Hybrid electric vehicle transmission system knowledge graph construction method, reasoning method and rapid design system

Abstract

The application discloses a kind of hybrid electric vehicle transmission system knowledge graph construction method, on the basis of establishing the knowledge base and instance base of hybrid electric vehicle transmission system, establish hybrid electric vehicle transmission system knowledge ontology model, in combination with the characteristics of hybrid electric vehicle transmission system, hybrid electric vehicle transmission system knowledge ontology model is set to include transmission scheme ontology, component structure ontology, parameter calculation model ontology and design process knowledge relationship set and then through entity recognition, relationship extraction and entity alignment knowledge processing, new knowledge can be accurately and efficiently extracted from big data, which is beneficial to knowledge mining and knowledge diffusion, and the application range of knowledge graph is improved from data retrieval and qualitative decision to comprehensive decision, thereby effectively solving complex problems in manufacturing scenarios. The application also discloses a principle block diagram of a hybrid electric vehicle transmission system knowledge graph reasoning method and a hybrid electric vehicle transmission system rapid design system.

Description

Technical Field

[0001] The present invention belongs to the technical field of transmission system design, and specifically provides a knowledge graph construction method, reasoning method and rapid design system for a hybrid vehicle transmission system. Background Art

[0002] The emergence of hybrid vehicles (HEVs) provides a balanced solution for both vehicle power performance and energy conservation and emission reduction. The design of HEVs in terms of transmission selection, structure, and parameter matching directly impacts vehicle power, range, fuel economy, and compactness.

[0003] The design process involves numerous constraints such as design standards, specifications, and process procedures. There are multiple structural options, selection issues such as parts selection, and complex process calculations and parameter matching, making it difficult to develop a fully intelligent design system. Consequently, designers still expend considerable time and effort completing a complete traditional system design. Furthermore, due to the varying performance requirements across different vehicle models, or even within the same vehicle, transmission system design involves significant repetitive work, leading to long product development cycles and high design costs.

[0004] Knowledge graph technology, based on knowledge engineering, can effectively integrate unstructured textual knowledge with heterogeneous data from multiple sources, establishing an entity-relationship network that visually displays data connections in a graphical format, thereby effectively improving data integration quality and enhancing data connectivity. Incorporating new-generation artificial intelligence algorithms such as deep learning for knowledge extraction can accurately and efficiently extract new knowledge from big data, facilitating knowledge mining and diffusion. Using new-generation artificial knowledge algorithms such as neural networks and deep learning, the application scope of knowledge graphs has been expanded from data retrieval and qualitative decision-making to comprehensive decision-making, effectively solving complex problems in manufacturing scenarios. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a hybrid vehicle transmission system knowledge graph construction method, reasoning method and rapid design method. The constructed knowledge graph can convert hybrid vehicle transmission system design methods, existing design cases and related expert experience into structured knowledge; the reasoning method can query and manage the required relevant knowledge in the knowledge graph; the rapid design system can intelligently recommend feasible hybrid vehicle transmission system design schemes and related design parameters based on the requirements of the design scheme.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention first proposes a method for constructing a knowledge graph of a hybrid vehicle transmission system, comprising the following steps:

[0008] Step 1: Constructing a knowledge ontology model of the hybrid vehicle transmission system

[0009] 11) Acquire knowledge related to hybrid vehicle transmission system design and build a knowledge base; integrate existing product cases of hybrid vehicle transmission systems to form a case library;

[0010] 12) Using a bottom-up construction approach, a triple-based hybrid vehicle powertrain knowledge ontology model is established, wherein the hybrid vehicle powertrain knowledge ontology model includes a powertrain scheme ontology, a component structure ontology, a parameter calculation model ontology, and a design process knowledge relationship set;

[0011] The transmission scheme ontology includes an element entity El, a variable entity Var, and a relationship set {Relations:Keys}; the entity type and property entity Pr of the element entity El, the entity types include resistive element, inertial element, capacitive element, potential source, current source, converter, gyrator, "0" junction, and "1" junction; the variable entity Var includes four types: mechanical translation, mechanical rotation, and hydraulic pressure; each type of variable entity Var includes six variable type entities VTy: potential, current, momentum, displacement, power, and energy;

[0012] The component structure ontology includes three levels: structural parts Tr, components Co and parts Pa. The structural entities at each level include five types of sub-entities and corresponding values, namely: type Ty, attribute At, technical parameter TPa, structural parameter SPa and performance parameter PPa; among them, type Ty refers to the function; attribute At represents the basic information of the product; technical parameter TPa represents the parameters of the structural entity during the design and control process; structural parameter SPa represents the spatial structure parameters of the structural entity; performance parameter PPa represents the design target parameter or performance performance parameter of the structural entity;

[0013] The parameter calculation model ontology includes parameter index Ind and calculation model CM; the parameter index Ind corresponds to the relevant parameters in the transmission scheme ontology and the component structure ontology and is described in a unified manner; the calculation model CM is used to select an appropriate calculation method and obtain the required relevant parameters;

[0014] The design process knowledge relationship set is used to associate the transmission solution ontology, component structure ontology and parameter calculation model ontology:

[0015] HVTSO={Entity∪Relation}=∑T i =TSO+CSO+PCMO+DS_Re

[0016] Wherein, HVTSO represents a hybrid vehicle transmission system knowledge ontology model; Entity represents an entity; Relation represents a mapping relationship set; T i represents the i th triple; TSO represents a transmission scheme ontology; CSO represents a component structure ontology; PCMO represents a parameter calculation model ontology; DS_Re represents a design process knowledge relationship set;

[0017] Step two: constructing a knowledge graph of a hybrid vehicle transmission system

[0018] 21) Bi-LSTM-CRF named entity recognition: the named entity recognition method based on Bi-LSTM-CRF identifies the named entity in the integrated electric drive structure field from unstructured text, takes unstructured text data as a knowledge source, and determines the type of named entity by analyzing the characteristics of the hybrid vehicle transmission system;

[0019] 22) Relationship extraction: the relationship extraction method based on the dependency syntax structure determines the core verb in the sentence by using dependency analysis, extracts the triplets with the verb as the core relationship, and thus realizes relationship extraction; the statistical model-based Chinese word segmentation method is adopted, jieba word segmentation and LTP sequence labeling word segmentation are used, and Unigram model and hidden Markov model are combined for word segmentation;

[0020] 23) Entity alignment: the similarity between each entity and other entities is calculated through cosine similarity, one of the entity pairs with similarity exceeding a threshold value is selected as a named entity, and other similar entities are stored as an attribute of the entity; the relationship between the entities is replaced, and entity alignment and relationship fusion are realized.

[0021] The application also provides a hybrid vehicle transmission system knowledge graph reasoning method, comprising the following steps:

[0022] S1: a hybrid vehicle transmission system knowledge graph is constructed based on the method of claim 1; all semantic relationships represented by triples are converted into dense low-dimensional real value vectors, the head entity is projected into the relationship space to obtain a head entity projection vector, the tail entity is projected into the relationship space to obtain a tail entity projection vector, and a scoring function is defined by using the head entity projection vector and the tail entity projection vector;

[0023] S2: the parts, functions and features in the knowledge graph are directly connected by the relationship r, and are represented by vectors, the entities of the knowledge graph are defined by attribute labels based on the ontology structure, and the label similarity between the entity nodes is defined;

[0024] S3: In the vector space, the cosine similarity calculation method is used to calculate the similarity between the control nodes in the knowledge graph. The distribution coordinate values ​​of the distribution vector reference are plotted into the vector space, and the cosine value of their angle is obtained to determine the similarity of the vectors, thereby determining the similarity between entities.

[0025] S4: Decompose the user input requirements into search conditions and output goals, where the search conditions are decomposed into condition constraints and performance indicators according to their nature. The condition constraints include function, feature and parameter constraints;

[0026] S5: Use graph search to perform entity mapping of functions, features, and parameter constraints in the hybrid electric vehicle transmission system knowledge graph, and obtain the corresponding function constraint entity set ∑Fu Eni , feature constraint entity set ∑Fe Eni and parameter constraint entity set ∑Pm Eni , using a combination of label similarity and semantic similarity based on cosine similarity to find the set of candidate transmission schemes ∑Pa_En that meet the functions and characteristics i ;

[0027] S6: According to the parameter entity set ∑Pm_En in the conditional constraint i , for the transmission scheme set ∑Pa_En to be selected i The corresponding computing resource entity is calculated to select the transmission scheme set ∑Pa_En that meets the parameter requirements in the constraint conditions. j ∈∑Pa_En i ;

[0028] S7: Based on the transmission scheme set ∑Pa_En j Carry out component design. When the transmission scheme is determined, the relevant parameters of the transmission scheme are used as the constraints of component design. According to the process of condition decomposition, entity mapping, matching screening, and calculation, a set of components that meet the conditions is designed and integrated into the candidate transmission scheme.

[0029] S8: For candidate transmission schemes, calculate their size, performance and other evaluation parameters in combination with the computing resource entities corresponding to the entities, and select the optimal transmission system design scheme;

[0030] S9: According to the output requirements, search for an entity or entity set that meets the requirements and is closest to the optimal transmission system design solution in a breadth-first search manner.

[0031] Furthermore, in step S1, the projection matrix of the head entity projection vector obtained by projecting the head entity onto the relational space is:

[0032] M rh =w r wh T +I m×n

[0033] Among them, M rh Represents the projection matrix of the head entity; w r The norm vector of the hyperplane representing the mapping relationship; w h T represents the projection of h on w_r; I m×n represents the m×n identity matrix;

[0034] The projection matrix of the tail entity projection vector obtained by projecting the tail entity into the relational space is:

[0035] M rt =w t w t T +I m×n

[0036] Among them, M rt represents the projection matrix of the tail entity; w r The norm vector of the hyperplane representing the mapping relationship; w t T represents the projection of t on w_r;

[0037] The scoring function is defined as:

[0038]

[0039] Among them, f r (h, t) indicates that the scoring function is defined using the head entity projection vector and the tail entity projection vector; h represents the head entity; t represents the tail entity; and r represents the mapping relationship from the head entity to the tail entity.

[0040] Furthermore, in step S2, the label similarity between entity nodes based on the Jaccard similarity coefficient is defined as:

[0041]

[0042] Among them, A and B represent entity nodes respectively.

[0043] Furthermore, in step S3, the cosine similarity between entities A(x1, y1) and B(x2, y2) is defined as:

[0044]

[0045] Among them, cosθ represents cosine similarity; x1 and y1 represent the projection coordinates of entity A in the relational space; x2 and y2 represent the projection coordinates of entity B in the relational space; A i ,Bi Represent the components of vectors A and B respectively; n represents the dimensions of vectors A and B.

[0046] The present invention also proposes a hybrid vehicle transmission system rapid design system, which includes a data layer, a knowledge graph layer and an application layer;

[0047] The data layer includes a knowledge base and an example base of hybrid vehicle transmission systems;

[0048] The knowledge graph layer includes a hybrid vehicle transmission system ontology model constructed based on a knowledge base and an instance base of the hybrid vehicle transmission system, and a hybrid vehicle transmission system knowledge graph obtained by subjecting the hybrid vehicle transmission system ontology model to knowledge processing including entity recognition, relationship extraction, and entity alignment;

[0049] An inference engine is provided in the application layer, and the inference engine designs a hybrid vehicle transmission system according to design requirements using the hybrid vehicle transmission system knowledge graph inference method described in any one of claims 2 to 5; the method for the inference engine to design the hybrid vehicle transmission system is: after inputting the design requirements and parameters, perform instance retrieval; determine whether there is an instance in the instance library with similarity that meets the requirements: if so, use the instance as the design result; if not, use the hybrid vehicle transmission system knowledge graph inference method to develop new products, and save the obtained new products to the instance library to supplement the hybrid vehicle transmission system knowledge graph.

[0050] Furthermore, there are three ways to develop new products using the knowledge graph reasoning method for hybrid vehicle transmission systems: local design, step-by-step design, and automatic design.

[0051] Furthermore, the local design method is:

[0052] 11) Select the component to be designed and enter the design requirements and parameters;

[0053] 12) Retrieve design solutions using knowledge graph reasoning methods for hybrid vehicle transmission systems;

[0054] 13) Determine whether the design solution meets the requirements: if yes, go to step 14); if not, go to step 12);

[0055] 14) Under the new design of the component, check or redesign the relevant parameters of other components coupled with the component through the knowledge graph of the hybrid vehicle transmission system;

[0056] 15) Determine whether the design solutions of other components coupled with the component meet the requirements: if so, obtain the design solution; if not, execute step 12).

[0057] Further, the step-by-step design method is:

[0058] 21) selecting a design range and related design parameters;

[0059] 22) obtaining a design scheme by using a hybrid vehicle transmission system knowledge graph reasoning method;

[0060] 23) judging whether the design scheme meets the requirements: if yes, step 24) is executed; if no, step 22) is executed;

[0061] 24) sequentially designing all parts by using a local design method until all parts are designed;

[0062] 25) verifying and evaluating the design scheme: if it meets the requirements, the design scheme is output; if it does not meet the requirements, local modification is performed by using a local design method until it meets the requirements.

[0063] Further, the automatic design method is:

[0064] 31) selecting a design range and related design parameters;

[0065] 32) obtaining a design scheme by using a hybrid vehicle transmission system knowledge graph reasoning method, and completing the design of all parts;

[0066] 33) verifying and evaluating the design scheme:

[0067] if it meets the requirements, the design scheme is output;

[0068] if the design scheme as a whole does not meet the requirements, step 32) is executed;

[0069] if the design scheme is locally not in conformity with the requirements, local modification is performed by using a local design method until it meets the requirements.

[0070] The present application has the following advantages:

[0071] The hybrid vehicle transmission system knowledge graph construction method of the application, on the basis of constructing the knowledge base and the instance base of the hybrid vehicle transmission system, establishes a hybrid vehicle transmission system knowledge ontology model, combines the characteristics of the hybrid vehicle transmission system, sets the hybrid vehicle transmission system knowledge ontology model as including a transmission scheme ontology, a component structure ontology, a parameter calculation model ontology and a design process knowledge relationship set, wherein the transmission scheme ontology includes element entities El, variable entities Var and a relationship set, and the element entities El and the variable entities Var are classified and associated, the component structure ontology is divided into three levels of structural parts Tr, components Co and parts Pa, and the sub-entities of the structural entities of each level are defined, and the parameter calculation model ontology and the design process knowledge relationship set are combined, which can effectively integrate the knowledge of unstructured text and various types of multi-source heterogeneous data, establish an entity relationship network, and intuitively display the association between data in the form of a graph, thereby effectively improving the data integration quality and enhancing the interconnectivity between data to meet the design requirements of the hybrid vehicle transmission system, and then through knowledge processing of entity recognition, relationship extraction and entity alignment, new knowledge can be accurately and efficiently extracted from big data, which is beneficial to knowledge mining and knowledge diffusion, and the application range of the knowledge graph is improved from data retrieval and qualitative decision making to comprehensive decision making, thereby effectively solving complex problems in manufacturing scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0072] In order to make the purpose, technical scheme and beneficial effects of the application more clear, the application provides the following drawings for illustration:

[0073] Figure 1 is a structural diagram of the transmission scheme ontology model;

[0074] Figure 2 is a structural diagram of the component structure ontology model;

[0075] Figure 3 is a structural diagram of the parameter calculation model ontology model;

[0076] Figure 4 is a structural diagram of the knowledge ontology model of the hybrid vehicle transmission system constructed by the design process knowledge relationship set in combination with the transmission scheme ontology, the component structure ontology and the parameter calculation model ontology model;

[0077] Figure 5 is a principle block diagram of the hybrid vehicle transmission system knowledge graph construction method of the application;

[0078] Figure 6 is a principle block diagram of the hybrid vehicle transmission system knowledge graph reasoning method of the application;

[0079] Figure 7The framework diagram of the quick design system of the hybrid vehicle transmission system of the application;

[0080] Figure 8 The flowchart of the basic method for the reasoning machine to design the hybrid vehicle transmission system;

[0081] Figure 9 The flowchart of the local design;

[0082] Figure 10 The flowchart of the step-by-step design;

[0083] Figure 11 The flowchart of the automatic design. DETAILED DESCRIPTION

[0084] The application will be further described below in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the application and implement it. The embodiments are not intended to limit the application.

[0085] I. Method for constructing a knowledge graph of a hybrid vehicle transmission system

[0086] The method for constructing a knowledge graph of a hybrid vehicle transmission system of the embodiment includes the following steps:

[0087] Step 1: Construct a knowledge ontology model of a hybrid vehicle transmission system

[0088] 11) According to the design process of a hybrid vehicle transmission system, obtain the knowledge related to the design of a hybrid vehicle transmission system from industry standards, enterprise design specifications, design manuals, existing design schemes, and expert experience, etc., to construct a knowledge base; integrate existing product cases of a hybrid vehicle transmission system to form an example library; the knowledge base is used to construct a knowledge ontology model of a hybrid vehicle transmission system, and provide a construction standard for a knowledge graph. The knowledge ontology model of a hybrid vehicle transmission system adopts a bottom-up construction method, represents the knowledge nodes and relationships in the form of triples, and is developed and managed through Protégé.

[0089] 12) A bottom-up construction method is adopted to establish a triple-based knowledge ontology model of a hybrid vehicle transmission system. The knowledge ontology model of a hybrid vehicle transmission system of the embodiment adopts a modular structure, and designs ontologies for transmission schemes, component structures, and parameter calculation models, respectively. The relationships between the three parts of the ontology are added through relevant design methods and expert experience to integrate them into a general ontology. This method effectively avoids knowledge redundancy and loss, and can integrate all the knowledge required in the design process.

[0090] Specifically, a triple-based knowledge ontology model T i=(h_En,r,t_En), including the three sub-ontologies of Transmission Solution Ontology (TSO), Component Structure Ontology (CSO), and Parametric Calculation Model Ontology (PCMO), as well as the set of design process knowledge relations (DS_Re = ∑Ds_r(En)) that connect the three sub-ontologies. i ,En j )), where En i Represents a node entity, h_En represents the head entity; t_En represents the tail entity; r represents the mapping relationship from the head entity to the tail entity.

[0091] (1) Transmission scheme

[0092] The transmission scheme ontology is based on the bond graph theory. The transmission scheme is represented by a bond model and then abstracted into entities and relationships. Node entities are used to describe components and component attributes, and relationships are used to describe potential flows and variables, generating a bond graph model represented by a graph network. The transmission scheme ontology of this embodiment includes the component entity El, the variable entity Var, and the relationship set {Relations:Keys}, as shown in the following example. Figure 1 The entity types and property entities Pr of the component entity El include resistive elements, inertial elements, capacitive elements, potential sources, current sources, converters, gyrators, "0" junctions, and "1" junctions, as shown in Table 1. The component properties in Table 1 are the property entities Pr.

[0093] Table 1 Component entity set

[0094]

[0095] The variable entity Var includes four types: mechanical translation, mechanical rotation, and hydraulic pressure; each type of variable entity Var includes six variable type entities VTy: potential, flow, momentum, displacement, power, and energy, as shown in Table 2.

[0096] Table 2 Variable entity set and its type classification

[0097]

[0098] The relationship collection {Relations:Keys} represents the key type, including the following types:

[0099] Table 3 Key type relationship set

[0100]

[0101] Note: ITC (Impedance-type causality), ATC (Admittance-type causality)

[0102] (2) Component structure ontology

[0103] The component structure ontology describes all parts entities in a unified model framework. As shown in Figure 2 , the component structure ontology of the present embodiment includes three levels of structural entities, i.e. structure Tr, assembly Co and part Pa, and each level of structural entity includes five types of sub-entities and corresponding values, i.e. type Ty, attribute At, technical parameter TPa, structural parameter SPa and performance parameter PPa. The type Ty refers to the functional role, such as the workpiece properties or roles of energy source, transmission, etc. The attribute At represents the basic product information including model number, production time, etc. The technical parameter TPa represents the parameters of the structural entity in the design and control process, including motor power, maximum torque, transmission ratio, etc. The structural parameter SPa represents the spatial structure parameters including weight, size, rotation angle, etc. The performance parameter PPa represents the design target parameters or performance parameters of the structural entity, including displacement, maximum speed, endurance, etc.

[0104] (3) Parameter calculation model ontology

[0105] The parameter calculation model ontology abstracts the calculation process and method, and is used to describe the main content of the calculation model and establish the relationship between the design process and the part entity. As shown in Figure 3 , the parameter calculation model ontology of the present embodiment includes parameter index Ind and calculation model CM. The parameter index Ind corresponds to the related parameters in the transmission scheme ontology and the component structure ontology and is uniformly described. The calculation model CM includes calculation formula, method, tool, etc. and is used to select the appropriate calculation method through the entity and obtain the required related parameters.

[0106] (4) Design process knowledge relationship set

[0107] As shown in Figure 4 , the design process knowledge relationship set is used to associate the transmission scheme ontology, the component structure ontology and the parameter calculation model ontology:

[0108] HVTSO = {Entity U Relation} = ∑T i = TSO + CSO + PCMO + DS_Re

[0109] Among them, HVTSO represents the knowledge ontology model of hybrid electric vehicle transmission system; Entity represents entity; Relation represents mapping relationship set; T i represents the i-th triple; TSO represents the transmission scheme ontology; CSO represents the component structure ontology; PCMO represents the parameter calculation model ontology; DS_Re represents the design process knowledge relationship set;

[0110] Step 2: Build a knowledge graph of hybrid vehicle transmission systems

[0111] In order to realize the extraction and storage of complex product design knowledge and realize knowledge reuse, considering the heterogeneity of design knowledge and the reuse needs of designers, this embodiment combines the characteristics of product design to construct Figure 5 The following is a framework for constructing a knowledge graph for a hybrid vehicle powertrain system. Knowledge processing is performed on database data, primarily including entity recognition, relationship extraction, and entity alignment. The extracted entities and relationships are stored as triples in a Neo4j database as a knowledge source to support knowledge retrieval, reuse, and decision-making. The steps involved are as follows:

[0112] 21) Bi-LSTM-CRF named entity recognition: This method uses Bi-LSTM-CRF-based named entity recognition to identify named entities in the field of integrated electric drive structures from unstructured text. It uses unstructured text data as a knowledge source based on design manuals, standards, expert experience, mature product design knowledge, etc., and determines the type of named entities by analyzing the characteristics of hybrid vehicle transmission systems.

[0113] This method uses a recurrent neural network to identify each Chinese character, using a BIO encoding scheme for labeling, where B indicates the start of an entity, I indicates the middle of an entity, and O indicates that the element is outside the entity recognition range. This transforms entity recognition into a sequence labeling problem, and uses a character-based Bi-LSTM-CRF combination model for entity recognition.

[0114] 22) Relation Extraction: A relation extraction method based on dependency syntactic structure uses dependency analysis to determine the core verbs in a sentence and extract triples with verbs as the core relationship, thereby achieving relation extraction; a Chinese word segmentation method based on a statistical model is adopted, using Jieba word segmentation and LTP sequence tagging word segmentation, combined with the Unigram model and Hidden Markov Model (HMM) for word segmentation.

[0115] 23) Entity Alignment: Calculate the similarity between each entity and other entities using cosine similarity and set a threshold. Select one of the entity pairs whose similarity exceeds the threshold as a named entity, and then store the other similar entities as attributes of that entity. Replace the relationships between entities to achieve entity alignment and relationship fusion. To ensure the accuracy of all entities, manual verification and validation can be used to ensure the accuracy of the fused entities and entity relationships.

[0116] The hybrid vehicle transmission system knowledge graph construction method of this embodiment establishes a hybrid vehicle transmission system knowledge ontology model based on the construction of the hybrid vehicle transmission system knowledge base and instance base. In combination with the characteristics of the hybrid vehicle transmission system, the hybrid vehicle transmission system knowledge ontology model is set to include a transmission scheme ontology, a component structure ontology, a parameter calculation model ontology and a design process knowledge relationship set. Among them, the transmission scheme ontology component entity El, variable entity Var and relationship set are classified and associated with the component entity El and the variable entity Var. The component structure ontology is divided into three levels: structure part Tr, component Co and component Pa. At the same time, the structure entity of each level is By defining the sub-entities of the entity, and then combining the parameter calculation model ontology and the design process knowledge relationship set, it can effectively integrate the knowledge of unstructured text and various types of multi-source heterogeneous data, establish an entity relationship network, and intuitively display the relationship between data in the form of a graph, thereby effectively improving the quality of data integration and enhancing the connectivity between data to meet the design requirements of hybrid vehicle transmission systems; then, through knowledge processing such as entity recognition, relationship extraction and entity alignment, new knowledge can be accurately and efficiently extracted from big data, which is conducive to knowledge mining and knowledge diffusion, and the application scope of knowledge graphs is upgraded from data retrieval and qualitative decision-making to comprehensive decision-making, thereby effectively solving complex problems in manufacturing scenarios.

[0117] The specific implementation of the hybrid vehicle transmission system knowledge graph reasoning method of the present invention is described below in conjunction with the above-mentioned hybrid vehicle transmission system knowledge graph construction method of this embodiment.

[0118] 2. Knowledge Graph Reasoning Method for Hybrid Electric Vehicle Transmission System

[0119] In order to use the hybrid vehicle transmission system knowledge graph to perform data retrieval, qualitative decision-making and comprehensive decision-making to complete the rapid design of the hybrid vehicle transmission system, this embodiment proposes a hybrid vehicle transmission system knowledge graph reasoning method, the principle of which is as follows: Figure 6 This method matches the user's input requirements, filters out the part entity set that meets the requirements, then calculates the optimal part entity with the optimal parameters in the entity set, and finally outputs the entity required by the user based on the optimal part entity.

[0120] Specifically, the hybrid vehicle transmission system knowledge graph reasoning method of this embodiment includes the following steps:

[0121] S1: Based on the method of claim 1, a knowledge graph of hybrid vehicle transmission system is constructed; all triples T i =(h_En,r,t_En) is converted into a dense low-dimensional real-valued vector, the head entity is projected into the relational space to obtain the head entity projection vector, the tail entity is projected into the relational space to obtain the tail entity projection vector, and the head entity projection vector and the tail entity projection vector are used to define the scoring function.

[0122] Specifically, the projection matrix of the head entity projection vector obtained by projecting the head entity into the relational space is:

[0123] M rh =w r w h T +I m×n

[0124] Among them, M rh Represents the projection matrix of the head entity; w r The norm vector of the hyperplane representing the mapping relationship; w h T represents the projection of h on w_r; I m×n represents the m×n identity matrix;

[0125] The projection matrix of the tail entity projection vector obtained by projecting the tail entity into the relational space is:

[0126] M rt =w t w t T +I m×n

[0127] Among them, M rt represents the projection matrix of the tail entity; w r The norm vector of the hyperplane representing the mapping relationship; w t T represents the projection of t on w_r.

[0128] The scoring function is defined as:

[0129]

[0130] Among them, f r (h, t) indicates that the scoring function is defined using the head entity projection vector and the tail entity projection vector; h represents the head entity; t represents the tail entity; and r represents the mapping relationship from the head entity to the tail entity.

[0131] S2: Connect the parts, functions, and features in the knowledge graph directly with the relationship r, and define the attribute labels of the entities in the knowledge graph using the ontology structure as vectors, and define the label similarity between entity nodes. Specifically, the label similarity between entity nodes based on the Jaccard similarity coefficient is defined as:

[0132]

[0133] Among them, A and B represent entity nodes respectively.

[0134] S3: In the vector space, use the cosine similarity calculation method to calculate the similarity between the control nodes in the knowledge graph. Draw the distribution coordinate values ​​of the distribution vector reference into the vector space, and obtain the cosine value of their angle to determine the similarity of the vectors, thereby determining the similarity between the entities. In this embodiment, the cosine similarity of entities A(x1, y1) and B(x2, y2) is defined as:

[0135]

[0136] Among them, cosθ represents cosine similarity; x1 and y1 represent the projection coordinates of entity A in the relational space; x2 and y2 represent the projection coordinates of entity B in the relational space; A i ,B i Represent the components of vectors A and B respectively; n represents the dimensions of vectors A and B.

[0137] S4: Decompose the user input requirements into search conditions and output goals, where the search conditions are decomposed into condition constraints and performance indicators according to their nature. Condition constraints include function, feature and parameter constraints.

[0138] S5: Use graph search to map functions, features, and parameter constraints in the hybrid electric vehicle powertrain knowledge graph and obtain the corresponding function constraint entity set. Feature Constraint Entity Set and parameter constraint entity sets The label similarity and the semantic similarity based on cosine similarity are combined to find the set of candidate transmission schemes ∑Pa_En that meet the functions and characteristics. i ;

[0139] S6: According to the parameter entity set ∑Pm_En in the conditional constraint i , for the transmission scheme set ∑Pa_En to be selected i The corresponding computing resource entity is calculated to select the transmission scheme set ∑Pa_En that meets the parameter requirements in the constraint conditions. j ∈∑Pa_En i ;

[0140] S7: Based on the set of transmission schemes ∑Pa_En j Part design is performed, in the case of determining the transmission scheme, relevant parameters in the transmission scheme are taken as constraint conditions of part design, and a set of parts meeting the conditions is designed according to the processes of condition decomposition, entity mapping, matching and screening, and calculation, and is integrated in the candidate transmission scheme.

[0141] S8: For the candidate transmission scheme, combining the calculation resource entity corresponding to the entity, size, performance and other evaluation parameters are calculated, and the optimal transmission system design scheme is selected.

[0142] S9: According to the output requirement, an entity or entity set meeting the requirement closest to the optimal transmission system design scheme is searched in a breadth-first search manner.

[0143] The specific implementation of the quick design system of the hybrid electric vehicle transmission system of the present application will be described below in combination with the hybrid electric vehicle transmission system knowledge graph reasoning method described above.

[0144] III. Quick design system of hybrid electric vehicle transmission system

[0145] Based on the established hybrid electric vehicle transmission system knowledge graph and the corresponding intelligent reasoning method, a comprehensive application system for the design of the hybrid electric vehicle transmission system of new energy vehicles is developed, and the system architecture is as shown in Figure 7 Specifically, the quick design system of the hybrid electric vehicle transmission system of the present application includes a data layer, a knowledge graph layer and an application layer. The data layer includes a knowledge base and an instance base of the hybrid electric vehicle transmission system. The knowledge graph layer includes a hybrid electric vehicle transmission system ontology model constructed based on the knowledge base and the instance base of the hybrid electric vehicle transmission system, and a hybrid electric vehicle transmission system knowledge graph obtained by knowledge processing including entity recognition, relationship extraction and entity alignment from the hybrid electric vehicle transmission system ontology model. The inference machine is provided in the application layer, and the inference machine designs the hybrid electric vehicle transmission system according to the design requirements by using the hybrid electric vehicle transmission system knowledge graph reasoning method described above.

[0146] Based on the existing data and technology layers, users can use the system to perform data retrieval, qualitative decision-making, and comprehensive decision-making, enabling rapid design of hybrid vehicle powertrains. Based on the design requirements and inputs, the system derives the key constraints and relevant parameters of the hybrid vehicle powertrain. It first searches the case library for examples. If the case library contains examples that meet the required similarity, the product is delivered to the user. If no examples meet the required similarity or the user is dissatisfied with the existing examples, the system develops a new product based on the hybrid vehicle powertrain knowledge graph, saves the product to the case library, and supplements the hybrid vehicle powertrain knowledge graph.

[0147] Specifically, such as Figure 8 As shown, the method for the inference engine of this embodiment to design the hybrid vehicle transmission system is: after inputting the design requirements and parameters, perform instance retrieval; determine whether there is an instance in the instance library that meets the similarity requirements: if so, use the instance as the design result; if not, use the hybrid vehicle transmission system knowledge graph reasoning method to develop new products, and save the obtained new products to the instance library to supplement the hybrid vehicle transmission system knowledge graph.

[0148] Specifically, in this embodiment, the hybrid vehicle powertrain knowledge graph reasoning method is used to develop new products in three ways: partial design, step-by-step design, and automatic design. Users can use partial design to quickly modify existing products. Alternatively, they can use step-by-step design or automatic design to quickly design hybrid vehicle powertrains, following the design steps for new energy hybrid vehicles.

[0149] (1) Local design

[0150] like Figure 9 As shown, the local design method is:

[0151] 11) Select the component to be designed and enter the design requirements and parameters;

[0152] 12) Retrieve design solutions using knowledge graph reasoning methods for hybrid vehicle transmission systems;

[0153] 13) Determine whether the design solution meets the requirements: if yes, go to step 14); if not, go to step 12);

[0154] 14) Under the new design of the component, check or redesign the relevant parameters of other components coupled with the component through the knowledge graph of the hybrid vehicle transmission system;

[0155] 15) Determine whether the design solutions of other components coupled with the component meet the requirements: if so, obtain the design solution; if not, execute step 12).

[0156] Local design is the design of a certain component, including selection, parameter design, etc. Its steps are as follows: Figure 8 As shown in the figure, the user first identifies the design component and sets the relevant constraints and conditions. The system then searches the hybrid vehicle drivetrain knowledge graph for a suitable design solution based on the required conditions until the user's needs are met. Under the new design solution, the system checks or redesigns the relevant parameters of the designed component and its coupled components using the hybrid vehicle drivetrain knowledge graph. If the parameters do not meet the requirements, an alternative solution is selected until the user's needs are met. Taking the design of the motor and transmission connection solution as an example, the user enters the motor shaft diameter, the transmission input shaft diameter, and the transmission angle (the angle between the motor shaft and the transmission input shaft). The system then calculates implicit constraints such as power and torque based on the hybrid vehicle drivetrain knowledge graph and intelligently recommends a reasonable connection solution.

[0157] (2) Step-by-step design

[0158] like Figure 10 As shown, the step-by-step design method is:

[0159] 21) Select the design scope and related design parameters;

[0160] 22) Use the knowledge graph reasoning method to obtain the design scheme of hybrid vehicle transmission system;

[0161] 23) Determine whether the design scheme meets the requirements: if yes, go to step 24); if not, go to step 22);

[0162] 24) Use the local design method to design all parts in sequence until all parts are designed;

[0163] 25) Verify and evaluate the design plan: if it meets the requirements, output the design plan; if it does not meet the requirements, make local modifications using local design methods until it meets the requirements.

[0164] Step-by-step design follows the principles of transmission mechanism design, using a localized design approach to design all transmission structures and components. Designing a transmission system using a step-by-step design approach primarily involves the following steps. First, the user selects the design scope and relevant design parameters, and the system automatically provides the relevant design solutions and steps. Taking overall design as an example, the system generates a transmission system solution that meets the design requirements. After the user selects a transmission solution, all components are designed sequentially. Within each component design process, a localized design approach is used to design each component element in turn, generating a component that meets the requirements. After all components are designed, the transmission solution is verified and evaluated, and suggestions for possible modifications are provided to the user. If it meets the user's requirements, the solution is delivered. If not, local modifications are made until the user's requirements are met.

[0165] (3) Automatic design

[0166] like Figure 11 As shown, the automatic design method is:

[0167] 31) Select the design scope and related design parameters;

[0168] 32) Use the knowledge graph reasoning method of the hybrid electric vehicle transmission system to obtain the design scheme and complete the design of all components;

[0169] 33) Verify and evaluate the design plan:

[0170] If it meets the requirements, the design plan is output;

[0171] If the overall design does not meet the requirements, proceed to step 32);

[0172] If part of the design does not meet the requirements, it will be modified partially using local design methods until it meets the requirements.

[0173] The above embodiments are merely preferred embodiments for the purpose of fully illustrating the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are within the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

Claims

1. A method for constructing a knowledge graph of a hybrid vehicle transmission system, characterized by: The steps include: Step 1: Constructing a knowledge ontology model of the hybrid vehicle transmission system 11) Acquire knowledge related to hybrid vehicle transmission system design and build a knowledge base; integrate existing product cases of hybrid vehicle transmission systems to form a case library; 12) Using a bottom-up construction approach, a triple-based hybrid vehicle powertrain knowledge ontology model is established, wherein the hybrid vehicle powertrain knowledge ontology model includes a powertrain scheme ontology, a component structure ontology, a parameter calculation model ontology, and a design process knowledge relationship set; The transmission scheme ontology includes an element entity El, a variable entity Var, and a relationship set {Relations:Keys}; the entity type and property entity Pr of the element entity El, the entity types include resistive element, inertial element, capacitive element, potential source, current source, converter, gyrator, 0 junction, and 1 junction; the variable entity Var includes four types: mechanical translation, mechanical rotation, and hydraulic pressure; each type of variable entity Var includes six variable type entities vTy: potential, current, momentum, displacement, power, and energy; The component structure ontology includes three levels: structural parts Tr, components Co and parts Pa. The structural entities at each level include five types of sub-entities and corresponding values, namely: type Ty, attribute At, technical parameter TPa, structural parameter SPa and performance parameter PPa; among them, type Ty refers to the function; attribute At represents the basic information of the product; technical parameter TPa represents the parameters of the structural entity during the design and control process; structural parameter SPa represents the spatial structure parameters of the structural entity; performance parameter PPa represents the design target parameter or performance performance parameter of the structural entity; The parameter calculation model ontology includes parameter index Ind and calculation model CM; the parameter index Ind corresponds to the relevant parameters in the transmission scheme ontology and the component structure ontology and is described in a unified manner; the calculation model CM is used to select an appropriate calculation method and obtain the required relevant parameters; The design process knowledge relationship set is used to associate the transmission solution ontology, component structure ontology and parameter calculation model ontology: HVTSO={Entity∪Relation}=∑T i =TSO+CSO+PCMO+DS_Re Among them, HVTSO represents the knowledge ontology model of hybrid electric vehicle transmission system; Entity represents entity; Relation represents mapping relationship set; T i represents the i-th triple; TSO represents the transmission scheme ontology; CSO represents the component structure ontology; PCMO represents the parameter calculation model ontology; DS_Re represents the design process knowledge relationship set; Step 2: Build a knowledge graph of hybrid vehicle transmission systems 21) Bi-LSTM-CRF named entity recognition: This method uses Bi-LSTM-CRF to identify named entities in the field of integrated electric drive structures from unstructured text. Using unstructured text data as the knowledge source, it determines the type of named entities by analyzing the characteristics of hybrid vehicle drive systems. 22) Relation Extraction: A relation extraction method based on dependency syntactic structure uses dependency analysis to determine the core verbs in a sentence and extract triples with verbs as the core relationship, thereby achieving relation extraction. A Chinese word segmentation method based on statistical models is used, using Jieba word segmentation and LTP sequence tagging for word segmentation, combined with the Unigram model and Hidden Markov Model for word segmentation. 23) Entity Alignment: Calculate the similarity between each entity and other entities through cosine similarity, select one of the entity pairs whose similarity exceeds the threshold as a named entity, and then store the other similar entities as an attribute of the entity; replace the relationship between entities to achieve entity alignment and relationship fusion.

2. A knowledge graph reasoning method for a hybrid electric vehicle transmission system, characterized by: The steps include: S1: constructing a knowledge graph of a hybrid vehicle transmission system based on the method according to claim 1; converting the semantic relations represented by all triples into dense low-dimensional real-valued vectors, projecting the head entity into the relational space to obtain a head entity projection vector, projecting the tail entity into the relational space to obtain a tail entity projection vector, and defining a scoring function using the head entity projection vector and the tail entity projection vector; S2: directly connect the parts, functions and features in the knowledge graph with the relationship r, and define the attribute labels of the entities in the knowledge graph with the ontology structure using vector representation, and define the label similarity between entity nodes; S3: In the vector space, the cosine similarity calculation method is used to calculate the similarity between the control nodes in the knowledge graph. The distribution coordinate values ​​of the distribution vector reference are plotted into the vector space, and the cosine value of their angle is obtained to determine the similarity of the vectors, thereby determining the similarity between entities. S4: Decompose the user input requirements into search conditions and output goals, where the search conditions are decomposed into condition constraints and performance indicators according to their nature. The condition constraints include function, feature and parameter constraints; S5: Use graph search to map functions, features, and parameter constraints in the hybrid electric vehicle powertrain knowledge graph and obtain the corresponding function constraint entity set. Feature Constraint Entity Set and parameter constraint entity sets The label similarity and the semantic similarity based on cosine similarity are combined to find the set of candidate transmission schemes ∑Pa_En that meet the functions and characteristics. i ; S6: According to the parameter entity set ∑Pm_En in the conditional constraint i , for the transmission scheme set ∑Pa_En to be selected i The corresponding computing resource entity is calculated to select the transmission scheme set ∑Pa_En that meets the parameter requirements in the constraint conditions. j ∈∑Pa_En i ; S7: Based on the transmission scheme set ∑Pa_En j Carry out component design. When the transmission scheme is determined, the relevant parameters of the transmission scheme are used as the constraints of component design. According to the process of condition decomposition, entity mapping, matching screening, and calculation, a set of components that meet the conditions is designed and integrated into the candidate transmission scheme. S8: For the candidate transmission schemes, calculate their sizes and performance evaluation parameters in combination with the computing resource entities corresponding to the entities, and select the optimal transmission system design scheme; S9: According to the output requirements, search for an entity or entity set that meets the requirements and is closest to the optimal transmission system design solution in a breadth-first search manner.

3. The hybrid vehicle powertrain knowledge graph reasoning method according to claim 2, characterized in that: In step S1, the projection matrix of the head entity projection vector obtained by projecting the head entity onto the relational space is: M rh =w r w h T +I m×n Among them, M rh Represents the projection matrix of the head entity; w r The norm vector of the hyperplane representing the mapping relationship; w h T represents the projection of h on w_r; I m×n represents the m×n identity matrix; The projection matrix of the tail entity projection vector obtained by projecting the tail entity into the relational space is: M rt =w r w t T +I m×n Among them, M rt represents the projection matrix of the tail entity; w r The norm vector of the hyperplane representing the mapping relationship; w t T represents the projection of t on w_r; The scoring function is defined as: Among them, f r (h, t) indicates that the scoring function is defined using the head entity projection vector and the tail entity projection vector; h represents the head entity; t represents the tail entity; and r represents the mapping relationship from the head entity to the tail entity.

4. The hybrid vehicle powertrain knowledge graph reasoning method according to claim 2, characterized in that: In step S2, the label similarity between entity nodes based on the Jaccard similarity coefficient is defined as: Among them, A and B represent entity nodes respectively.

5. The hybrid vehicle powertrain knowledge graph reasoning method according to claim 2, characterized in that: In step S3, the cosine similarity between entities A(x1, y1) and B(x2, y2) is defined as: Among them, cosθ represents cosine similarity; x1 and y1 represent the projection coordinates of entity A in the relational space; x2 and y2 represent the projection coordinates of entity B in the relational space; A i ,B i Represent the components of vectors A and B respectively; n represents the dimensions of vectors A and B.

6. A hybrid vehicle transmission system rapid design system, characterized by: Includes data layer, knowledge graph layer and application layer; The data layer includes a knowledge base and an example base of hybrid vehicle transmission systems; The knowledge graph layer includes a hybrid vehicle transmission system ontology model constructed based on a knowledge base and an instance base of the hybrid vehicle transmission system, and a hybrid vehicle transmission system knowledge graph obtained by subjecting the hybrid vehicle transmission system ontology model to knowledge processing including entity recognition, relationship extraction, and entity alignment; An inference engine is provided in the application layer, and the inference engine designs a hybrid vehicle transmission system according to design requirements using the hybrid vehicle transmission system knowledge graph inference method described in any one of claims 2 to 5; the method for the inference engine to design the hybrid vehicle transmission system is: after inputting the design requirements and parameters, perform instance retrieval; determine whether there is an instance in the instance library with similarity that meets the requirements: if so, use the instance as the design result; if not, use the hybrid vehicle transmission system knowledge graph inference method to develop new products, and save the obtained new products to the instance library to supplement the hybrid vehicle transmission system knowledge graph.

7. The hybrid vehicle transmission system rapid design system according to claim 6, characterized in that: There are three ways to develop new products using the knowledge graph reasoning method of hybrid vehicle transmission system, including local design, step-by-step design and automatic design.

8. The hybrid vehicle transmission system rapid design system according to claim 7, characterized in that: The method of the local design is: 11) Select the component to be designed and enter the design requirements and parameters; 12) Retrieve design solutions using knowledge graph reasoning methods for hybrid vehicle transmission systems; 13) Determine whether the design solution meets the requirements: if yes, go to step 14); if not, go to step 12); 14) Under the new design of the component, check or redesign the relevant parameters of other components coupled with the component through the knowledge graph of the hybrid vehicle transmission system; 15) Determine whether the design solutions of other components coupled with the component meet the requirements: if so, obtain the design solution; if not, execute step 12).

9. The hybrid vehicle transmission system rapid design system according to claim 7, characterized in that: The step-by-step design approach is: 21) Select the design scope and related design parameters; 22) Use the knowledge graph reasoning method to obtain the design scheme of hybrid vehicle transmission system; 23) Determine whether the design scheme meets the requirements: if yes, go to step 24); if not, go to step 22); 24) Use the local design method to design all parts in sequence until all parts are designed; 25) Verify and evaluate the design plan: if it meets the requirements, output the design plan; if it does not meet the requirements, make local modifications using local design methods until it meets the requirements.

10. The hybrid vehicle transmission system rapid design system according to claim 7, characterized in that: The automatic design method is: 31) Select the design scope and related design parameters; 32) Use the knowledge graph reasoning method of the hybrid electric vehicle transmission system to obtain the design scheme and complete the design of all components; 33) Verify and evaluate the design plan: If it meets the requirements, the design plan is output; If the overall design does not meet the requirements, proceed to step 32); If part of the design does not meet the requirements, it will be modified partially using local design methods until it meets the requirements.

Images