Knowledge graph-based helicopter structural member inspection planning and compiling method

Abstract

The invention provides a helicopter structural part inspection planning compiling method based on a knowledge graph. The method comprises the following steps: issuing the knowledge graph based on GraphML; reconstructing a knowledge graph structure oriented to a three-dimensional detection planning environment; extracting MBD model detection information of the helicopter structural component to obtain a to-be-detected element set; a three-dimensional inspection planning scheme set meeting requirements is obtained based on map matching; recommending an inspection planning scheme based on to-be-inspected element collaborative filtering and a minimum use frequency elimination and replacement mechanism; in the method, a detection planning business rule of the helicopter structural component is converted into a knowledge graph, retrieval recommendation is performed on the basis of detection demand information improved from a structural component model in a detection scheme compiling link to obtain a detection planning scheme set, and on the basis, a collaborative filtering and least use frequency elimination conversion method is applied to obtain a detection planning scheme set. And an optimal detection planning scheme is recommended.

Description

Technical Field

[0001] The present application belongs to the field of digital detection and automation technology, and in particular, relates to a knowledge graph-based inspection planning method for helicopter structural parts. Background Art

[0002] The production mode of helicopter structural parts is characterized by multiple varieties and small batches. With the continuous deepening of the application of intelligent manufacturing production lines in the manufacturing of helicopter structural parts, the production capacity and production rhythm of helicopter structural parts have gradually increased, which has put forward higher requirements on the efficiency of detection planning in the process preparation link before manufacturing.

[0003] Traditional inspection planning methods rely heavily on the experience and operational proficiency of technicians, and are unable to meet the current needs for efficient production of helicopter structural parts. Summary of the invention

[0004] Purpose of the invention: Based on this background, this method introduces knowledge graphs into the detection planning process to improve the efficiency and intelligence level of detection planning.

[0005] The present application provides a method for compiling a helicopter structural component inspection plan based on a knowledge graph, the method comprising: Release of knowledge graph based on GraphML; Reconstruction of knowledge graph structure for 3D inspection planning environment; Extract the MBD model inspection information of helicopter structural parts and obtain the collection of elements to be inspected; Obtain a collection of 3D inspection planning solutions that meet the requirements based on graph matching; The inspection planning scheme is recommended based on collaborative filtering of the elements to be inspected and the least frequently used elimination and replacement mechanism. Preferably, the GraphML-based knowledge graph publishing includes: Convert the detection planning knowledge graph in Neo4j into a GraphML file to complete the publication of the knowledge graph.

[0006] Preferably, the knowledge graph structure reconstruction for the three-dimensional detection planning environment includes: Based on the knowledge graph structure reconstruction algorithm, the knowledge graph structure is reconstructed.

[0007] Preferably, the extraction of helicopter structural component MBD model detection information to obtain a collection of elements to be detected includes: The CATITPS, CATITTRS and CATIRGE series interfaces in CATIA CAA are used to extract annotation type, theoretical value, upper and lower deviations, datum and geometric feature data from the helicopter component MBD model. Then, structured parsing and storage are performed in accordance with GD&T rules to obtain a collection of elements to be inspected.

[0008] Preferably, the obtaining of a collection of three-dimensional inspection planning solutions that meet the requirements based on graph matching includes: Based on the 3D inspection planning rule matching algorithm, a collection of 3D inspection planning solutions that meet the requirements is obtained.

[0009] Preferably, the recommendation of the inspection planning scheme based on collaborative filtering of elements to be inspected and the least frequently used elimination and replacement mechanism includes: Based on collaborative filtering and least frequent elimination algorithm, a test planning scheme is obtained.

[0010] Preferably, the method further comprises: Collaborative filtering and least frequent elimination algorithm are used to recommend the most optimized inspection planning scheme.

[0011] Preferably, the method further comprises: Generate a GraphML knowledge graph.

[0012] Beneficial technical effects of this application: In this method, the inspection planning business rules of helicopter structural parts are converted into a knowledge graph. In the inspection plan preparation stage, retrieval and recommendation are performed based on the inspection requirement information improved from the structural parts model to obtain a collection of inspection planning schemes. On this basis, collaborative filtering and least frequent replacement algorithms are applied to recommend the most optimized inspection planning scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a flow chart of a method for preparing a helicopter structural component inspection plan based on a knowledge graph provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] See also Figure 1 , the present application provides a method for compiling a helicopter structural component inspection plan based on a knowledge graph, comprising the following steps: Step 1: Publishing knowledge graph based on GraphML The knowledge information of helicopter structural parts stored in the knowledge graph is stored in the form of Neo4j graph database, and the three-dimensional inspection planning environment of helicopter structural parts is a three-dimensional industrial software based on the CAD engine, which is the CATIA environment in this method. Therefore, the inspection planning knowledge graph in Neo4j is first converted into a GraphML file to complete the publication of the knowledge graph.

[0015] Step 2: Reconstruct the knowledge graph structure for the 3D detection planning environment.

[0016] The algorithm for reconstructing the knowledge graph structure is as follows: Input: KG.graphml file; Output: graph-based knowledge graph object (KG). BEGIN reads KG.graphml file IF the file does not exist, exit the entire process ELSE Go to step 5 WHILE locate the tag "graph" Enter the tag "graph" to create a knowledge graph instance Graph WHILE locate any tag IF label is "node" Create a node instance Node and get the nodeId and Label Traverse the "data" tag to get all the attributes of the Node Add the Node to the nodes collection of the Graph ELSE IF label is "edge" Create an edge instance Edge and obtain the edge ID and Label Get the source node and target node of the edge from the node collection of the Graph, put the target node into the outNodes collection of the source node, and put the source node into the inNodes collection of the target node Traverse the "data" tag to get all Edge properties Add Node to the edges collection of Graph Exit the loop and return to GraphEND Step 3: Extraction of helicopter structural parts MBD model detection information.

[0017] The CATITPS, CATITTRS and CATIRGE series interfaces in CATIA CAA are used to extract annotation type, theoretical value, upper and lower deviations, datum and geometric feature data from the helicopter component MBD model. Then, structured parsing and storage are performed in accordance with GD&T rules to obtain a collection of elements to be inspected.

[0018] Step 4: Obtain a collection of 3D inspection planning solutions that meet the requirements based on graph matching 3D inspection planning rule matching algorithm Input: MBD model, 3D inspection planning rule knowledge graph Output: 3D inspection planning scheme collection BEGIN Extract the current inspection planning link fact set Take out rules WHILE Starting from the starting node of the current link, BFS search can verify conditions ( ) nodes and calculate

[0019] IF The next logical operation is "AND" DFS lookup can verify conditions ( )node IF cannot find the next condition that satisfies the condition ( ) Node OR The next logical operation is "or" Return to step 5 ELSE Go to step 12 ELSE Return to step 5 IF

[0020] Iterating over a collection , starting from the starting node of the current link, BFS searches for the conclusion that meets the conditions ( ) node, add it to the conclusion set ELSE Go to step 15 ENDStep 5: Recommendation of inspection planning schemes based on collaborative filtering of elements to be inspected and LFU (least frequent use) replacement mechanismInput of inspection planning scheme recommendation algorithm based on collaborative filtering of elements to be inspected and LFU replacement mechanism: Elements to be inspected , the set of solutions obtained by rule matching , a collection of similar cases obtained through case matching ,in For all the schemes in the rule base, For all cases in the case library.

[0021] Output: sorted set of solutions .

[0022] The middle plan for the elements to be tested Priority Matrix ,in For the plan The priority of the current element to be inspected.

[0023] calculate The eigenvector of as well as The eigenvector of Among them, each component of the feature vector represents a quantitative attribute of the same type of basic elements in the attribute set of the element to be inspected (or case, the element to be inspected becomes a case after decision), such as quality characteristic accuracy requirements, geometric parameters of associated features, etc.

[0024] Normalize the data in the feature vector and the calculation method is as follows

[0025] calculate and . (Using Pearson's correlation coefficient as the calculation method)

[0026] The feature vector Substituting into the above formula, we get

[0027] Will Recorded as , and get the similarity matrix

[0028] Get every case A collection of solutions

[0029] calculate and The intersection of

[0030] Calculate according to the order of the solutions in the set Each solution The priority for this case , the calculation method is as follows:

[0031] in for In the collection . (For example: if it is ranked first, ) The priority calculated in the previous step is recorded as the priority matrix

[0032] The similarity matrix With the priority matrix Multiply them together and you get Each solution Elements to be inspected Priority Matrix

[0033] According to the priority matrix Aggregate solutions Sorting is performed. When the calculated priorities are the same, according to the LFU principle, the solution with more usage times within a certain period of time is placed in front. If the number of usage times is the same, the solution added to the rule base later is placed in front.

Claims

1. A method for preparing inspection plan for helicopter structural parts based on knowledge graph, characterized in that: The method comprises: Release of knowledge graph based on GraphML; Reconstruction of knowledge graph structure for 3D inspection planning environment; Extract the MBD model inspection information of helicopter structural parts and obtain the collection of elements to be inspected; Obtain a collection of 3D inspection planning solutions that meet the requirements based on graph matching; The inspection planning scheme is recommended based on collaborative filtering of the elements to be inspected and the least frequently used elimination and replacement mechanism.

2. The method according to claim 1, characterized in that The GraphML-based knowledge graph publishing includes: Convert the detection planning knowledge graph in Neo4j into a GraphML file to complete the publication of the knowledge graph.

3. The method according to claim 2, characterized in that The knowledge graph structure reconstruction for the three-dimensional detection planning environment includes: Based on the knowledge graph structure reconstruction algorithm, the knowledge graph structure is reconstructed.

4. The method according to claim 3, characterized in that The MBD model detection information of the helicopter structure is extracted to obtain a collection of elements to be inspected, including: The CATITPS, CATITTRS and CATIRGE series interfaces in CATIA CAA are used to extract annotation type, theoretical value, upper and lower deviations, datum and geometric feature data from the helicopter component MBD model. Then, structured parsing and storage are performed in accordance with GD&T rules to obtain a collection of elements to be inspected.

5. The method according to claim 4, characterized in that The method of obtaining a collection of three-dimensional inspection planning solutions that meet the requirements based on graph matching includes: Based on the 3D inspection planning rule matching algorithm, a collection of 3D inspection planning solutions that meet the requirements is obtained.

6. The method according to claim 5, characterized in that The inspection planning scheme recommended based on collaborative filtering of elements to be inspected and the least frequently used elimination and replacement mechanism includes: Based on collaborative filtering and least frequent elimination algorithm, a test planning scheme is obtained.

7. The method according to claim 6, characterized in that The method also includes: applying collaborative filtering and least-used-frequency elimination algorithms to recommend an optimal inspection planning solution.

8. The method according to claim 7, characterized in that The method further comprises: Generate a GraphML knowledge graph.

Images