Seabed data shelter distribution design method based on knowledge graph

Abstract

The invention provides a submarine data shelter distribution design method based on a knowledge graph, and belongs to the field of ocean engineering. According to the method, the relationship among various information of typical submarine data shelters in history can be integrated by constructing a distribution design knowledge graph system framework of the submarine data shelters; through a mode of fusing different knowledge maps, redundant information can be merged to improve reusability of knowledge objects, information contradictions among different knowledge maps are reduced, a system error-tolerant rate is reduced, and the fused knowledge maps are more universal in a submarine data shelter project; according to the submarine data square cabin obtained through the optimal process of submarine data square cabin distribution design of the method, factors in all aspects and the relation between different elements are comprehensively considered, and the constructed submarine data square cabin is more environmentally friendly and safer.

Description

Technical Field

[0001] The present invention relates to the field of marine engineering, and in particular to a method for designing the distribution of seabed data cabins based on a knowledge graph. Background Art

[0002] With the continuous advancement of the information industry, the demand for data storage is increasing day by day. Since the data storage process requires the release of a large amount of heat, the data storage points need to be cooled and heat-dissipated. Placing the data cabin in the marine environment for storage and taking away the heat through undercurrents and high-speed water flow is the cheapest way of heat dissipation and is widely respected in the industry.

[0003] Different sea area locations, different tidal conditions in different seasons in the same sea area, and different plane positions of individual cabins all affect the heat dissipation effect of the submarine data center (cabin group). How to arrange the submarine data cabin on the seabed is a major problem in the industry.

[0004] The present invention proposes a method for designing the distribution of seabed data shelters based on knowledge graphs. The method constructs a knowledge graph system framework for the distribution of seabed data shelters, establishes the relationship between various information in the knowledge graph system, formulates constraint rules, draws on experience from similar engineering cases (machine learning), intelligently selects the best solution parameter configuration from the knowledge graph system, and proposes the optimal design solution for the distribution of seabed data shelters. Summary of the invention

[0005] The purpose of the present invention is to overcome the above problems existing in the prior art and greatly improve its technical effect on the basis of the original technology; firstly, the present invention provides a method for designing the distribution of seabed data cabins based on knowledge graph, which method includes:

[0006] (I) The object-oriented method is used to divide the knowledge graph system of the seabed data shelter distribution design into the design element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph; the object-oriented method means that the established knowledge graph focuses on universality, classifies the cumbersome process flow into objects according to their commonalities, combines the information and the behavioral source of the information as an object mapping set, and forms a behavioral, functional, modular and reusable universal structural system; the design element knowledge graph is constructed through knowledge triples<Component,Component_feature,Feature> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Feature is the tail entity, which indicates the feature information of the design element; Component_feature is the mapping relationship between the head entity and the tail entity; the layout parameter knowledge graph is represented by triples<Component,Component_parameter,Parameter> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Parameter is the tail entity, which indicates the specific layout parameter information; Component_parameter is the mapping relationship between the head entity and the tail entity, which is the specific layout method, which is determined by the construction plan and the design calculation report; the layout effect knowledge graph refers to the quality control of the knowledge graph, and the layout effect knowledge graph is constructed through triples<Item,Item_requirement,Requirement> Item is the head entity, which consists of the quality control item code Item-ID and the quality control item name Item-Name; Requirement is the tail entity, which indicates the quality control item threshold; Item_requirement is the mapping relationship between the head entity and the tail entity, which is determined by the data cabin maximum operating temperature requirement table, material performance table, equipment manual, and relevant industry technical standards and specifications;

[0007] (II) The method of entity alignment of graph convolutional neural network (GCN) is used to fuse the element knowledge graph, layout parameter knowledge graph and layout effect knowledge graph; the formula for knowledge fusion is:

[0008]

[0009] Among them, s is the activation function, P is an n×n adjacency matrix, representing the association information between two knowledge graphs, and n is the number of feature nodes; I is the n×n identity matrix; H input is the input data driven by the encoding in a knowledge graph, which is the vertex feature matrix; for The diagonal node degree matrix, W is the weight matrix, H out put Encode the associated output data matrix in another knowledge graph;

[0010] (III) After the knowledge graph is integrated through (I) and (II), an upper ontology of construction knowledge based on the knowledge graph is formed; subsequently, the upper ontology of the integrated knowledge graph is used to perform scheme preview and design parameter intelligent decision-making on the design data information of the specific project; finally, the upper ontology of the integrated knowledge graph is processed by distributing the design knowledge graph of the submarine data cabins of specific historical projects, and the data sources of the submarine data cabins of each specific project are combined to perform scheme preview and design parameter intelligent decision-making on the upper ontology of the integrated knowledge graph, thereby generating an instantiation of the upper ontology of the first part of the construction knowledge, and generating a knowledge graph of the submarine data cabins of specific historical projects. Usually, the data sources of the submarine data cabins of different projects correspond to different intelligent decision-making schemes for design parameters;

[0011] (iv) forming a data knowledge base of the seabed data cube through the data sources of the seabed data cubes of specific historical projects, and forming a graph knowledge base of the seabed data cube through the knowledge graphs of the seabed data cubes of specific historical projects created by (i), (ii) and (iii);

[0012] (V) When constructing the knowledge graph for the target seabed data cabin, first, the design factor similarity calculation method is used to find the seabed data cabins of historical specific projects similar to the target seabed data cabin in the data knowledge base; then, the approximate process of the distribution design of the target seabed data cabin is obtained through the knowledge graph in the graph knowledge base corresponding to the seabed data cabins of similar historical specific projects; finally, the constraint rule decision tree method is used to conditionally constrain each key part and functional link in the approximate process, so as to realize the intelligent planning of the distribution design of the target seabed data cabin.

[0013] Specifically, the object-oriented method is used to design the knowledge graph system for the distribution of the seabed data cabins, including: the knowledge graph is usually structured in triples<h,r,e> The knowledge information is described, where h and e are nodes in the knowledge graph, and r is the relationship between nodes in the knowledge graph; the creation of the construction knowledge graph includes extracting the construction knowledge information from the unstructured text, representing the construction information in a structured form, and then fusing the interrelated construction information to generate a visual construction knowledge graph; the design knowledge graph system refers to the representation of design knowledge.

[0014] Specifically, the graph convolutional neural network (GCN) entity alignment method includes: the method uses GCN to learn and embed graph structure data, thereby realizing entity matching in different knowledge graphs; entity alignment is a key task in knowledge graph fusion, and its goal is to identify and match different nodes representing the same real-world entities in different knowledge graphs; when using the graph convolutional neural network (GCN) entity alignment method to fuse the element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph, the element knowledge graph and the layout parameter knowledge graph are first fused through the knowledge fusion formula, and then the fused knowledge graph entities and entities in the layout effect knowledge graph are fused, and finally the element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph are fused into a knowledge graph with the main entity characteristics of the three.

[0015] Specifically, the design data information involved in the specific project of using the seabed data cabin data source to conduct scenario preview and intelligent decision-making on design parameters for the upper ontology of the fused knowledge graph includes: the purpose of the scenario preview is to predict and optimize the performance of the knowledge graph in various scenarios by simulating and evaluating different ontology design schemes before actual application; the scenario preview steps of the knowledge graph by the data source include: data source analysis, ontology mapping, data integration and scenario preview; the purpose of the intelligent decision-making of design parameters is to manage, analyze and utilize large amounts of data more efficiently and accurately, so as to support various intelligent applications and decision-making processes; the steps of the intelligent decision-making of design parameters include: determining the design target data information, parameter selection, determination, implementation and testing of intelligent decision-making methods.

[0016] Specifically, the design factor similarity calculation method includes: using an edit distance algorithm for similarity calculation, traversing and comparing the data S of a specific target seabed data cabin with a data knowledge base corresponding to a typical engineering case solution of a seabed data cabin in history as conditional data T, performing similarity analysis on the description text statements of the two, and selecting a scheme including the one with the highest similarity and several typical schemes with high similarity as preliminary design schemes; the edit distance algorithm refers to the minimum number of editing operations required to change from the target data S to the conditional data T, represented by the symbol LD; the steps of calculating the similarity are:

[0017] Based on the target seabed data cabin distribution design knowledge graph and extract its data source, select m key indicator parameters, and construct a 1×m target data matrix. Select m key indicator parameters and construct a 1×m target data matrix S = (s1, s2, ...s m ); Taking the typical design schemes in history as dependent variables, extract the knowledge of the upper ontology of the submarine data cabin distribution design knowledge data source, select n key indicator parameters as independent variables, and construct a 1×n conditional data matrix T=(t1, t2, …t n ); construct a (m+1)×(n+1) order matching relationship matrix D between S and T:

[0018] D (m+1)×(n+1) ={d i,j}(0≤i≤m,0≤j≤n,)

[0019] Define the element d in the matrix D according to the following relationship i，j :

[0020]

[0021] Among them, a ij To edit the cost, express it according to the following rules:

[0022]

[0023] After the matrix D is constructed, the lower right corner element d mn That is, the edit distance LD between the target data matrix S and the conditional data matrix T, that is, LD = d mn ; At this time, the factor similarity calculation formula is:

[0024]

[0025] Here, max(m, n) means taking the larger value between m and n as the value of max(m, n).

[0026] Specifically, the method of using constraint rule decision tree to conditionally constrain each key part and functional link in the approximate process includes: the constraint rule decision tree method is a tool for decision support and conditional control; each node of the decision tree represents a decision point and condition, each edge represents the result of the condition, and the leaf node represents the final decision result and action step; the steps of using the constraint rule decision tree method to conditionally constrain each key part and functional link in the approximate process are: defining key decision points, establishing a decision tree, applying constraint rules and executing decisions; optimizing the process through the main node information of the decision tree after constraints, obtaining the optimal process for the distribution design of the target seabed data cabin, and providing guidance for the engineering construction of the seabed data cabin.

[0027] The beneficial effects of the present invention are:

[0028] The present invention provides a method for designing the distribution of seabed data cubes based on knowledge graphs. By constructing a knowledge graph system framework for the distribution design of seabed data cubes, the relationship between various information of typical seabed data cubes in history can be integrated. By fusing different knowledge graphs, redundant information can be merged to improve the reusability of knowledge objects, reduce information contradictions between different knowledge graphs, and reduce the system fault tolerance, so that the fused knowledge graph is more versatile in the seabed data cube project. The seabed data cube obtained by the optimal process of the seabed data cube distribution design of the method comprehensively considers various factors and the relationship between different elements, and will be more environmentally friendly and safe. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 :A flow chart of a method for designing the distribution of seabed data cabins based on a knowledge graph of the present invention.

[0030] Figure 2 : A schematic diagram of creating a knowledge graph of an undersea data cabin according to the present invention.

[0031] Figure 3 : A flow chart of a decision tree of constraint rules for the distribution design of seabed data cabins of the present invention. DETAILED DESCRIPTION

[0032] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings; it should be understood that the specific embodiments given here are only used to illustrate and explain the present invention and cannot be used to limit the present invention.

[0033] It should be noted that many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention may also have other implementations and variations thereof. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0034] like Figure 1As shown, a flow chart of a method for designing the distribution of seabed data shelters based on a knowledge graph according to an embodiment of the present invention comprises: step S100, using an object-oriented method to divide the knowledge graph system of the distribution design of seabed data shelters into a design element knowledge graph, a layout parameter knowledge graph and a layout effect knowledge graph; the object-oriented method refers to the establishment of a knowledge graph that focuses on universality, classifies the cumbersome process flows into objects according to their commonalities, combines information and the behavioral source generated by the information as an object mapping set, and forms a behavioral, functional, modular and reusable universal structural system; the design element knowledge graph is constructed through knowledge triples<Component,Component_feature,Feature> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Feature is the tail entity, which indicates the feature information of the design element; Component_feature is the mapping relationship between the head entity and the tail entity; the layout parameter knowledge graph is represented by triples<Component,Component_parameter,Parameter> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Parameter is the tail entity, which indicates the specific layout parameter information; Component_parameter is the mapping relationship between the head entity and the tail entity, which is the specific layout method, which is determined by the construction plan and the design calculation report; the layout effect knowledge graph refers to the quality control of the knowledge graph, and the layout effect knowledge graph is constructed through triples<Item,Item_re quirement,Requirement> Indicated, Item is the head entity, which consists of the quality control item code Item-ID and the quality control item name Item-Name; Requirement is the tail entity, which indicates the quality control item threshold; Item_requirement is the mapping relationship between the head entity and the tail entity, which is determined by the data cabin maximum working temperature requirement table, material performance table, equipment manual, and relevant industry technical standards and specifications; Step S200, using the graph convolutional neural network GCN entity alignment method to fuse the element knowledge graph, layout parameter knowledge graph and layout effect knowledge graph; Step S300, after the knowledge graph is fused through step S100 and step S200, an upper-level construction knowledge ontology based on the knowledge graph is formed;Subsequently, the seabed data cube data source is used to perform scheme preview and design parameter intelligent decision-making on the upper ontology of the fused knowledge graph for the design data information involved in the specific project; finally, the seabed data cubes of specific historical projects are used to distribute the design knowledge graph, and the data sources of the seabed data cubes of each specific project are combined to perform scheme preview and design parameter intelligent decision-making on the upper ontology of the fused knowledge graph for the design data information involved in the specific project, and generate an instantiation of the first part of the upper ontology of the construction knowledge, and generate a knowledge graph of the seabed data cubes of specific historical projects. Usually, the data sources of the seabed data cubes of different projects correspond to different design parameter intelligent decision-making schemes; step S400, the seabed data cubes are composed of the data sources of the seabed data cubes of specific historical projects. The data knowledge base of the seabed data cube is composed of the seabed data cube knowledge graph of the historical specific project created by step S100, step S200 and step S300 to form the graph knowledge base of the seabed data cube; step S500, when constructing the knowledge graph of the target seabed data cube, firstly, the design factor similarity calculation method is used to find the seabed data cube of the historical specific project similar to the target seabed data cube in the data knowledge base; then, the general process of the distribution design of the target seabed data cube is obtained through the knowledge graph in the graph knowledge base corresponding to the seabed data cube of the similar historical specific project; finally, the constraint rule decision tree method is used to conditionally constrain each key part and functional link in the general process, so as to realize the intelligent planning of the distribution design of the target seabed data cube. ;

[0035] Specifically, the method first divides the knowledge graph system of historical typical seabed data cabin distribution design into design element knowledge graph, layout parameter knowledge graph and layout effect knowledge graph through object-oriented method; then, the three established knowledge graphs are integrated into one knowledge graph, and the scheme preview and design parameter intelligent decision-making are carried out through the data source to instantiate the integrated knowledge graph; when it is necessary to carry out distribution design of the target seabed data cabin, the design factor similarity calculation method is first used to find the seabed data cabin of the historical specific project similar to the target seabed data cabin in the data knowledge base; then, the approximate process of the target seabed data cabin distribution design is obtained through the knowledge graph in the graph knowledge base corresponding to the seabed data cabin of the similar historical specific project; finally, the constraint rule decision tree method is used to conditionally constrain each key part and functional link in the approximate process, so as to realize the intelligent planning of the target seabed data cabin distribution design.

[0036] Step S100, using an object-oriented method to divide the knowledge graph system of the seabed data shelter distribution design into a design element knowledge graph, a layout parameter knowledge graph, and a layout effect knowledge graph; the object-oriented method means that the established knowledge graph focuses on universality, classifies the cumbersome process flows into objects according to their commonalities, combines the information and the behavioral source of the information as an object mapping set, and forms a behavioral, functional, modular, and reusable universal structural system; the design element knowledge graph is constructed through knowledge triples<Component,Component_feature,Featur e> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Feature is the tail entity, which indicates the feature information of the design element; Component_feature is the mapping relationship between the head entity and the tail entity; the layout parameter knowledge graph is represented by triples<Component,Component_parameter,Parameter> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Parameter is the tail entity, which indicates the specific layout parameter information; Component_parameter is the mapping relationship between the head entity and the tail entity, which is the specific layout method, which is determined by the construction plan and the design calculation report; the layout effect knowledge graph refers to the quality control of the knowledge graph, and the layout effect knowledge graph is obtained through triples<Item,Item_requirement,Requirement> It is represented by Item, which is the head entity, and the head entity consists of the quality control item code Item-ID and the quality control item name Item-Name; Requirement is the tail entity, which indicates the quality control item threshold; Item_requirement is the mapping relationship between the head entity and the tail entity, which is determined by the data cabin maximum operating temperature requirement table, material performance table, equipment manual, and relevant industry technical standards and specifications.

[0037] In the above embodiment, specifically, the knowledge graph is usually structured in triples.<h,r,e> The knowledge information is described, where h and e are nodes in the knowledge graph, and r is the relationship between nodes in the knowledge graph; the creation of the construction knowledge graph includes extracting the construction knowledge information from the unstructured text, representing the construction information in a structured form, and then fusing the interrelated construction information to generate a visual construction knowledge graph; the design knowledge graph system refers to the representation of design knowledge.

[0038] In the above embodiments, preferably, the number of data cabins and the design methods are different, and the extracted knowledge is different; for example: a single-layer submarine data cabin with a gravel foundation trench as the supporting cabin arranged horizontally, a single-layer submarine data cabin with a pile-column cap beam as the supporting cabin arranged horizontally, and a double-layer submarine data cabin with a gravel foundation trench as the supporting cabin arranged horizontally (the lower cabin is supported by the gravel foundation trench, and the upper cabin is supported by the steel frame), and a double-layer submarine data cabin with a pile-column cap beam as the supporting cabin arranged horizontally (the lower cabin is supported by the gravel foundation trench, and the upper cabin is supported by the steel frame), etc., when extracting relevant design knowledge, the extracted knowledge nodes will be different.

[0039] In the above embodiment, preferably, Component1 is used to represent the second-layer data cabin, Component2 represents the steel frame, and Component3 represents the concrete corridor; Feature1 is used to represent that the element has a hanging feature; Component2_feature1 is used to represent the relationship between the second-layer data cabin and the steel frame, which may specifically include the fixing method, the bearing capacity requirements of the steel frame, etc.; Component 3_feature1 is used to represent the relationship between the second-layer data cabin and the concrete corridor, which may specifically include the corridor height requirements, etc.

[0040] In the above embodiment, preferably, Component is used to represent the data cabin, Parameter represents the time interval and layer distance of the lateral arrangement of the cabin, and Component_parameter is the temperature row numerical simulation calculation model.

[0041] In the above embodiment, preferably, under the Parameter submarine data cabin horizontal two-layer arrangement scheme, Item indicates that the cabin is arranged horizontally with a time interval of 0.8m, a layer spacing of 1.2m, and a bracket cantilever length of 5m. Requirement is the specific threshold of the item in It em under the relevant mapping relationship Item_requirement such as the data cabin working temperature requirement and the connection key deformation requirement under Item. The cabin is affected by water flow. Under normal load operation, the actual maintained temperature does not exceed 35°C, and the flexible deformation between the cabin and the corridor does not exceed 50mm.

[0042] Step S200: Using the graph convolutional neural network (GCN) entity alignment method to fuse the element knowledge graph, the layout parameter knowledge graph, and the layout effect knowledge graph.

[0043] In the above embodiment, specifically, the formula for performing knowledge fusion is:

[0044]

[0045] Among them, s is the activation function, P is an n×n adjacency matrix, representing the association information between two knowledge graphs, and n is the number of feature nodes; I is the n×n identity matrix; H input is the input data driven by the encoding in a knowledge graph, which is the vertex feature matrix; for The diagonal node degree matrix, W is the weight matrix, H out put Encodes the associated output data matrix in another knowledge graph.

[0046] In the above embodiment, specifically, the graph convolutional neural network GCN entity alignment method uses GCN to learn and embed graph structure data, thereby realizing entity matching in different knowledge graphs; entity alignment is a key task in knowledge graph fusion, and its goal is to identify and match different nodes representing the same real-world entities in different knowledge graphs; when using the graph convolutional neural network GCN entity alignment method to fuse the element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph, the element knowledge graph and the layout parameter knowledge graph are first fused through the knowledge fusion formula, and then the fused knowledge graph entities and the entities in the layout effect knowledge graph are fused, and finally the element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph are fused into a knowledge graph with the main entity characteristics of the three.

[0047] Step S300, after the knowledge graph is integrated through step S100 and step S200, an upper ontology of construction knowledge based on the knowledge graph is formed; then, the submarine data cabin data source is used to perform scheme preview and design parameter intelligent decision-making on the upper ontology of the integrated knowledge graph, and the design data information involved in the specific project; finally, the submarine data cabins of specific historical projects are used to distribute the design knowledge graph, and the data sources of the submarine data cabins of each specific project are combined to perform scheme preview and design parameter intelligent decision-making on the upper ontology of the integrated knowledge graph, and a method for processing the design data information involved in the specific project is generated, to generate an instantiation of the first part of the upper ontology of construction knowledge, and to generate a knowledge graph of the submarine data cabins of specific historical projects. Usually, the data sources of the submarine data cabins of different projects correspond to different design parameter intelligent decision-making schemes.

[0048] In the above embodiment, specifically, the purpose of the scenario preview is to predict and optimize the performance of the knowledge graph in various scenarios by simulating and evaluating different ontology design schemes before actual application; the scenario preview steps of the data source on the knowledge graph include: data source analysis, ontology mapping, data integration and scenario preview; the purpose of the intelligent decision-making of design parameters is to more efficiently and accurately manage, analyze and utilize large amounts of data, so as to support various intelligent applications and decision-making processes; the steps of the intelligent decision-making of design parameters include: determining design target data information, parameter selection, determination, implementation and testing of intelligent decision-making methods.

[0049] In the above embodiments, preferably, combining the data source of the seabed data cabins of each specific project to perform scheme preview and intelligent decision-making on design parameters on the upper-level ontology of the fused knowledge graph can remove unimportant information; for example: in the single-layer seabed data cabin scheme with a gravel foundation trench supporting the cabin arranged horizontally, the influencing factors related to the supporting steel frame are filtered, and more attention is paid to the influence of aspects such as foundation bearing capacity and uneven deformation.

[0050] Step S400, a data knowledge base of the seabed data cube is formed through the data sources of the seabed data cubes of specific historical projects, and a graph knowledge base of the seabed data cube is formed through the seabed data cube knowledge graphs of specific historical projects created through steps S100, S200 and S300.

[0051] In the above embodiment, specifically, the data information corresponding to the main nodes of the seabed data cube knowledge graph of the specific historical project is extracted through the main nodes of the obtained knowledge graph of the seabed data cube of the specific historical project to form a data knowledge base of the seabed data cube; at the same time, the correspondence between the knowledge in the data knowledge base and the knowledge graph in the graph knowledge base is established.

[0052] Step S500, when constructing a knowledge graph for the target seabed data cabin, first, a design factor similarity calculation method is used to find seabed data cabins of historical specific projects similar to the target seabed data cabin in the data knowledge base; then, an approximate process of the distribution design of the target seabed data cabin is obtained through the knowledge graph in the graph knowledge base corresponding to the seabed data cabins of similar historical specific projects; finally, a constraint rule decision tree method is used to conditionally constrain each key part and functional link in the approximate process, so as to realize intelligent planning of the distribution design of the target seabed data cabin.

[0053] In the above embodiment, specifically, the edit distance algorithm of similarity calculation is adopted, the data S of the specific target seabed data cabin and the data knowledge base corresponding to the typical engineering case scheme of the seabed data cabin in history are traversed and compared as the condition data T, and the similarity analysis is performed on the description text sentences of the two, and the one with the highest similarity and several typical schemes with high similarity are selected as the preliminary design scheme; the edit distance algorithm refers to the minimum number of editing operations required to change from the target data S to the condition data T, which is represented by the symbol LD; the steps of calculating the similarity are:

[0054] Based on the distribution of the target seabed data cabin, a knowledge graph is designed and its data source is extracted. m key indicator parameters are selected to construct a 1×m target data matrix. m key indicator parameters are selected to construct a 1×m target data matrix S = (s1, s2, …s m ); Taking the typical design schemes in history as dependent variables, extract the knowledge of the upper ontology of the submarine data cabin distribution design knowledge data source, select n key indicator parameters as independent variables, and construct a 1×n conditional data matrix T=(t1, t2, …t n ); construct a (m+1)×(n+1) order matching relationship matrix D between S and T:

[0055] D (m+1)×(n+1) ={d i,j}(0≤i≤m,0≤j≤n,)

[0056] Define the element d in the matrix D according to the following relationship i，j :

[0057]

[0058] Among them, a ij To edit the cost, express it according to the following rules:

[0059]

[0060] After the matrix D is constructed, the lower right corner element d mn That is, the edit distance LD between the target data matrix S and the conditional data matrix T, that is, LD = d mn ; At this time, the factor similarity calculation formula is:

[0061]

[0062] Here, max(m, n) means taking the larger value between m and n as the value of max(m, n).

[0063] In the above embodiment, specifically, the constraint rule decision tree method is a tool for decision support and conditional control; each node of the decision tree represents a decision point and condition, each edge represents the result of the condition, and the leaf node represents the final decision result and action step; the steps of using the constraint rule decision tree method to conditionally constrain each key part and functional link in the approximate process are: defining key decision points, establishing a decision tree, applying constraint rules and executing decisions; optimizing the process through the main node information of the decision tree after constraints, obtaining the optimal process for the distribution design of the target seabed data cabin, and providing guidance for the engineering construction of the seabed data cabin.

[0064] like Figure 2 As shown: a schematic diagram of creating a knowledge graph of a submarine data shelter of the present invention; in the schematic diagram, the creation of the knowledge graph of the submarine data shelter is divided into three parts; the first part is the upper layer ontology of the knowledge of the submarine data shelter distribution design; the second part is the data source; the third part is the creation of the knowledge graph; wherein, the upper layer ontology of the knowledge of the submarine data shelter distribution design includes: the extraction of the knowledge concept of design elements, the extraction of the knowledge concept of layout parameters and the extraction of the knowledge concept of quality control requirements; which is equivalent to the knowledge graph of design elements, the knowledge graph of layout parameters and the knowledge graph of layout effects; wherein, the process of extracting the knowledge concept of design elements is the extraction of the concept attributes of design elements, the extraction of the characteristic information of design elements and the extraction of the mapping relationship of design elements; the knowledge concept of layout parameters The extraction includes: layout parameter concept attribute extraction, layout parameter key information extraction and layout parameter mapping relationship extraction; quality control requirement knowledge concept extraction includes: quality evaluation index concept attribute extraction, technical standards and specifications knowledge extraction and key control item threshold mapping relationship extraction; then, the data source of the seabed data cabin is obtained, including construction geology, construction sea conditions, cabin structure performance and implementation technical standards, and the knowledge graph after the fusion of the seabed data cabin distribution design knowledge layer ontology is subjected to instance extraction, concept instance, relationship extraction, instance attribute extraction, design entity graph mapping, design knowledge fusion, ternary node and edge construction through the data source, and finally the seabed data cabin distribution design knowledge graph of the specific project is generated.

[0065] like Figure 3As shown: is a flow chart of a decision tree of constraint rules for the distribution design of seabed data cabins of the present invention; the flow chart mainly includes 5 decision points of yes / no, which are used to constrain the distribution process of seabed data cabins; the first decision point is to constrain whether the seabed data cabin to be built is super-large, if it is super-large, the step of multi-layer arrangement of the cabin is entered; if it is not super-large, the step of single-layer arrangement of the cabin is entered; the second decision point is to constrain whether the deflection of the steel frame cantilever is too large, if the deflection of the steel frame cantilever is too large, the data cabin is arranged longitudinally; if the cantilever deflection is not large, the data cabin is arranged horizontally; the third decision point is to constrain whether the bearing capacity of the seabed is sufficient, if the bearing capacity is sufficient, the judgment of anti-overturning capacity is entered; if the bearing capacity is insufficient, the cabin can only be arranged in a single layer; the fourth decision point is to resist overturning The constraint of whether the capacity is met is that if the anti-overturning capacity is sufficient, the spacing and layer spacing of the arranged data cabins are optimized; if the anti-overturning capacity is insufficient, the cabins can only be arranged in a single layer; the last is the constraint of the heat dissipation effect of the arranged cabins; if the heat dissipation effect meets the requirements, the design plan of the submarine data cabin is generated, and a series of parameter configurations of the submarine data cabin are generated; if the heat dissipation effect does not meet the requirements, the spacing and layer spacing of the cabins are continued to be optimized until the requirements are met; a series of parameter configurations generated by the flowchart are conditionally constrained for each key part and functional link in the approximate process, and the process is optimized through the constraint conditions to obtain the best process for the distribution design of the submarine data cabins to be built, and the submarine data cabins are constructed according to the obtained best process, so that the constructed submarine data cabins are safer and more environmentally friendly.

[0066] It should be understood that the above-mentioned embodiments are one or more embodiments of the present invention, and there are many other embodiments and variations thereof based on the present invention; the variations and modifications made by ordinary technicians in this industry through the present invention without making pioneering innovations all fall within the scope of protection of the present invention.

Claims

1. A method for designing the distribution of submarine data shelters based on knowledge graph, characterized in that: The method comprises: 1) The object-oriented method is used to divide the knowledge graph system of the seabed data shelter distribution design into the design element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph; the object-oriented method means that the established knowledge graph focuses on universality, classifies the cumbersome process flow into objects according to their commonalities, combines the information and the behavior source generated by the information as an object mapping set, and forms a behavioral, functional, modular and reusable universal structure system; the design element knowledge graph is constructed through knowledge triples<Component,Component_feature,Feature> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Feature is the tail entity, which indicates the feature information of the design element; Component_feature is the mapping relationship between the head entity and the tail entity; the layout parameter knowledge graph is represented by triples<Component,Component_parameter,Parameter> Indicates that Component is the head entity, which consists of the design element code Component-ID, the design element name Component-Name and the design element basic information description Component-Info; Parameter is the tail entity, which indicates the specific layout parameter information; Component_parameter is the mapping relationship between the head entity and the tail entity, which is the specific layout method, which is determined by the construction plan and the design calculation report; the layout effect knowledge graph refers to the quality control of the knowledge graph, and the layout effect knowledge graph is constructed through triples<Item,Item_requirement,Requirement> Item is the head entity, which consists of the quality control item code Item-ID and the quality control item name Item-Name; Requirement is the tail entity, which indicates the quality control item threshold; Item_requirement is the mapping relationship between the head entity and the tail entity, which is determined by the data cabin maximum operating temperature requirement table, material performance table, equipment manual, and relevant industry technical standards and specifications; 2) The graph convolutional neural network (GCN) entity alignment method is used to fuse the element knowledge graph, layout parameter knowledge graph, and layout effect knowledge graph. The formula for knowledge fusion is: Among them, s is the activation function, P is an n×n adjacency matrix, representing the association information between two knowledge graphs, and n is the number of feature nodes; I is the n×n identity matrix; H input is the input data driven by the encoding in a knowledge graph, which is the vertex feature matrix; for The diagonal node degree matrix, W is the weight matrix, H output Encode the associated output data matrix in another knowledge graph; 3) After the knowledge graph is integrated through 1) and 2), an upper ontology of construction knowledge based on the knowledge graph is formed; then, the upper ontology of the integrated knowledge graph is used to perform scheme preview and design parameter intelligent decision-making on the design data information involved in the specific project; finally, the upper ontology of the integrated knowledge graph is processed by the method of distributing the design knowledge graph of the submarine data cabins of specific historical projects and combining the data sources of the submarine data cabins of each specific project to perform scheme preview and design parameter intelligent decision-making on the upper ontology of the integrated knowledge graph, so as to generate an instantiation of the upper ontology of the first part of the construction knowledge, and generate the knowledge graph of the submarine data cabins of specific historical projects. Usually, the data sources of the submarine data cabins of different projects correspond to different intelligent decision-making schemes for design parameters; 4) The data sources of the submarine data cubes of specific historical projects are used to form a data knowledge base of the submarine data cubes, and the knowledge graphs of the submarine data cubes of specific historical projects created through 1), 2) and 3) are used to form a graph knowledge base of the submarine data cubes; 5) When constructing the knowledge graph for the target seabed data cabin, first, the design factor similarity calculation method is used to find the seabed data cabins of historical specific projects that are similar to the target seabed data cabin in the data knowledge base; then, the approximate process of the distribution design of the target seabed data cabin is obtained through the knowledge graph in the graph knowledge base corresponding to the seabed data cabins of similar historical specific projects; finally, the constraint rule decision tree method is used to conditionally constrain each key part and functional link in the approximate process, so as to realize the intelligent planning of the distribution design of the target seabed data cabin.

2. The method for designing the distribution of seabed data shelters based on knowledge graph according to claim 1 is characterized in that: The object-oriented method is used to design the knowledge graph system for distributing the seabed data cabins, including: the knowledge graph is usually structured as a triple<h,r,e> The knowledge information is described, where h and e are nodes in the knowledge graph, and r is the relationship between nodes in the knowledge graph; the creation of the construction knowledge graph includes extracting the construction knowledge information from the unstructured text, representing the construction information in a structured form, and then fusing the interrelated construction information to generate a visual construction knowledge graph; the design knowledge graph system refers to the representation of design knowledge.

3. The method for designing the distribution of seabed data shelters based on knowledge graph according to claim 1 is characterized in that: The graph convolutional neural network (GCN) entity alignment method includes: the method uses GCN to learn and embed graph structure data, thereby realizing entity matching in different knowledge graphs; entity alignment is a key task in knowledge graph fusion, and its goal is to identify and match different nodes representing the same real-world entities in different knowledge graphs; when using the graph convolutional neural network (GCN) entity alignment method to fuse the element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph, the element knowledge graph and the layout parameter knowledge graph are first fused through the knowledge fusion formula, and then the fused knowledge graph entities and entities in the layout effect knowledge graph are fused, and finally the element knowledge graph, the layout parameter knowledge graph and the layout effect knowledge graph are fused into a knowledge graph with the main entity features of the three.

4. The method for designing the distribution of seabed data shelters based on knowledge graph according to claim 1 is characterized in that: The design data information involved in the specific project of using the seabed data cabin data source to conduct solution preview and design parameter intelligent decision-making on the upper ontology of the fused knowledge graph includes: the purpose of the solution preview is to predict and optimize the performance of the knowledge graph in various scenarios by simulating and evaluating different ontology design schemes before actual application; the solution preview steps of the knowledge graph by the data source include: data source analysis, ontology mapping, data integration and solution preview; the purpose of the intelligent decision-making of design parameters is to manage, analyze and utilize a large amount of data more efficiently and accurately, so as to support various intelligent applications and decision-making processes; the steps of the intelligent decision-making of design parameters include: determining the design target data information, parameter selection, determination, implementation and testing of intelligent decision-making methods.

5. The method for designing the distribution of seabed data shelters based on knowledge graph according to claim 1 is characterized in that: The design factor similarity calculation method comprises: using an edit distance algorithm for similarity calculation, traversing and comparing the data S of a specific target seabed data cabin with a data knowledge base corresponding to a typical engineering case solution of a seabed data cabin in history as conditional data T, performing similarity analysis on the description text sentences of the two, and selecting a scheme including the one with the highest similarity and several typical schemes with high similarity as preliminary design schemes; the edit distance algorithm refers to the minimum number of editing operations required to change from the target data S to the conditional data T, represented by the symbol LD; the steps of calculating the similarity are: Based on the target seabed data cabin distribution design knowledge graph and extract its data source, select m key indicator parameters, and construct a 1×m target data matrix. Select m key indicator parameters and construct a 1×m target data matrix S = (s1, s2, ...s m ); Taking the typical design schemes in history as dependent variables, extract the knowledge of the upper ontology of the submarine data cabin distribution design knowledge data source, select n key indicator parameters as independent variables, and construct a 1×n conditional data matrix T=(t1, t2, …t n ); construct a (m+1)×(n+1) order matching relationship matrix D between S and T: D (m+1)×(n+1) ={d i,j }(0≤i≤m,0≤j≤n,) Define the element d in the matrix D according to the following relationship i，j : Among them, a ij To edit the cost, express it according to the following rules: After the matrix D is constructed, the lower right corner element d mn That is, the edit distance LD between the target data matrix S and the conditional data matrix T, that is, LD = d mn ; At this time, the factor similarity calculation formula is: Here, max(m, n) means taking the larger value between m and n as the value of max(m, n).

6. The method for designing the distribution of seabed data shelters based on knowledge graph according to claim 1, characterized in that: The method of using constraint rule decision tree to conditionally constrain each key part and functional link in the approximate process includes: the constraint rule decision tree method is a tool for decision support and condition control; each node of the decision tree represents a decision point and a condition, each edge represents the result of the condition, and the leaf node represents the final decision result and action step; the steps of using the constraint rule decision tree method to conditionally constrain each key part and functional link in the approximate process are: defining key decision points, establishing a decision tree, applying constraint rules and executing decisions; optimizing the process through the main node information of the decision tree after constraints, obtaining the optimal process of the target seabed data cabin distribution design, and providing guidance for the engineering construction of the seabed data cabin.

Images