A systematic multidimensional digital characterization method and system for aero-engines

Abstract

This invention relates to the field of aero-engine digitization, and discloses a systematic multi-dimensional digital representation method and system for aero-engines. The method involves decomposing the component structure of an aero-engine hierarchically to form an engine product decomposition structure. Based on the product hierarchy and inclusion relationships of this decomposition structure, a multi-dimensional representation framework for the aero-engine is constructed. Feature factors for each dimension are obtained by extracting associated features from the models or data derived from each dimension. Using these extracted feature factors as variable ontologies, semantic web technology is employed to form a skeleton variable map with these feature factors as its framework. Data is dynamically input into the skeleton variable map to form an aero-engine representation space based on this map. Machine learning algorithms are used to perform representation traversal calculations and analyses on the aero-engine representation space to obtain the representation traversal calculation and analysis results. This method solves the representation challenges of existing representation technologies for complex systems like aero-engines.

Description

Technical Field

[0001] This invention relates to the field of aero-engine digitization, and discloses a systematic multidimensional digital representation method and system for aero-engines. Background Technology

[0002] In engineering, representation refers to the process and methods of transforming the key attributes, behaviors, or relationships of engineering physical systems, equipment, or technical problems into a structured, quantifiable form to describe, analyze, or manipulate them. Its core lies in formal abstraction to serve engineering design, simulation, optimization, or decision-making. The core elements of engineering representation include: object simplification, mathematical / symbolic mapping, visualization / datafication, and computability.

[0003] Existing characterization techniques require engineers with advanced technical knowledge. Geometric and graphical characterization necessitates knowledge in structural and finite element methods, while mathematical and symbolic characterization requires a strong mathematical foundation. Data-driven characterization demands expertise in data acquisition, cleaning, and big data, and multi-scale and multi-physics characterization requires the integration of interdisciplinary knowledge. These specialized fields require the efforts of multiple teams. Furthermore, existing characterization techniques are designed from the perspective of their respective professional fields, rather than representing the entire system from a top-down perspective. This lack of systematic approach results in incomplete characterization, necessitating the systematic construction of a characterization framework. Moreover, the simplification of objects during characterization has specific adaptability and limitations. These simplifications become ineffective under different conditions and environments. For example, in the famous Leaning Tower of Pisa experiment, using two iron balls allowed for the neglect of air resistance's impact on the falling speed. However, when the test objects were replaced with feathers and iron balls, this simplification yielded completely opposite experimental results. The main problem in engineering often lies in the fact that before we have fully understood the mechanism, we may not know whether a certain factor will have a significant impact on the result. If we simplify things at this time, it may lead to incorrect results. Summary of the Invention

[0004] The purpose of this invention is to provide a systematic multidimensional digital characterization method and system for aero-engines, which can systematically establish a complete top-down engine characterization system through a hierarchical and structured characterization system, thus solving the characterization problems of existing characterization technologies for complex systems such as aero-engines.

[0005] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is as follows:

[0006] A systematic multidimensional digital representation method for aero-engines includes:

[0007] Using the aero-engine product decomposition structure as the pillar, and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical structured organization of the representation system, multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, a multi-dimensional representation framework for aero-engines is constructed.

[0008] A model analysis method is used to extract associated features from the models or data derived from each dimension, resulting in feature factors for each dimension. These feature factors include design parameters, calculation parameters, and boundary conditions for the corresponding aero-engine components extracted from CAD or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logical models. The calculation parameters include load parameters and material performance parameters from the CAE model. The logical model includes a hierarchical model expressing parent-child nodes in a tree structure, as well as a many-to-many network model, data flow graph, and entity relationship graph.

[0009] Using the feature factors extracted from each dimension as variable ontology, semantic web technology is used to convert all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data into graph training data, and machine learning training is used to form a skeleton variable graph with feature factors as the skeleton;

[0010] Using all or part of the aero-engine design data, production data, simulation history data, test history data, and operation and use history data as spatial training data, the skeleton variable map is input into the space, and machine learning algorithms are used to construct an aero-engine representation space based on the skeleton variable map.

[0011] Machine learning algorithms were used to perform representation traversal calculation and analysis on the aero-engine representation space, and the results of the representation traversal calculation and analysis were obtained.

[0012] Furthermore, it also includes:

[0013] Based on the results of the representation traversal calculation and analysis, and combined with the deterministic rule model, the skeleton variable map is optimized to obtain the optimized skeleton variable map; the deterministic rules include design rules, process rules, and usage rules in engineering practice;

[0014] Spatial training data is input into the optimized skeleton variable map to form an aero-engine characterization space based on the optimized skeleton variable map;

[0015] Machine learning algorithms were used to perform representation traversal calculation and analysis on the aero-engine representation space based on the optimized skeleton variable map, and the results of the representation traversal calculation and analysis were obtained.

[0016] Furthermore, methods for constructing an aero-engine representation space based on a skeleton variable map using machine learning algorithms include:

[0017] Based on the feature factors in the spatial training data, machine learning algorithms are used to establish correlations between the corresponding feature factors and the nodes of the skeleton variable map, as well as to establish correlations between the feature factors.

[0018] Using the correlated feature factors as the representation factors of the representation space, the representation factors that are correlated with each node in the skeleton variable map and the corresponding correlation descriptions are combined to form a representation model. All representation models are combined to form the aero-engine representation space based on the skeleton variable map.

[0019] Furthermore, it also includes:

[0020] After establishing the representation space, all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data accumulated over a certain period of time are input into the representation space as incremental data, and new feature factors are extracted from the incremental data.

[0021] Using the representation space as the traversal space, and relying on the correlation threshold between each representation factor in the representation space as the evolution direction, new feature factors are added to the current representation model for dynamic combination to form different new representation models.

[0022] Sensitivity analysis of each characterization factor in the newly added characterization model is conducted, and the results of the sensitivity analysis are used as the basis for evaluating the effectiveness of the combination. New characterization models whose sensitivity analysis results are greater than the preset limit value are considered to be effective new characterization models.

[0023] The effective new representation models are added to the representation space to participate in the representation traversal calculation and evolution.

[0024] Furthermore, the characterization dimensions in the multi-dimensional characterization framework for aero-engines include engine product decomposition structure dimension, external characteristic parameter dimension, functional / performance parameter dimension, physical characteristic dimension, manufacturing process dimension, environmental factor dimension, mission sequence dimension, and failure dimension.

[0025] To achieve the above-mentioned technical effects, the present invention also provides a systematic multidimensional digital characterization system for aero-engines, used to implement the aforementioned systematic multidimensional digital characterization method for aero-engines, comprising:

[0026] The representation framework construction module is used to take the aero-engine product decomposition structure as the pillar and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical structured organization of the representation system. Multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, a multi-dimensional representation framework for aero-engines is constructed.

[0027] The feature factor extraction module is used to extract associated features from models or data derived from each dimension using model analysis methods, thereby obtaining feature factors for each dimension. These feature factors include design parameters, calculation parameters, and boundary conditions for corresponding aero-engine components extracted from CAD or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logical models. The calculation parameters include load parameters and material performance parameters from the CAE model. The logical model includes hierarchical models expressing parent-child nodes in a tree structure, as well as many-to-many network models, data flow graphs, and entity relationship graphs.

[0028] The variable graph generation module is used to use the feature factors extracted from various dimensions as variable ontology. It uses semantic web technology to convert all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data into graph training data, and forms a skeleton variable graph with feature factors as the skeleton through machine learning training.

[0029] The characterization space generation module is used to input all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data as spatial training data into the skeleton variable map, and to construct an aero-engine characterization space based on the skeleton variable map using machine learning algorithms.

[0030] The representation analysis module is used to perform representation traversal calculation and analysis on the aero-engine representation space using machine learning algorithms, and obtain the representation traversal calculation and analysis results.

[0031] Furthermore, the variable map generation module includes a variable map optimization unit, which optimizes the skeleton variable map based on the representation traversal calculation and analysis results, combined with a deterministic rule model, to obtain an optimized skeleton variable map; the deterministic rules include design rules, process rules, and usage rules in engineering practice;

[0032] The representation space generation module includes a representation space optimization unit, which is used to input spatial training data into the optimized skeleton variable map to form an aero-engine representation space based on the optimized skeleton variable map.

[0033] The characterization analysis module is also used to perform characterization traversal calculation analysis on the aero-engine characterization space based on the optimized skeleton variable map using machine learning algorithms, and obtain the characterization traversal calculation analysis results.

[0034] Furthermore, the representation space generation module also includes:

[0035] The correlation establishment unit is used to establish correlations between the corresponding feature factors and the nodes of the skeleton variable map based on the feature factors in the spatial training data, and to establish correlations between the feature factors.

[0036] The representation model combination unit is used to take the related feature factors as the representation factors of the representation space. It combines the related representation factors and the corresponding related descriptions of each node in the skeleton variable map to form a representation model. All representation models are combined to form the aero-engine representation space based on the skeleton variable map.

[0037] Furthermore, in the characterization framework construction module, the characterization dimensions in the multi-dimensional characterization framework for aero-engines include engine product decomposition structure dimension, external characteristic parameter dimension, functional / performance parameter dimension, physical characteristic dimension, manufacturing process dimension, environmental factor dimension, mission sequence dimension, and failure dimension.

[0038] Compared with the prior art, the beneficial effects of this invention are:

[0039] By establishing a correlation and quantitative relationship model between characterization factors and specific targets, a strong correlation is established between characterization factors and targets. Then, through a hierarchical and structured characterization system, the characterization complexity problem of complex systems such as aero-engines is solved. Compared with traditional characterization and modeling techniques, the aero-engine characterization space constructed by this invention can comprehensively reflect information on all aspects of engine design, production, simulation, testing and operation, making the characterization more systematic and complete. Attached Figure Description

[0040] Figure 1 This is a flowchart of the systematic multidimensional digital characterization method for aero-engines in Example 1;

[0041] Figure 2 This is a block diagram of the systematic multidimensional digital characterization system for aero-engines in Example 1;

[0042] Figure 3 This is a flowchart of the systematic multidimensional digital characterization method for aero-engines in Example 2;

[0043] The module comprises: 1. Representation framework construction module; 2. Feature factor extraction module; 3. Variable map generation module; 301. Variable map optimization unit; 4. Representation space generation module; 401. Representation space optimization unit; 402. Correlation establishment unit; 403. Representation model combination unit; and 5. Representation analysis module. Detailed Implementation

[0044] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0045] Example 1

[0046] See Figures 1 to 2 A systematic multidimensional digital representation method for aero-engines includes:

[0047] Using the aero-engine product decomposition structure as the pillar, and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical structured organization of the representation system, multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, a multi-dimensional representation framework for aero-engines is constructed.

[0048] A model analysis method is used to extract associated features from the models or data derived from each dimension, resulting in feature factors for each dimension. These feature factors include design parameters, calculation parameters, and boundary conditions for the corresponding aero-engine components extracted from CAD or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logical models. The calculation parameters include load parameters and material performance parameters from the CAE model. The logical model includes a hierarchical model expressing parent-child nodes in a tree structure, as well as a many-to-many network model, data flow graph, and entity relationship graph.

[0049] Using the feature factors extracted from each dimension as variable ontology, semantic web technology is used to convert all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data into graph training data, and machine learning training is used to form a skeleton variable graph with feature factors as the skeleton;

[0050] Using all or part of the aero-engine design data, production data, simulation history data, test history data, and operation and use history data as spatial training data, the skeleton variable map is input into the space, and machine learning algorithms are used to construct an aero-engine representation space based on the skeleton variable map.

[0051] Machine learning algorithms were used to perform representation traversal calculation and analysis on the aero-engine representation space, and the results of the representation traversal calculation and analysis were obtained.

[0052] In this embodiment, a structured hierarchical representation organization is constructed based on the Product Breakdown Structure (PBS). A general representation framework for aero-engines is built from multiple dimensions, and feature factors of models or data in each dimension are extracted using model analysis methods. Then, using the feature factors of each dimension as the core, semantic web technology is used to convert various types of data into graph training data, and a skeleton variable graph is formed through machine learning training. By inputting aero-engine-related data into the skeleton variable graph, a representation space is constructed using machine learning algorithms, forming a general aero-engine representation system. By performing representation traversal calculation analysis on the aero-engine representation space, the representation traversal calculation analysis results can be obtained. This invention establishes a strong correlation between representation factors and targets by establishing a correlation and quantitative relationship model between representation factors and specific targets. Then, through a hierarchical and structured representation system, it solves the representation complexity problem of complex systems such as aero-engines. Compared with traditional representation and modeling techniques, the aero-engine representation space constructed by this invention can comprehensively reflect information from all aspects of engine design, production, simulation, testing, and operation, making the representation more systematic and complete.

[0053] Based on the same inventive concept, this embodiment also provides a systematic multidimensional digital representation system for aero-engines, used to implement the aforementioned systematic multidimensional digital representation method for aero-engines, including:

[0054] The representation framework construction module 1 is used to take the aero-engine product decomposition structure as the pillar, and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical structured organization of the representation system. Multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, a multi-dimensional representation framework for aero-engines is constructed.

[0055] Feature factor extraction module 2 is used to extract associated features from the models or data derived from each dimension using model analysis methods, thereby obtaining feature factors for each dimension. The feature factors include design parameters, calculation parameters, and boundary conditions of the corresponding aero-engine components extracted from CAD models or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logic models. The calculation parameters include load parameters and material performance parameters in the CAE model. The logic model includes a hierarchical model expressing parent-child nodes in a tree structure, as well as a many-to-many network model, data flow graph, and entity relationship graph.

[0056] The variable graph generation module 3 is used to use the feature factors extracted from each dimension as variable ontology, and to use semantic web technology to convert all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data into graph training data, and to form a skeleton variable graph with feature factors as the skeleton through machine learning training.

[0057] The characterization space generation module 4 is used to input all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data as spatial training data into the skeleton variable map, and to construct an aero-engine characterization space based on the skeleton variable map using machine learning algorithms.

[0058] The representation analysis module 5 is used to perform representation traversal calculation and analysis on the aero-engine representation space using machine learning algorithms, and obtain the representation traversal calculation and analysis results.

[0059] In this embodiment, the variable map generation module 3 includes a variable map optimization unit 301, which is used to optimize the skeleton variable map based on the representation traversal calculation and analysis results and in combination with a deterministic rule model to obtain an optimized skeleton variable map; the deterministic rules include design rules, process rules and usage rules in engineering practice;

[0060] The representation space generation module 4 includes a representation space optimization unit 401, which is used to input spatial training data into the optimized skeleton variable map to form an aero-engine representation space based on the optimized skeleton variable map.

[0061] Based on this, the representation analysis module 5 is also used to perform representation traversal calculation analysis on the aero-engine representation space based on the optimized skeleton variable map using machine learning algorithms, and obtain the representation traversal calculation analysis results.

[0062] In this embodiment, the representation space generation module 4 further includes:

[0063] The correlation establishment unit 402 is used to establish correlations between the corresponding feature factors and the nodes of the skeleton variable map based on the feature factors in the spatial training data using a machine learning algorithm, and to establish correlations between the feature factors.

[0064] The representation model combination unit 403 is used to use the correlated feature factors as the representation factors of the representation space, and to form a representation model together with the correlated representation factors and corresponding correlation descriptions of each node in the skeleton variable map. All representation models are combined to form an aero-engine representation space based on the skeleton variable map.

[0065] Example 2

[0066] See Figure 3 A systematic multidimensional digital representation method for aero-engines includes:

[0067] Step 1: Using the aero-engine product decomposition structure as the pillar, and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical structured organization of the representation system, multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, a multi-dimensional representation framework for aero-engines is constructed.

[0068] In this embodiment, based on the Product Breakdown Structure (PBS) of aero-engines, an 8-Dimensional Representation Framework (8DRF) for aero-engines is constructed, which includes:

[0069] Dimension 1: Engine Product Decomposition Structure Dimension;

[0070] Dimension 2: External characteristic parameters, including: aircraft type compatibility, flight characteristic parameters, environmental characteristic requirements, user characteristic requirements, etc., which can be decomposed based on the requirements of relevant parties for engine operation, use and maintenance, etc.

[0071] Dimension 3: Functional / performance parameter dimension, including indicators such as engine thrust, EGT, vibration spectrum, as well as external characteristics-function / performance relationship, function / performance-physical environment relationship, etc. Tensor decomposition can be performed based on the engine performance index system, and horizontal and vertical relationships can be expressed.

[0072] Dimension 4: Physical characteristics, including: external dimensions, shape, weight, center of gravity, moment of inertia, material properties (creep coefficient, fatigue crack propagation rate, etc.), internal and external physical environment, etc.

[0073] Dimension 5: Manufacturing process dimension, including: manufacturing tolerances, assembly clearances, etc.;

[0074] Dimension 6: Environmental factors, including: altitude, temperature, humidity, Mach number, etc.

[0075] Dimension 7: Mission sequence dimension, including: flight cycle, mission profile, etc.;

[0076] Dimension 8: Failure and failure dimension, including FMEA, FTA, etc.

[0077] Step 2: Use model analysis methods to extract associated features from the models or data derived from each dimension, obtaining feature factors for each dimension. These feature factors include design parameters, calculation parameters, and boundary conditions for the corresponding aero-engine components extracted from CAD or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logical models. The calculation parameters include load parameters and material performance parameters from the CAE model. The logical model includes hierarchical models expressing parent-child nodes in a tree structure, as well as many-to-many network models, data flow graphs, and entity relationship graphs.

[0078] In this embodiment, key variables include the single independent variable upon which a univariate function and its derivative depend, and multiple independent variables upon which a multivariate function and its partial derivatives depend; state vectors include each variable in the minimum independent variable group (such as capacitor voltage and inductor current in a circuit system, displacement and velocity in a mechanical system); input and output quantities include external forces such as applied force and voltage as control inputs, and observable physical quantities such as sensor measurements as outputs; nodes include parent-child nodes in a tree structure, many-to-many nodes in a graph structure, and active nodes in an activity graph; logical relationship features include hierarchical relationships, sequential relationships, and associative relationships.

[0079] This embodiment takes the feature factor extraction of the engine product decomposition structure dimension as an example to explain the feature factor extraction process of the corresponding dimension in this step in detail. The specific steps are as follows:

[0080] 2.1 Based on the product breakdown structure (PBS) of the engine, construct the engine technology breakdown structure (TBS) and work breakdown structure (WBS), and based on the work breakdown structure (WBS), complete the construction of the engine schedule breakdown and resource breakdown structure (OBS), and based on the resource breakdown structure (OBS), complete the construction of the engine cost breakdown structure (CBS).

[0081] 2.2 The model analysis method is used to extract the associated features of the model derived from the Product Breakdown Structure (PBS) of the engine product; the model derived from the Product Breakdown Structure (PBS) of the engine product includes CAD model, CAE model, differential equation / partial differential equation / state equation / logic model.

[0082] Furthermore, for the feature factor extraction of the "functional / performance parameter dimension", based on the existing aero-engine indicator system, we can extract the indicator factors of layer L and layer L+1 (L takes values ​​of 0, 1, 2, 3... layer 0 indicator factors are the top-level target factors of the engine) from the indicator tree, as well as the correlation between layer L and layer L+1 indicator factors (including but not limited to inclusion relationship, correlation, etc.), and construct the target factor tree of the characterization system to obtain the feature factors of the functional / performance parameter dimension.

[0083] The 8DRF (Engine 8-Dimensional Representation Framework) extracts features for other dimensions primarily through semantic recognition, large language models, knowledge graphs, and model parsing (applicable to rule-based models, computational models, task models, behavioral models, etc.). For example, it extracts engine model adaptability features, mission scenario features, flight characteristics, mission environment characteristics, and user characteristics from textual descriptions, graphics, and model data in the external characteristic dimension.

[0084] In this embodiment, feature extraction is the foundation of the present invention. Through the APIs of different CAD / CAE and other modeling tools, and through adapter development, fine-grained analysis of the model can be achieved, and various feature parameters, such as design size parameters, material parameters, load and boundary conditions, can be obtained from the model and data to form a set of feature parameters.

[0085] Step 3: Using the feature factors extracted from each dimension as variable ontology, semantic web technology is used to convert all or part of the aero-engine design data, production data, simulation history data, test history data, and operation and use history data into graph training data. Through machine learning training, a skeleton variable graph with feature factors as the skeleton is formed.

[0086] The variable ontology is a semantic data model used to define the concepts of engine variables and the attributes that can be used to describe them. In this embodiment, the variable ontology has three main components:

[0087] Concepts (classes): Different concepts / types that exist in the data;

[0088] Relationship: An attribute that connects two concepts;

[0089] Attributes: Attributes that describe a single concept.

[0090] Multiple variable entities can be designed based on multi-dimensional, multi-view scenarios and business needs of the engine. Each variable entity can define variables, variable relationships, variable attribute information, etc. Variable entity design is performed via canvas drag-and-drop, providing a WYSIWYG experience. Variable entities support version control to support changes and flexibility in the graph.

[0091] Variable graphs are an application based on knowledge graphs. They are databases that graphically represent the relationships between different variable entities. They connect variable information of different types, dimensions, and levels (such as targets, entities, events, concepts, parameters, attributes, etc.) through relationships to form an organic variable knowledge network. The variable graph will be constructed based on the engine variable ontology design. In this embodiment, general graph construction tools can be used, taking the feature set, indicator system, and variable set collected by the engine as input, and following the step-by-step guidance to construct multiple graphs.

[0092] Step 4: Input all or part of the aero-engine design data, production data, simulation history data, test history data, and operation and use history data as spatial training data into the skeleton variable map, and use machine learning algorithms to construct an aero-engine representation space based on the skeleton variable map;

[0093] In this embodiment, firstly, based on the feature factors in the spatial training data, a machine learning algorithm is used to establish the correlation between the corresponding feature factors and the nodes of the skeleton variable map, and to establish the correlation between the feature factors.

[0094] Using the correlated feature factors as the representation factors of the representation space, the representation factors that are correlated with each node in the skeleton variable map and the corresponding correlation descriptions are combined to form a representation model. All representation models are combined to form the aero-engine representation space based on the skeleton variable map.

[0095] In practical applications, this embodiment can also optimize the skeleton variable map based on the representation traversal calculation and analysis results, combined with a deterministic rule model, to obtain an optimized skeleton variable map. The deterministic rules include design rules, process rules, and usage rules in engineering practice. Spatial training data is input into the optimized skeleton variable map to form an aero-engine representation space based on the optimized skeleton variable map.

[0096] Step 5: Use machine learning algorithms to perform representation traversal calculation and analysis on the aero-engine representation space, and obtain the representation traversal calculation and analysis results.

[0097] In this embodiment, after establishing the representation space, all or part of the aero-engine design data, production data, simulation history data, test history data, and operational history data accumulated over a certain period can be input into the representation space as incremental data in stages, and new feature factors can be extracted from the incremental data. Then, the representation space is used as a traversal space, and the correlation threshold between each representation factor in the representation space is used as the evolution direction to dynamically combine the current representation model with new feature factors to form different new representation models. Sensitivity analysis is performed on each representation factor in the new representation model, and the sensitivity analysis results are used as the basis for evaluating the effectiveness of the combination. New representation models whose sensitivity analysis results are greater than the preset limit value are considered valid new representation models. Valid new representation models are added to the representation space to participate in representation traversal calculation and evolution.

[0098] Furthermore, the multi-dimensional characterization framework for aero-engines includes the engine product decomposition structure dimension, external characteristic parameter dimension, functional / performance parameter dimension, physical characteristic dimension, manufacturing process dimension, environmental factor dimension, mission sequence dimension, and failure dimension.

[0099] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A systematic multidimensional digital representation method for aero-engines, characterized in that, include: Using the aero-engine product decomposition structure as the pillar, and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical and structured organization of the representation system, multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, a multi-dimensional representation framework for aero-engines is constructed. Among them, the representation dimensions in the multi-dimensional representation framework for aero-engines include engine product decomposition structure dimension, external characteristic parameter dimension, functional / performance parameter dimension, physical characteristic dimension, process manufacturing dimension, environmental factor dimension, mission sequence dimension, and failure dimension. A model analysis method is used to extract associated features from the models or data derived from each dimension, resulting in feature factors for each dimension. These feature factors include design parameters, calculation parameters, and boundary conditions for the corresponding aero-engine components extracted from CAD or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logical models. The calculation parameters include load parameters and material performance parameters from the CAE model. The logical model includes a hierarchical model expressing parent-child nodes in a tree structure, as well as a many-to-many network model, data flow graph, and entity relationship graph. Using the feature factors extracted from each dimension as variable ontology, semantic web technology is used to convert all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data into graph training data, and machine learning training is used to form a skeleton variable graph with feature factors as the skeleton; All or part of the aero-engine design data, production data, simulation history data, test history data, and operation and use history data are used as spatial training data input into the skeleton variable map. Based on the feature factors in the spatial training data, machine learning algorithms are used to establish correlations between the corresponding feature factors and the nodes of the skeleton variable map, as well as between the feature factors. Using the correlated feature factors as the representation factors of the representation space, the representation factors that are correlated with each node in the skeleton variable map and the corresponding correlation descriptions are combined to form a representation model. All representation models are combined to form the aero-engine representation space based on the skeleton variable map. Machine learning algorithms were used to perform representation traversal calculation and analysis on the aero-engine representation space, and the results of the representation traversal calculation and analysis were obtained.

2. The systematic multidimensional digital characterization method for aero-engines according to claim 1, characterized in that, Also includes: Based on the results of the representation traversal calculation and analysis, and combined with the deterministic rule model, the skeleton variable map is optimized to obtain the optimized skeleton variable map; the deterministic rules include design rules, process rules, and usage rules in engineering practice; Spatial training data is input into the optimized skeleton variable map to form an aero-engine characterization space based on the optimized skeleton variable map; Machine learning algorithms were used to perform representation traversal calculation and analysis on the aero-engine representation space based on the optimized skeleton variable map, and the results of the representation traversal calculation and analysis were obtained.

3. The systematic multidimensional digital characterization method for aero-engines according to claim 1, characterized in that, Also includes: After establishing the representation space, all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data accumulated over a certain period of time are input into the representation space as incremental data, and new feature factors are extracted from the incremental data. Using the representation space as the traversal space, and relying on the correlation threshold between each representation factor in the representation space as the evolution direction, new feature factors are added to the current representation model for dynamic combination to form different new representation models. Sensitivity analysis of each characterization factor in the newly added characterization model is conducted, and the results of the sensitivity analysis are used as the basis for evaluating the effectiveness of the combination. New characterization models whose sensitivity analysis results are greater than the preset limit value are considered to be effective new characterization models. The effective new representation models are added to the representation space to participate in the representation traversal calculation and evolution.

4. A systematic multidimensional digital representation system for aero-engines, used to implement the systematic multidimensional digital representation method for aero-engines as described in claim 1, characterized in that, include: The representation framework construction module is used to construct a multi-dimensional representation framework for aero-engines, using the aero-engine product decomposition structure as the pillar and the node objects of each layer of the aero-engine product decomposition structure as the carrier of the hierarchical and structured organization of the representation system. Multiple representation dimensions are recursively derived from the node objects of each layer of the aero-engine product decomposition structure. According to the product hierarchy and inclusion relationship of the aero-engine product decomposition structure, the representation dimensions in the multi-dimensional representation framework for aero-engines include engine product decomposition structure dimension, external characteristic parameter dimension, functional / performance parameter dimension, physical characteristic dimension, process manufacturing dimension, environmental factor dimension, mission sequence dimension, and fault failure dimension. The feature factor extraction module is used to extract associated features from models or data derived from each dimension using model analysis methods, thereby obtaining feature factors for each dimension. These feature factors include design parameters, calculation parameters, and boundary conditions for corresponding aero-engine components extracted from CAD or CAE models, as well as variables, initial conditions, state vectors, input / output quantities, nodes, and logical relationship features extracted from differential equations, partial differential equations, state equations, and logical models. The calculation parameters include load parameters and material performance parameters from the CAE model. The logical model includes hierarchical models expressing parent-child nodes in a tree structure, as well as many-to-many network models, data flow graphs, and entity relationship graphs. The variable graph generation module is used to use the feature factors extracted from various dimensions as variable ontology. It uses semantic web technology to convert all or part of the aero-engine design data, production data, simulation history data, test history data and operation and use history data into graph training data, and forms a skeleton variable graph with feature factors as the skeleton through machine learning training. The representation space generation module is used to input all or part of the aero-engine design data, production data, simulation history data, test history data, and operational history data as spatial training data into the skeleton variable map. Based on the feature factors in the spatial training data, a machine learning algorithm is used to establish correlations between the corresponding feature factors and the nodes of the skeleton variable map, as well as between the feature factors. The correlated feature factors are used as representation factors of the representation space. The representation factors correlated with each node in the skeleton variable map and their corresponding correlation descriptions are combined to form a representation model. All representation models are combined to form an aero-engine representation space based on the skeleton variable map. The representation analysis module is used to perform representation traversal calculation and analysis on the aero-engine representation space using machine learning algorithms, and obtain the representation traversal calculation and analysis results.

5. The systematic multidimensional digital characterization system for aero-engines according to claim 4, characterized in that, The variable map generation module includes a variable map optimization unit, which optimizes the skeleton variable map based on the representation traversal calculation and analysis results and in combination with a deterministic rule model to obtain an optimized skeleton variable map; the deterministic rules include design rules, process rules and usage rules in engineering practice; The representation space generation module includes a representation space optimization unit, which is used to input spatial training data into the optimized skeleton variable map to form an aero-engine representation space based on the optimized skeleton variable map. The characterization analysis module is also used to perform characterization traversal calculation analysis on the aero-engine characterization space based on the optimized skeleton variable map using machine learning algorithms, and obtain the characterization traversal calculation analysis results.

Images