Multi-level equipment cross-platform fault knowledge recommendation and adaptive transfer method

Abstract

Provided in the present invention is a multi-level equipment cross-platform fault knowledge recommendation and adaptive transfer method, comprising: constructing an association relationship between product structure tree nodes and a fault knowledge base; calculating the similarity between first-level nodes of a start-domain product structure tree and a target-domain product structure tree, so as to obtain a similar first-level node pair set; calculating the similarity between child node sets of a similar parent node pair, and outputting a similar second-level node pair set; repeating the previous steps on a next-level node until a terminal node is reached, and outputting a final similar node pair set; calculating the fault case knowledge similarity between every two nodes in the similar node pair set, and excluding, from the similar node pair set, node pairs having relatively low similarities; and selecting a fault knowledge type that needs to be transferred, acquiring and copying, from node pairs having relatively high similarities, fault knowledge data of a start domain, modifying into a target domain a node to which related data is bound, and writing the related data into a database. The present invention can solve the problem of rapid transfer of fault knowledge between different items in avionics equipment.

Description

A multi-level equipment cross-platform fault knowledge recommendation and adaptive migration method TECHNICAL FIELD

[0001] The present application relates to the technical field of avionics, in particular to a multi-level equipment cross-platform fault knowledge recommendation and adaptive migration method. BACKGROUND

[0002] The aviation electronic information equipment product structure is complex, and the equipment depth is highly integrated, and the construction workload of the fault knowledge base is huge. At the same time, the aviation electronic information equipment platform is multiple, and there is a certain similarity between the product structure trees of different platforms, but there is a lack of good method to guide the fault knowledge migration between platforms, and the new platform often needs to rebuild the entire fault knowledge base, which is a huge workload and affects the development of aviation electronic information equipment model task. If the similarity between the fault knowledge bases of two projects can be identified, the fault knowledge migration work can be effectively guided, the rapid construction of the fault knowledge base of the new project can be realized, the repeated work of the design and operation personnel can be reduced, the aviation electronic equipment development efficiency can be improved, and the cross-project and cross-enterprise promotion of the fault knowledge intelligent platform can be promoted.

[0003] SUMMARY

[0004] The present application aims to provide a multi-level equipment cross-platform fault knowledge recommendation and adaptive migration method to solve the problem of rapid migration of fault knowledge between different projects in aviation electronic equipment.

[0005] The multi-level equipment cross-platform fault knowledge recommendation and adaptive migration method provided by the present application comprises the following steps:

[0006] S1, an association relationship between a product structure tree node and a fault knowledge base is constructed;

[0007] S2, the similarity of the parent nodes of the source domain product structure tree and the target domain product structure tree is calculated, and a node pair set with higher similarity is outputted;

[0008] S3, the similarity of the child node set of the similar parent node pair is calculated, the overall similarity of the child node set is weighted with the similarity calculated in step S2, and a parent node pair set with higher similarity is outputted;

[0009] S4, for the next level node similarity, steps S2-S3 are repeated until the terminal node, and the final similar node pair set is outputted;

[0010] S5, the fault case knowledge similarity of the two nodes in the similar node pair set is calculated, and the node pair with lower similarity is excluded from the similar node pair set;

[0011] S6, selecting the type of fault knowledge to be migrated, obtaining and copying the fault knowledge data of the source domain in the node pair with higher similarity, and modifying the associated node of the data to the target domain and writing into the database.

[0012] Further, step S1 comprises:

[0013] By manual or automatic association, the FMECA, failure mode library, maintenance process library and / or failure case library are hung on the corresponding sub-nodes for the source domain product structure tree; the existing target fault knowledge related data and failure case data are hung on the corresponding sub-nodes for the target domain product structure tree.

[0014] Further, step S2 comprises:

[0015] The product structure tree is converted into an algorithm-recognizable format;

[0016] The similarity of the first-level nodes of the source domain product structure tree and the target domain product structure tree is calculated; the node pairs with higher similarity and greater than the threshold value are added to the set of first-level nodes considered to be preliminarily similar;

[0017] All first-level nodes in the target domain product structure tree are traversed, and the set R1 of similar first-level node pairs is output.

[0018] Further, the calculation of node similarity is to calculate the literal similarity of the nodes by using Tanimoto coefficient.

[0019] Further, the calculation method of calculating the literal similarity of the nodes by using Tanimoto coefficient is as follows:

[0020] Wherein, E j (A, B) represents the similarity of vector A and vector B, A and B represent the vectors converted from the literal of the two nodes respectively, A·B represents the vector product, and ‖A‖ 2 represents the modulus of the vector.

[0021] Further, step S3 comprises:

[0022] For the similar first-level node pair (TN1, N1) ∈ R1, the similarity of the sub-nodes of TN1 node and the sub-nodes of N1 node is calculated respectively, a group of nodes with the highest similarity and greater than the threshold value are added to the set R2 of second-level nodes considered to be preliminarily similar, and the step is repeated for all sub-nodes of TN1 node;

[0023] Finally, the proportion of the similar sub-nodes found in TN1 node is calculated and weighted with the similarity of the parent node to obtain the final node similarity;

[0024] Repeat step S3 for each similar primary node pair in the similar primary node pair set R1 to obtain the final similarity of each node and eliminate the similar primary node pairs whose similarity does not meet the requirement from the similar primary node pair set R1.

[0025] Further, the overall similarity calculation formula of the child node set is:

[0026] The final node similarity is E(TN1, N1) = aE j + bE p

[0027] wherein E j is the literal similarity of the primary node itself, E p is the overall similarity of the child nodes, a and b are weighting coefficients, and a + b = 1.

[0028] Further, step S4 includes:

[0029] Repeat steps S2-S3 for all similar secondary node pairs in the similar secondary node pair set R2, eliminate the similar secondary node pairs whose final similarity does not meet the requirement from the similar secondary node pair set R2, and obtain a similar tertiary node pair set R3. If the tertiary node is a terminal node, jump to step S5, otherwise continue steps S2-S3 until the terminal node is reached.

[0030] In the case of retaining the node hierarchy relationship, merge the similar node pair sets of the nodes at each level to obtain the recommended similar node pair set R.

[0031] Further, step S5 includes:

[0032] For all similar node pair sets that need to be analyzed for migratability, if the fault case knowledge set of the target domain node reaches the set proportion of the target domain, extract the case knowledge base thereof as a comparison unit, carry out comparison between the source domain case set A and the target domain case set B based on case matching, set a similarity threshold, and calculate the migratability of the knowledge based on case coverage for the source domain case set A and the target domain case set B respectively, and eliminate the nodes with lower migratability from the recommended similar node pair set R.

[0033] Further, step S6 includes:

[0034] Automatically recommend the similar node pair set generated in step S5, confirm whether all nodes need to be migrated by the user, select the fault knowledge types that need to be migrated, automatically obtain and copy the fault knowledge data of the source domain in the similar node pair with higher similarity in the background, modify the belonging node bound to the relevant data to the target domain, and write into the database.

[0035] In conclusion, by using the technical scheme, the application has the following beneficial effects:

[0036] The application can realize rapid construction of a new project fault knowledge base, reduce repetitive work of designers, improve reusability of fault knowledge of old equipment, and has important significance and value. By identifying similarity between two project fault product structures and the knowledge base, rapid construction of a new project fault knowledge base is guided, reuse of fault knowledge of old projects is realized, repetitive work of designers and maintenance personnel is reduced, aviation electronic equipment development efficiency is improved, and cross-project and cross-enterprise promotion of the fault knowledge intelligent platform is promoted. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as a limitation on the scope. Other related drawings can also be obtained by those skilled in the art without creative labor on the basis of the drawings.

[0038] Fig. 1 is a flowchart of a multi-level equipment cross-platform fault knowledge recommendation and adaptive migration method according to an embodiment of the application.

[0039] Fig. 2 is a schematic diagram of a fault case similarity determination method according to an embodiment of the application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the application more clear, the technical solutions in the embodiments of the application will be described clearly and completely in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are some embodiments of the application, but not all embodiments. The components of the embodiments of the application described and shown in the drawings can be arranged and designed in various different configurations.

[0041] Therefore, the following detailed description of the embodiments of the application provided in the drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the application. All other embodiments obtained by those skilled in the art on the basis of the embodiments in the application without creative labor are within the scope of protection of the application.

[0042] EMBODIMENT

[0043] As shown in Fig. 1, the embodiment proposes a multi-level equipment cross-platform fault knowledge recommendation and adaptive migration method, including the following steps:

[0044] S1, an association relationship between a product structure tree node and a fault knowledge base is constructed;

[0045] Specifically, by means of manual or automatic association, FMECA, failure mode library, maintenance process library, failure case library and the like are hung on the corresponding sub-nodes for the source domain product structure tree; the existing target failure knowledge related data and failure case data are hung on the corresponding sub-nodes for the target domain product structure tree.

[0046] S2, similarity of parent nodes of the source domain product structure tree and the target domain product structure tree is calculated, and a node pair set with higher similarity is output.

[0047] Specifically, the product structure tree is converted into an algorithm recognizable format, and in the embodiment, the format is a multi-dimensional list unfolded in the hierarchical relationship of the product structure tree, for example, a three-layer product structure tree of a source domain is defined and converted into: Start_tree={ [N1, N2], [[N1_1, N1_2, N1_3, N1_4], [N2_1, N2_2, N2_3] ], [[[N1_1_1, N1_1_2, N1_1_3], […], …, […], [N2_3_1, N2_3_2]]]}

[0048] Wherein, N1 and N2 are first-level parent nodes, N1_1, N1_2, N1_3 and N1_4 are sub-nodes of N1, and N1_1_1, N1_1_2 and N1_1_3 are lower-level sub-nodes of N1_1.

[0049] Similarly, a three-layer product structure tree of a target domain is defined and converted into: Target_tree={ [TN1, TN2, TN3], [[TN1_1, TN1_2TN1_3], [TN2_1, TN2_2], [TN3_1, TN3_2] ], [[[TN1_1_1, TN1_1_2, TN1_1_3], […], …, […], [TN3_2_1, TN3_2_2]]]}

[0050] Then, the similarity of the first-level nodes of the source domain product structure tree Start_tree and the target domain product structure tree Target_tree is calculated. For the TN1 node, the Tanimoto coefficient is used to calculate the literal similarity of the TN1 node and the N1 node and the N2 node respectively, if E j (TN1, N1) > E j (TN1, N2), and E j (TN1, N1) is greater than a threshold value, (TN1, N1) is added to the first-level node set considered to be preliminarily similar, all first-level nodes in the target domain product structure tree are traversed, and a similar first-level node pair set R1 is output.

[0051] Wherein, the calculation method of calculating the literal similarity of the nodes by using the Tanimoto coefficient is as follows:

[0052] wherein A, B represent the vectors converted from two node texts respectively, each element in the set represents a dimension in the vector, and the value in each dimension is usually between [0, 1], A·B represents the vector product, and ‖A‖ 2 represents the modulus of the vector, i.e.

[0053] S3, the similarity of the child node set of the similar parent node pair is calculated, the overall similarity of the child node set is weighted with the similarity calculated in step S2, and the parent node pair set with higher similarity is output.

[0054] Specifically, for a similar first-level node pair (TN1, N1) e R1, the similarity of the child node TN1_1 of the TN1 node and the child nodes N1_1, N1_2, N1_3, N1_4 of the N1 node is calculated respectively, if the most similar node pair is (TN1_1, N1_2) and the similarity is greater than a threshold, (TN1_1, N1_2) is added to the similar second-level node pair set R2 considered to be preliminarily similar, and all child nodes of TN1 are traversed to repeat the step. Finally, the proportion of the similar child nodes found in the TN1 node is calculated and weighted as the overall similarity of the child node set and the parent node similarity to obtain the final node similarity.

[0055] The overall similarity calculation formula of the child node set is:

[0056] The final node similarity is E (TN1, N1) = aE j + bE p

[0057] wherein E j is the text similarity of the first-level node itself, E p is the overall similarity of the child node, and a, b are weighting coefficients, a + b = 1.

[0058] For each similar first-level node pair in the similar first-level node pair set R1, step S3 is repeated to obtain the final similarity of each node, and the similar first-level node pair whose similarity does not meet the requirement is removed from the similar first-level node pair set R1.

[0059] S4, for the next level node similarity, steps S2-S3 are repeated until the terminal node, and the final similar node pair set is output.

[0060] Specifically, for all similar secondary node pairs in the similar secondary node pair set R2, steps S2-S3 are repeated, the similar secondary node pairs whose final similarity does not meet the requirement are removed from the similar secondary node pair set R2, and the similar tertiary node pair set R3 is obtained. Since the example product structure tree is three layers, the tertiary node is a terminal node, and step S5 is jumped to, otherwise steps S2-S3 are continued until the terminal node is reached.

[0061] The similar node pair set R is obtained by merging the similar node pair sets of each hierarchical node under the condition of retaining the hierarchical relationship of the nodes.

[0062] S5, the similarity of the fault case knowledge of each two nodes in the similar node pair set is calculated, and the node pair with lower similarity is excluded from the similar node pair set.

[0063] Specifically, for all similar node pair sets that need to be analyzed for migratability, if the fault case knowledge set of the target domain node reaches 1 / 3 of the target domain, the case knowledge base thereof is extracted as a comparison unit, and the comparison between the source domain case set A and the target domain case set B is carried out based on case matching. A similarity threshold is set, as shown in FIG. 2, and the migratability of the knowledge is calculated based on the case coverage for the source domain case set A and the target domain case set B respectively. The node with lower migratability is removed from the recommended similar node pair set R.

[0064] S6, the type of fault knowledge that needs to be migrated is selected, the fault knowledge data of the source domain in the node pair with higher similarity is obtained and copied, and the related data binding is modified to belong to the target domain and written into the database.

[0065] Specifically, the similar node pair set generated in step S5 is automatically recommended, the user confirms whether all nodes need to be migrated, selects the type of fault knowledge (fault mode, FMECA, maintenance process, etc.) that needs to be migrated, and the background automatically obtains and copies the fault knowledge data of the source domain in the node pair with higher similarity, modifies the related data binding to belong to the target domain, and writes into the database.

[0066] The multilevel equipment cross-platform fault knowledge recommendation and adaptive migration method can be applied to multiple fields. First, the association between the product structure tree nodes and the fault knowledge base is constructed; the node similarity of the source domain product structure tree and the target domain structure tree is calculated, and the node pair set with higher similarity is output; the fault case similarity of each node pair is calculated, and the node pair with lower similarity is excluded from the similar node set; the fault knowledge data of the source domain in the node pair with higher similarity is obtained and copied, and the relevant data binding is modified to the target domain and written into the database. The application can realize the rapid construction of the fault knowledge base of the new project, reduce the repetitive work content of the designer, improve the reusability of the fault knowledge of the old equipment, and has important significance and value. By identifying the similarity between the fault product structure and the knowledge base of two projects, the rapid construction of the fault knowledge base of the new project is guided, the reuse of the fault knowledge of the old project is realized, the repetitive work of the designers and operators is reduced, the efficiency of the avionics equipment development is improved, and the cross-project and cross-enterprise promotion of the fault knowledge intelligent platform is promoted.

[0067] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration, characterized in that, The method comprises the following steps: S1, constructing the association relationship between the product structure tree nodes and the fault knowledge base; S2, calculating the similarity of the parent nodes of the source domain product structure tree and the target domain product structure tree, and outputting the node pair set with higher similarity; S3, calculating the similarity of the child node set of the similar parent node pair, weighting the overall similarity of the child node set with the similarity calculated in step S2, and outputting the parent node pair set with higher similarity; S4, repeating steps S2-S3 for the next level node similarity until the terminal node is reached, and outputting the final similar node pair set; S5, calculating the fault case knowledge similarity of each two nodes in the similar node pair set, and excluding the node pair with lower similarity from the similar node pair set; S6, selecting the fault knowledge type to be migrated, obtaining and copying the fault knowledge data of the source domain in the node pair with higher similarity, modifying the associated node of the relevant data to the target domain, and writing into the database. 2.The multi-tiered equipped cross-platform fault knowledge recommendation and adaptive migration method of claim 1, wherein, Step S1 comprises: Through manual or automatic association, the FMECA, fault mode library, maintenance process library and / or fault case library are hung on the corresponding child nodes for the source domain product structure tree; the existing target fault knowledge related data and fault case data are hung on the corresponding child nodes for the target domain product structure tree.

3. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 1, wherein, Step S2 comprises: The product structure tree is converted into a format recognizable by the algorithm; The similarity of the first level nodes of the source domain product structure tree and the target domain product structure tree is calculated; the node pair with higher similarity and greater than the threshold value is added to the first level node set considered to be preliminarily similar; All first level nodes in the target domain product structure tree are traversed, and the similar first level node pair set R1 is output. The node similarity is calculated by using the Tanimoto coefficient to calculate the literal similarity of the nodes.

4. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 3, wherein, Step S3 comprises:

5. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 4, wherein, The calculation method of calculating the word similarity of nodes by using the Tanimoto coefficient is as follows: wherein E j (A, B) denotes the similarity between vector A and vector B, A, B are respectively represented as vectors converted from two node texts, A·B denotes vector multiplication, and ‖A‖ 2 denotes the modulus of a vector.

6. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 3, wherein, For the similar first level node pair (TN1, N1) ∈ R1, the similarity of the child nodes of TN1 and the child nodes of N1 is calculated respectively, a group of nodes with the highest similarity and greater than the threshold value is added to the second level node pair set R2 considered to be preliminarily similar, and the step is repeated for all child nodes of TN1; Finally, the proportion of the similar child nodes found in the TN1 node is calculated and weighted as the overall similarity of the child node set and the parent node similarity to obtain the final node similarity; For each similar first level node pair in the similar first level node pair set R1, step S3 is repeated to obtain the final similarity of each node, and the similar first level node pair with a similarity that does not meet the requirement is excluded from the similar first level node pair set R1. Step S4 comprises:

7. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 6, wherein, The overall similarity calculation formula of the child node set is: The final node similarity is: E(TN1,N1) = aE j + βE p wherein E j is the text similarity of the primary node itself, E p is the overall similarity of its child nodes, and α, β are weighting coefficients, with α + β = 1.

8. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 6, wherein, For all similar second level node pairs in the similar second level node pair set R2, steps S2-S3 are repeated, the similar second level node pair with a final similarity that does not meet the requirement is excluded from the similar second level node pair set R2, and the similar third level node pair set R3 is obtained; if the third level node is a terminal node, step S5 is jumped to, otherwise steps S2-S3 are continued until the terminal node is reached; The similar node pair sets of the nodes at each level are combined to obtain the recommended similar node pair set R under the condition that the node hierarchy relationship is preserved. Step S5 comprises:

9. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 8, wherein, ​ For all similarity node pair sets requiring migration analysis, if the target domain node exists in the fault case knowledge set reaching the set proportion of the target domain, extract its case knowledge base as the comparison unit, carry out the comparison between the source domain case set A and the target domain case set B based on the case matching; set the similarity threshold, and calculate the migration of the knowledge according to the case coverage of the source domain case set A and the target domain case set B respectively, and remove the nodes with lower migration from the recommended similarity node pair set R.

10. The multi-tiered equipped cross-platform failure knowledge recommendation and adaptive migration method of claim 1, wherein, Step S6 comprises: The automatic recommendation step S5 produces a similarity node pair set, the user confirms whether all nodes need to be migrated, selects the fault knowledge type that needs to be migrated, the background automatically acquires and copies the fault knowledge data of the source domain in the node pair with higher similarity, modifies the belonging node bound to the related data to the target domain, and writes into the database. ​

Images