computing resources of the machine, is responsible for accepting the task allocated by the Resource Manager module, and creates a container and recycles resources.

[0061] Under the support of the NameNode module, the DataNode module and the resource management and task scheduler, the mass data is stored in the data warehouse.

[0062] The foregoing working process of the data warehouse-text knowledge fusion is as follows:

[0063] The data warehouse-text knowledge fusion expression technology based on the graph data model defines the data warehouse and the existing text of the enterprise as resources in the graph data model, connects the data warehouse and the text through a directed line segment; the text extracts knowledge through a search technology, the data warehouse extracts knowledge through data mining, and the graph traversal technology is used to extract the correlation knowledge between the data warehouse and the text, so as to realize the deep mining and extraction of knowledge.

[0064] The architecture of the foregoing system plug-in function enhancement mechanism comprises a user operation interface, a plug-in task pool and a plug-in; the user operation interface provides a plug-in management and plug-in arrangement function interface; the plug-in creates a corresponding plug-in task pool when used; and the running mode of the system plug-in function enhancement mechanism is based on a workflow scheduling engine, and is specifically as follows:

[0065] The user uploads the plug-in to the intelligent knowledge fusion platform, the platform saves the plug-in to the local; then creates a process, the platform generates a zip package by arranging the plug-in and uploads it to the workflow scheduling engine, the workflow scheduling engine returns a workflow Id; then when the user executes / schedules the plug-in, the platform submits a task to the resource management and task scheduler through the workflow engine, and finally operates the data warehouse to realize the function of the entire plug-in.

Claims

1. A smart knowledge fusion system for testing mechanical and hydraulic products of turboshaft and turboprop engines, characterized in that: The system includes distributed massive experimental data storage, data warehouse-text knowledge fusion, and a system plugin enhancement mechanism. Based on distributed massive data storage, it provides a flexible and scalable distributed storage solution for the massive amounts of data generated from experiments. On this basis, the massive data is organized through a data warehouse. Based on data warehouse-text knowledge fusion, unstructured text data is linked with the structured massive data stored in the data warehouse via URLs, achieving the fusion of unstructured and structured data and laying the foundation for subsequent knowledge extraction and data mining. The system plugin enhancement mechanism enables the addition and updating of decision-making algorithms at any time. The data warehouse-text knowledge fusion process is as follows: The data warehouse-text knowledge fusion and expression technology based on graph data model defines the data warehouse and the enterprise's existing text as resources in the graph data model, and connects the data warehouse and text through directed line segments; the text uses search technology to assist in knowledge extraction, the data warehouse uses data mining to assist in knowledge extraction, and at the same time, it uses graph traversal technology to realize the extraction of data warehouse-text correlation knowledge, and finally realizes in-depth knowledge mining and extraction. The architecture of the system plugin functionality enhancement mechanism includes a user interface, a plugin task pool, and plugins. The user interface provides plugin management and orchestration functions. A corresponding plugin task pool is created when a plugin is used. The system plugin functionality enhancement mechanism operates based on a workflow scheduling engine, as detailed below: Users upload plugins to the intelligent knowledge fusion system, which saves the plugins locally. Then, a workflow is created, and the system generates a zip package by orchestrating the plugins and uploads it to the workflow scheduling engine. The workflow scheduling engine then returns the workflow ID. When a user executes / schedules a plugin, the system submits the task to the resource manager and task scheduler through the workflow engine, and finally operates the data warehouse to realize the functionality of the entire plugin.

2. The intelligent knowledge fusion system for testing mechanical and hydraulic products of turboshaft and turboprop engines according to claim 1, characterized in that, The flexible and scalable distributed storage solution involves dividing massive amounts of data into multiple data blocks for storage.

3. The intelligent knowledge fusion system for testing mechanical and hydraulic products of turboshaft and turboprop engines according to claim 2, characterized in that, The working process of distributed massive data storage includes: The NameNode module provides metadata management services and is responsible for managing the file namespace of the entire distributed massive data storage. The DataNode module acts as a worker node for distributed massive data storage, responsible for storing the segmented data blocks and providing data read and write services to the client.

4. The intelligent knowledge fusion system for testing mechanical and hydraulic products of turboshaft and turboprop engines according to claim 3, characterized in that, The write process for distributed massive data storage is executed as follows: First, the Client establishes a connection with the NameNode, views the metadata information through the NameNode, and returns the information to the Client. Then, the Client receives the list of DataNodes returned by the NameNode and establishes a connection with them to prepare for receiving data. The Client divides the file into blocks and sends the data blocks and the list information returned by the NameNode to the DataNodes.

5. The intelligent knowledge fusion system for testing mechanical and hydraulic products of turboshaft and turboprop engines according to claim 3, characterized in that, The read process for distributed massive data storage is executed as follows: The client establishes a connection with the NameNode. The NameNode obtains the data block mapping information of the required file and the list of DataNodes storing the data blocks by querying the metadata information, and sends it to the client. After receiving the information, the client finds the nearest available DataNode to retrieve the data blocks.

6. The intelligent knowledge fusion system for testing mechanical and hydraulic products of turboshaft and turboprop engines according to claim 3, characterized in that, The process of distributed massive data storage also includes: The Resource Manager module in the resource management and task scheduler accepts distributed computing programs submitted by users and allocates resources for them; it manages and monitors the resource status of each Node Manager module to facilitate load balancing. The Node Manager module in the resource management and task scheduler manages the computing resources of the machine, is responsible for accepting tasks assigned by the Resource Manager module, and creating containers and reclaiming resources; With the support of the NameNode module, DataNode module, and resource management and task scheduler, massive amounts of data are stored in the data warehouse.

Images