A method for constructing a digital engineering model of an aerospace engine

Abstract

This invention belongs to the field of digital engineering technology, specifically to a method for constructing a digital engineering model for an aircraft engine. The specific technical solution is: a digital engineering model construction method that constructs an intelligent digital engineering model that fully corresponds to the physical architecture based on the matching relationship between the physical object corresponding to the model to be constructed and the actual working state of the physical object. The digital engineering model disclosed in this invention features physical rule operation, close performance tracking, and extremely fast dynamic response. It can also simultaneously utilize both complete machine test data and component system-level test data, effectively improving the accuracy and speed of aircraft engine digital engineering.

Description

Technical Field

[0001] The invention belongs to the field of digital engineering technology, and in particular relates to a method for constructing a digital engineering model of an aero-engine. Background Art

[0002] The digital transformation of aircraft engines is a crucial technological approach for revolutionizing their product R&D, manufacturing, and support models, and the digital engineering model for aircraft engines is a core component of this digital transformation. During the demonstration phase, the digital engineering model can be integrated into a simulation environment to generate design solutions, analyzing trade-offs among them based on performance, cost, and risk, ultimately identifying the optimal alternative. During the development phase, the digital engineering model conducts digital testing and appraisal based on the equipment design plan, accelerating the development process. During the equipment production phase, the digital engineering model establishes a digital quality assurance process based on the automated collection and analysis of production data, optimizing and controlling the manufacturing process and improving manufacturing maturity. During the equipment support phase, the digital engineering model utilizes automatically acquired equipment status information to enable precise equipment maintenance and support based on predictive analysis.

[0003] However, the working principles of aircraft engines are complex, with numerous control parameters, cross-integration of structured data and unstructured data, and parameter collection is frequent, large in volume, diverse in structure, and highly timely. Therefore, how to use operation and maintenance data to build digital engineering models is a huge challenge.

[0004] Traditional aircraft engine models first establish component characteristics for components, systems, or elements based on test or simulation data. Using the component approach, they then establish the connections between these components based on the physical operating mechanisms of the aircraft engine. Matching operating points are then iterated using the shared operating equations. For example, in the case of a turbofan engine air path model, matching operating points are found on the respective characteristic diagrams using the shared operating equations for each component, including the inlet, fan, compressor, combustor, high-pressure turbine, low-pressure turbine, bypass, mixing chamber, afterburner, and tail nozzle. Air system models, based on bleed air, cooling, and sealing terminal parameters, use the flow and energy balances of each flow path element as shared operating equations to find matching operating points on the element characteristic data. However, this approach struggles to couple the air path, air, oil, accessories, and control systems into a unified model framework. Furthermore, once established, component characteristics remain fixed, making them incapable of tracking aircraft engine performance degradation and, therefore, unsuitable for digital engineering throughout the aircraft engine lifecycle. Furthermore, adaptive models based on traditional aircraft engine models can partially adjust component characteristics, but not all component, system, and element characteristics, making them unsuitable for onboard aircraft engine digital models. At the same time, traditional aircraft engine models cannot reflect individual differences and cannot be used for digital testing.

[0005] Existing methods also use data-driven aeroengine model building. This involves inputting aeroengine test or simulation data into a "black box" intelligent network to train the aeroengine model network. Based on the structure of the aeroengine model being built, this method collects input, output, and state variable data from the aeroengine state variable model. After denoising this data, it uses this data as training data to train a neural network model, resulting in an aeroengine intelligent network model. Based on this established aeroengine intelligent network model, the chain rule is used to calculate the partial derivatives of the output to the input using the partial derivative representation of the aeroengine model parameters, thereby establishing a data-driven aeroengine state variable model. However, this method is limited by the physical laws of the aeroengine, which can lead to the risk of exceeding these laws. Furthermore, the amount of data required to be collected and processed is enormous.

[0006] Existing technologies also include aircraft engine models driven by data and physics models. These models typically calibrate traditional aircraft engine models and then use deep learning techniques to build data-driven models. Alternatively, they dynamically adjust traditional aircraft engine models using artificial intelligence. However, these methods still rely on traditional aircraft engine models and fail to integrate the gas path, air, oil, accessories, and control systems into a unified model framework. This results in limited accuracy and limited dynamic adjustment speed.

[0007] Therefore, if a method for constructing a digital engineering model of an aero-engine driven by an intelligent architecture can be provided, it will have important scientific research value and practical significance. Summary of the Invention

[0008] The purpose of the present invention is to provide a method for an aero-engine digital engineering model driven by an intelligent architecture.

[0009] To achieve the above-mentioned purpose of the invention, the technical solution adopted by the present invention is: a digital engineering model construction method, based on the physical object corresponding to the model to be constructed to which the digital engineering object belongs and the actual working matching relationship of the physical object, to construct an intelligent digital engineering model that is completely corresponding to the physical architecture.

[0010] Preferably, the structure of the physical object corresponding to the model to be constructed is divided into: system and components from large to small.

[0011] Preferably, the method comprises the following steps:

[0012] (1) List the systems and components to be considered based on the type of aircraft engine to which the digital object belongs;

[0013] (2) establishing corresponding sub-networks for the components and systems described in step (1);

[0014] (3) Connecting the mainline system and component sub-networks constructed in step (2) through the actual working matching relationship between the model to be constructed and the corresponding physical object;

[0015] (4) According to the purpose of the digital engineering model, a body network training layer is added to the front end of the input of the digital engineering model established in step (3) to represent the dynamic characteristics or directly input the test run and high-altitude test environment parameter characteristics, and a feature mapping network is added to the final output end of the model as the final output parameter to form a component system-level digital engineering model of the whole machine.

[0016] Preferably, the structure of the physical object corresponding to the model to be constructed is divided into: system, component, and element from large to small.

[0017] Preferably, based on the data source, the physical measurement parameters corresponding to the model to be constructed are classified into corresponding parts, systems or elements according to the measurement locations.

[0018] Preferably, feature screening is performed according to the purpose of building the digital model, redundant parameter variables are removed, and complexity is reduced.

[0019] Preferably, the input features or input parameters in step (4) are activated in batches in a time series based on the connection structure or workflow of the physical object corresponding to the model to be constructed.

[0020] Preferably, based on the component system-level digital engineering model of the whole machine, the test data of the component or system is refined, the processing object is transferred from the component and system to the element, and the element network is connected according to the component composition structure and working matching relationship to form a refined component system sub-network, thereby obtaining the component-level digital engineering model of the whole machine.

[0021] Preferably, based on the component-level digital engineering model of the whole machine, according to the data source, the physical measurement parameters corresponding to the model to be constructed are classified into the corresponding components according to the measurement position, and the features are screened according to the purpose of the digital model to remove redundant parameter variables and reduce the complexity; then, based on the physical measurement connection structure or workflow corresponding to the model to be constructed, the corresponding input parameters are activated in batches in a time series.

[0022] Correspondingly, the digital engineering model construction method is applied in constructing a digital engineering model of an aero-engine.

[0023] The present invention has the following beneficial effects:

[0024] This invention proposes a digital engineering technology for aircraft engines driven by intelligent architecture. This technology integrates the physical architecture of aircraft engines into intelligent network design, enabling efficient integration of multiple systems in digital space, enabling digital models to closely track aircraft engine performance and reflect individual differences, and can be extended to the digital engineering of other mechanical systems. By building an intelligent architecture and utilizing data from aircraft engine R&D, testing, manufacturing, and operation, this method uses intelligent methods to construct and train digital models. This supports all engineering and management activities throughout the aircraft engine life cycle, from concept design to retirement disposal, including digital testing, health management, digital maintenance, extended service, reverse engineering, intelligent control, and production monitoring. This shortens the aircraft engine R&D cycle, reduces testing costs, improves equipment safety, and ensures fast and reliable maintenance.

[0025] Compared to traditional aircraft engine models, this invention offers the advantages of high accuracy and the ability to leverage both complete aircraft test data and component and element test data. It integrates the aircraft engine's physical architecture into an intelligent network, ensuring a one-to-one correspondence between each subnetwork and its matching relationships with the actual aircraft engine's physical architecture. It integrates the aircraft engine's gas path, air, oil, control, and accessory systems with high accuracy, and each subnetwork can be further subdivided, allowing continued utilization of component and element test data.

[0026] Compared to aviation engine models driven by data and physical models, the present invention also has the advantages of high precision, high speed, and the ability to utilize component and element test data. The intelligent network architecture of the present invention completely breaks free from the limitations of fixed mathematical and physical equations, and thus completely breaks free from the limitations of traditional mathematical models. The intelligent network and the actual engine are highly integrated, offering the advantage of high precision. At the same time, the intelligent network architecture of the present invention is simply a network architecture, without the addition of mathematical and physical models, and has the advantage of high speed. Moreover, the network of the present invention can be further refined, and component and element test data can be utilized.

[0027] Compared to data-driven aircraft engine models, this method requires less data and does not violate physical constraints. Each subnetwork of the intelligent network and its matching relationships correspond one-to-one with the physical architecture of an actual aircraft engine. This subdivided network eliminates the need for a single, all-in-one training process, thus requiring less data. Furthermore, because the subnetworks and their matching relationships are highly correlated and correspond one-to-one with the actual aircraft engine architecture, physical constraints are avoided.

[0028] In summary, the digital engineering model disclosed in the present invention has the characteristics of operating according to physical rules, closely tracking performance and dynamic response, and can simultaneously use whole-machine test data and component system-level test data, effectively improving the accuracy and speed of digital engineering of aircraft engines. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a diagram showing the working principle of a mixed exhaust turbofan engine;

[0030] Figure 2 The overall flow chart for constructing a digital engineering model of an aero-engine at the component system level using the method of the present invention;

[0031] Figure 3 This is a diagram of the neural network structure between the nozzle training network layer and the parameter output layer;

[0032] Figure 4 The present invention is a flow chart for refining the digital engineering model of an aero-engine component level compressor using the method of the present invention. DETAILED DESCRIPTION

[0033] The present invention provides a method for constructing a digital engineering model of an aerospace engine. The method primarily involves constructing an intelligent digital engineering model that fully corresponds to the physical architecture based on the aerospace engine type to which the digital engineering object belongs and its actual operational matching relationship. The specific process for constructing the digital engineering model is as follows:

[0034] 1. Based on the aircraft engine type to which the digital object belongs, list the systems, components, and elements to be considered. Taking the component-system-level digital engineering model as an example, these components include but are not limited to the inlet, fan, compressor, combustor, high-pressure turbine, low-pressure turbine, bypass, mixing chamber, afterburner, nozzle, and other components, as well as the lubricating oil system, control system, air system, and accessory systems. If, based on the components and systems, you want to further refine the component-level digital engineering model, you can further refine each component and system to the component level. For example, you can refine the compressor components to the first and second stages, and the air system to pipes, gaps, chambers, seals, and so on. After listing all the systems, components, and elements to be considered, first establish the drive system and component-level digital engineering models of the entire aircraft in steps 2-7. Based on this, you can then establish the component-level digital engineering model of the entire aircraft in steps 8-9.

[0035] 2. Establish corresponding sub-networks for the components and systems described in step 1. Each sub-network can be initially selected or designed based on the characteristics of the digital model object and the corresponding parameter attributes. Subsequently, during the data training process, the internal nodes of the sub-network and the connections between them will be rapidly adjusted and streamlined. The optimal adjustment and streamlining approach is to refer to the maximum entropy acceleration training principle described in the applicant's prior patent application CN 111967202 A, and gradually streamline each sub-network during the training process.

[0036] 3. Connect the subnetworks of the main flow path components constructed in Step 2 (i.e., the systems and components described in Step 1) using the actual operational matching relationships of the aircraft engine. The entire main flow path digital model framework encompasses all aircraft engine components to be considered. The number of subnetworks within the framework exactly matches the number of corresponding physical components, encompassing all components and systems listed in Step 1. This forms the main structure of the component system-level digital engineering model for the entire aircraft.

[0037] 4. Add subnetworks corresponding to the air system, oil system, and control system and connect them to the main digital model structure established in Step 3. The upstream and downstream connections of the subnetworks in this step are determined based on the specific location and function of the system in the aircraft engine, which will change the number of input and output features of the main subnetwork in Step 3. The input of each subnetwork of the component and system consists of two parts: one is the abstract output features of the upstream subnetwork connected to the subnetwork, and the other is the measured parameter features of the actual physical structure corresponding to the subnetwork.

[0038] 5. According to the purpose of the digital engineering model, add an airframe network training layer to the front end of the input of the digital engineering model established in step 4 to represent the dynamic characteristics, or directly input the test run and high-altitude test environment parameter characteristics without considering the airframe, and add a feature mapping network as the final output parameter at the final output end of the model to form a component system-level digital engineering model of the whole machine.

[0039] 6. Based on the data source, classify the aircraft engine measurement parameters according to the measurement location to the corresponding parts, systems, or components (if applicable), and perform feature screening based on the purpose of the digital model to remove redundant parameter variables and reduce complexity. It should be noted that the feature screening method can be a parameter correlation analysis screening method, that is, establishing correlation calculations between parameters and eliminating redundant parameter variables with low correlation with the results. Other methods can also be used. Feature screening is a mature existing technology, and those skilled in the art can select it according to their needs, so it will not be elaborated on in detail.

[0040] 7. The input features of the component system-level digital engineering model constructed based on steps 2-5 are not input at the same time. Instead, the corresponding input parameters are activated in batches in a time series based on the connection structure or workflow. That is, the data is input in a hierarchical manner according to the parameter test position to drive the established digital engineering model.

[0041] 8. Based on the component and system sub-networks of the component system-level complete machine digital engineering model constructed in steps 2-7, continue to refine them according to the test data of the corresponding components or systems. The process of establishing the "refined component sub-network" refers to steps 3 and 4, and the processing object is transferred from components and systems to elements, that is, the composition structure of the component system is split, the component test measurement parameters are classified, and the "element network" is connected according to the component composition structure and working matching relationship to form a "refined component system sub-network". For example, the compressor components are refined into the first level, the second level, etc., and the air system is refined into pipelines, gaps, chambers, seals, etc., and so on. By replacing the component system sub-network in the component system-level complete machine digital engineering model with the corresponding refined component sub-network, a component-level complete machine digital engineering model is formed.

[0042] 9. Based on the component-level complete machine digital engineering model, execute steps 6 and 7 to drive the corresponding component-level aero-engine digital engineering model. For example, by refining the corresponding sub-network of the compressor, the inter-stage test data of the compressor component test can be input into the digital engineering model to drive the corresponding component-level complete machine digital engineering model.

[0043] The following will be combined with the accompanying drawings to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0044] Example 1: Constructing a digital engineering model for a hybrid exhaust turbofan engine

[0045] This embodiment focuses on the lubricating oil system, air system, and control system to build a digital engineering model for a hybrid exhaust turbofan engine. It should be noted that the network training layer described in this embodiment and the subnetwork in the solution are the same description. The subnetwork is the name of the digital engineering model construction process, while the network training layer represents the name of the actual application training.

[0046] Figure 1The basic structure and connection form of the aircraft engine of this embodiment. The main components of the aircraft engine include: an air inlet 101, a fan 102, a compressor 103, a combustion chamber 104, a high-pressure turbine 105, a low-pressure turbine 106, a mixing chamber 107, an outer duct 11, an afterburner 108, and a nozzle 109. It also includes a lubricating oil system 31, an air system 21, and a control system 41. According to the components and systems, all measured characteristic parameters contained in the data source are divided and filtered. The division method can be: dividing by the position of the component to which the parameter belongs, that is, the parameter characteristics, for example, dividing according to rules such as airflow parameters and control parameters; the screening method can be: using a parameter correlation analysis screening method, that is, establishing a correlation calculation between parameters and eliminating redundant parameter variables with low correlation with the results. Of course, in practice, according to actual needs, other rules can be followed to divide and filter the characteristic parameters.

[0047] like Figure 1 As shown, the working process of this aircraft engine is as follows: the engine draws air from the atmosphere 110 through the air inlet 101, and the air is pressurized by the fan 102. Subsequently, a portion of the air passes through the duct 11 and enters the mixing chamber 107 and nozzle 109. Another portion of the air is further pressurized by the compressor 103 and enters the combustion chamber 104, where it burns with the fuel, increases its temperature, and expands through the high-pressure turbine 105 and low-pressure turbine 106 to produce work. This portion of air then mixes with the air from the duct 11 in the mixing chamber 107 and enters the afterburner 108. The air is combusted again and discharged through the nozzle 109 back to the atmosphere 110, thereby generating thrust for the engine and aircraft. Furthermore, the air system 21 cools the high-pressure turbine 105 by introducing high-pressure air from the compressor 103. The lubricating oil system 31 lubricates the fan 102, compressor 103, and high and low-pressure turbines 105. The control system 41 controls all control variables.

[0048] Figure 2 This is the overall architecture diagram of the component system-level aero-engine digital engineering model presented in this embodiment. The following sub-network layers are constructed using the traditional neural network construction method. The main structure and process of the model are as follows:

[0049] S01: The S01 layer of the digital engineering model is the airframe training network layer; its input is the engine's overall global propulsion parameters (such as time propulsion or spatial global propulsion parameters, or test environment parameters such as test runs), and its output serves as one of the input parameters of the S02 inlet training network layer.

[0050] S02: S02 is the inlet training network layer; its input parameters are the output parameters of S01 and the intake parameters. The output parameters serve as one of the input parameters for the fan training network layer S03. Inlet parameters are the key engine inlet parameters specified by the aerospace engine discipline, including but not limited to atmospheric pressure, temperature, flight Mach number, and inlet outlet total pressure and temperature. These inlet parameters, as well as subsequent fan and compressor parameters, are all key parameters specified by the aerospace engine discipline. The specific parameters listed here are only examples.

[0051] S03: S03 is the fan training network layer. Its input parameters are the output parameters of the inlet training network layer S02, the output parameters of the lubricating system training network layer S3, the output parameters of the control system training network layer S4, and fan parameters. The output parameters serve as one of the input parameters of the compressor training network layer S04. Fan parameters include but are not limited to the fan inlet and outlet total temperature and pressure, pressure ratio, fan efficiency, speed, and shaft power.

[0052] S04: S04 is the compressor training network layer. Its input parameters are the output parameters of the fan training network layer S03, the output parameters of the lubricating system training network layer S3, the output parameters of the control system training network layer S4, and compressor parameters. The output parameters serve as one of the input parameters of the combustion chamber training network layer S05. Compressor parameters include but are not limited to inlet and outlet total temperature and pressure, pressure ratio, compressor efficiency, speed, and shaft power.

[0053] S05: S05 is the combustion chamber training network layer. Its input parameters are the output parameters of the compressor training network layer S04, the output parameters of the control system training network layer S4, and the combustion chamber parameters. The output parameters serve as one of the input parameters of the high-pressure turbine training network layer S06. Combustion chamber parameters include but are not limited to fuel flow rate, inlet and outlet total temperature and pressure, and combustion efficiency.

[0054] S06: S06 is the high-pressure turbine training network layer. Its input parameters are the output parameters of the combustion chamber turbine training network layer S05, the output parameters of the air system training network layer S2, the output parameters of the lubricating oil system training network layer S3, and the high-pressure turbine parameters. The output parameters serve as one of the input parameters of the low-pressure turbine training network layer S07. High-pressure turbine parameters include but are not limited to inlet and outlet total temperature and pressure, pressure drop ratio, efficiency, speed, shaft power, and turbine work.

[0055] S07: S07 is the low-pressure turbine training network layer. Its input parameters are the output parameters of the high-pressure turbine training network layer S06, the output parameters of the lubricating oil system training network layer S3, the output parameters of the air system network training layer S2, and the low-pressure turbine parameters. The output parameters serve as one of the input parameters of the mixing chamber training network layer S08. Low-pressure turbine parameters include but are not limited to inlet and outlet total temperature and pressure, pressure drop ratio, efficiency, speed, shaft power, and turbine work.

[0056] S08: S08 is the mixing chamber training network layer. Its input parameters are the output parameters of the low-pressure turbine training network layer S07, the output parameters of the bypass network layer S1, and the mixing chamber parameters. The output parameters serve as one of the input parameters of the afterburner training network layer S09. Mixing parameters include but are not limited to the mixing chamber inlet and outlet total temperature and pressure.

[0057] S09: S09 is the afterburner training network layer. Its input parameters are the output parameters of the mixing chamber training network layer S08 and the afterburner parameters. The output parameters serve as one of the input parameters of the nozzle training network layer S10. Afterburner parameters include but are not limited to inlet and outlet total temperature and pressure, and fuel flow rate.

[0058] S10: S10 is the nozzle training network layer; its input parameters are the output parameters of the afterburner training network layer S08, the output parameters of the control system training network layer S4, and the nozzle parameters. The output parameters are finally passed to the parameter output layer S11 through the multi-layer neural training network layer. The construction method of the multi-layer neural training network layer is carried out using the general neural network construction method, which will not be described in detail here. For details, please refer to Figure 3 conduct.

[0059] S11: S11 is the final parameter output layer of the network; the output parameters are mainly set to the main output performance indicators of the aircraft engine, including thrust, specific fuel consumption rate, exhaust temperature, etc., and can also be determined according to actual engineering needs and the purpose of digital engineering model training.

[0060] S1: S1 is the outer duct network training layer; its input parameters are the output parameters of the fan training network layer S03 and the internal parameters of the outer duct system, and the output parameters serve as the input parameters of the mixing chamber network training layer S08.

[0061] S2: S2 is the air system network training layer; its input parameters are the output parameters of the compressor training network layer S04 and the internal parameters of the air system. The output parameters serve as the input parameters of the turbine network training layers S06 and S07.

[0062] S3: S3 is the oil system network training layer; its input parameters are the internal parameters of the oil system, and its output parameters serve as the input parameters of S03, S04, S06 and S07.

[0063] S4: S4 is the control system network training layer; its input parameters are the internal parameters of the control system, and its output parameters serve as the input parameters of S03-S05 and S10.

[0064] Based on the maximum entropy acceleration training principle described in patent CN 111967202 A, the network can be gradually streamlined during training to form a corresponding component system-level aeroengine digital engineering model. Test run data and high-altitude test data are input hierarchically according to test locations to drive the component system-level aeroengine digital engineering model.

[0065] The structural diagram of the component system-level digital engineering model of the aircraft engine described in this embodiment is completely consistent with the principles and architecture followed by the actual operation of the aircraft engine in terms of architectural design; the digital model constructed in this embodiment is completely based on the actual working matching and architecture of the aircraft engine, and is therefore more in line with the engine operation laws and has higher accuracy and reliability.

[0066] like Figure 3 As shown, the connection between the nozzle training network layer S10 (whose results contain all the parameter information of the aircraft engine) and the final parameter output layer S11 is used as an example to simply illustrate the connection between layers. There are multiple hidden layers and fully connected layers between S10 and S11, and there are multiple neuron nodes 201 in each layer. The connection structure between layers is the basic structure in deep learning, which belongs to mature existing technology and is not the focus of protection of this invention, so it will not be elaborated in depth.

[0067] Example 2

[0068] This embodiment is based on the compressor test data, and expands the digital engineering model corresponding to the hybrid exhaust turbofan engine in the first embodiment to form the following Figure 4 The component-level digital engineering model of the entire aero-engine with detailed compressor components is shown. The main structure and process of the model are as follows:

[0069] Similar to establishing the component system-level digital engineering model of the whole machine in Example 1, the compressor component is split into multiple stages of blade elements, and a corresponding element network is established for each blade stage, which is mapped one-to-one to the corresponding blade stage according to the inter-stage measurement parameters in the compressor test.

[0070] The blades of an axial-flow compressor are arranged in a sequential order. Based on the workflow, the input to the component network corresponding to each blade stage includes the output characteristic parameters of the previous-stage component subnetwork and the corresponding measured characteristic parameters. These input parameters are integrated and fed into a specific network structure, which then outputs abstract characteristics. Finally, based on the modeling objectives, a mapping network is added after the final-stage compressor component network to output the corresponding compressor characteristic parameters.

[0071] The refined component subnetwork established above replaces the corresponding component subnetwork in the component system-level digital engineering model of the entire aircraft engine in Example 1, forming a component-level digital engineering model of the entire aircraft engine. The model replaces the input characteristic parameters and output parameters of the corresponding component. Network connections are refined here for operations such as component bleed air. For example, if bleed air occurs in the penultimate compressor stage, the corresponding component network output of the refined component subnetwork is connected to the bleed air endpoint.

[0072] Based on the maximum entropy acceleration training principle described in patent CN 111967202 A, the network can be gradually streamlined during training to form a corresponding component-level aeroengine digital engineering model. Test run data, high-altitude test bench data, and component-level test data are input hierarchically based on test locations to drive the component-level aeroengine digital engineering model.

[0073] It should be noted that, in addition to the compressor, replacement of other components and systems can also be carried out with reference to this embodiment.

[0074] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various deformations, modifications, and substitutions made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A method for constructing a digital engineering model, characterized by: Based on the matching relationship between the physical object and the model to be built, an intelligent digital engineering model that fully corresponds to the physical architecture is constructed. The steps include: (1) List the systems and components to be considered based on the type of aircraft engine to which the digital object belongs; (2) establishing corresponding sub-networks for the components and systems described in step (1); (3) Connecting the mainline system and component sub-networks constructed in step (2) through the actual working matching relationship between the model to be constructed and the corresponding physical object; (4) Add the sub-networks corresponding to the air system, lubricating oil system and control system and connect them to the main structure of the digital model established in step (3); according to the purpose of building the digital engineering model, add the body network training layer to the front end of the input of the digital engineering model established in step (3) to represent the dynamic characteristics or directly input the test run and high-altitude test environment parameter characteristics, and add the feature mapping network as the final output parameter at the final output end of the model to form a component system-level digital engineering model of the whole machine; Step (4) inputting features or input parameters is to activate the corresponding input parameters in batches in a time series based on the connection structure or workflow of the physical object corresponding to the model to be constructed; The following sub-network layers are constructed according to the traditional neural network construction method. The main structure and process of the model are as follows: S01: The S01 layer of the digital engineering model is the airframe training network layer; its input is the overall global propulsion parameters of the engine, and its output is one of the input parameters of the S02 inlet training network layer; S02: S02 is the air intake training network layer; its input parameters are the output parameters of S01 and the air intake parameters. The output parameters serve as one of the input parameters of the fan training network layer S03; S03: S03 is the fan training network layer. Its input parameters are the output parameters of the inlet training network layer S02, the output parameters of the lubricating system training network layer S3, the output parameters of the control system training network layer S4, and the fan parameters. The output parameters serve as one of the input parameters of the compressor training network layer S04. S04: S04 is the compressor training network layer; Its input parameters are the output parameters of the fan training network layer S03, the output parameters of the lubricating oil system training network layer S3, the output parameters of the control system training network layer S4, and the compressor parameters. The output parameters serve as one of the input parameters of the combustion chamber training network layer S05. S05: S05 is the combustion chamber training network layer; Its input parameters are the output parameters of the compressor training network layer S04 and the output parameters of the control system training network layer S4 and the combustion chamber parameters. The output parameters serve as one of the input parameters of the high-pressure turbine training network layer S06. S06: S06 is the high-pressure turbine training network layer; Its input parameters are the output parameters of the combustion chamber turbine training network layer S05, the output parameters of the air system network training layer S2, the output parameters of the lubricating oil system training network layer S3, and the high-pressure turbine parameters. The output parameters serve as one of the input parameters of the low-pressure turbine training network layer S07. S07: S07 is the low-pressure turbine training network layer; its input parameters are the output parameters of the high-pressure turbine training network layer S06, the output parameters of the lubricating oil system training network layer S3, the output parameters of the air system network training layer S2, and the low-pressure turbine parameters. The output parameters serve as one of the input parameters of the mixing chamber training network layer S08; S08: S08 is the mixing chamber training network layer; Its input parameters are the output parameters of the low-pressure turbine training network layer S07, the output parameters of the outer bypass network layer S1, and the mixing chamber parameters. The output parameters serve as one of the input parameters of the afterburner training network layer S09. S09: S09 is the afterburner training network layer; its input parameters are the output parameters of the mixing chamber training network layer S08 and the afterburner parameters. The output parameters serve as one of the input parameters of the nozzle training network layer S10; S10: S10 is the nozzle training network layer; Its input parameters are the output parameters of the afterburner training network layer S08, the output parameters of the control system training network layer S4, and the nozzle parameters. The output parameters are finally transmitted to the parameter output layer S11 through the multi-layer neural training network layer. S11: S11 is the final parameter output layer of the network; the output parameters are mainly set to the main output performance indicators of the aircraft engine; S1: S1 is the duct network training layer; its input parameters are the output parameters of the fan training network layer S03 and the internal parameters of the duct system. The output parameters serve as the input parameters of the mixing chamber network training layer S08; S2: S2 is the air system network training layer; its input parameters are the output parameters of the compressor training network layer S04 and the internal parameters of the air system. The output parameters serve as the input parameters of the turbine network training layers S06 and S07; S3: S3 is the oil system network training layer; its input parameters are the internal parameters of the oil system, and its output parameters serve as the input parameters of S03, S04, S06 and S07; S4: S4 is the control system network training layer; its input parameters are the internal parameters of the control system, and the output parameters serve as the input parameters of S03-S05 and S10; the compressor components are split into multi-stage blade elements, and a corresponding element network is established for each blade stage. According to the inter-stage measurement parameters in the compressor test, they are mapped one by one to the corresponding blade stage. The input of the element network corresponding to each blade stage includes the output characteristic parameters and corresponding measurement characteristic parameters of the previous level element sub-network. The refined component sub-network is established to form an element-level digital engineering model of the entire aircraft engine.

2. The digital engineering model construction method according to claim 1, characterized in that: Divide the structure of the physical object corresponding to the model to be constructed into: systems and components from large to small.

3. The digital engineering model construction method according to claim 1, characterized in that: Divide the structure of the physical object corresponding to the model to be constructed into: systems, parts, and components from large to small.

4. The digital engineering model construction method according to claim 3, characterized in that: Based on the data source, the physical measurement parameters corresponding to the model to be constructed are classified into the corresponding parts, systems or components according to the measurement location.

5. The digital engineering model construction method according to claim 4, characterized in that: Feature screening is performed according to the purpose of digital model construction to remove redundant parameter variables and reduce complexity.

6. The digital engineering model construction method according to claim 5, characterized in that: Based on the component-system-level digital engineering model of the whole machine, the test data of the component or system is refined, and the processing object is transferred from the component and system to the element. The element network is connected according to the component composition structure and working matching relationship to form a refined component system sub-network, thereby obtaining the component-level digital engineering model of the whole machine.

7. The digital engineering model construction method according to claim 6, characterized in that: Based on the component-level digital engineering model of the whole machine, according to the data source, the physical measurement parameters corresponding to the model to be built are classified into the corresponding components according to the measurement position, and the features are screened according to the purpose of the digital model to remove redundant parameter variables and reduce the complexity; then, based on the physical measurement connection structure or workflow corresponding to the model to be built, the corresponding input parameters are activated in batches in a time series.

8. Application of the digital engineering model construction method according to any one of claims 1 to 7 in constructing a digital engineering model of an aerospace engine.

Images