Parameter optimization method and device based on distributed artificial intelligence hybrid modeling
By using distributed artificial intelligence hybrid modeling, combining neural networks and mechanistic models, the modeling challenge of complex coupling relationships between components in aero-engine systems has been solved, achieving efficient and accurate multi-condition parameter optimization and improving the efficiency and accuracy of aero-engine design optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- AECC SICHUAN GAS TURBINE RES INST
- Filing Date
- 2026-01-16
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies struggle to effectively handle nonlinear coupling relationships between components and optimize parameters under multiple operating conditions in aero-engine systems, resulting in bottlenecks in efficiency and accuracy for traditional modeling methods.
A distributed artificial intelligence hybrid modeling approach is adopted, which deeply integrates neural network layers and mechanistic models through distributed AI entities and combines reinforcement learning training strategies to construct a distributed optimization model, thereby realizing the modeling of complex coupling relationships between components and the optimization of parameters under multiple operating conditions.
It shortens the optimization cycle, improves modeling efficiency and accuracy, can complete initial training within 30 minutes, achieves global optimization under multiple working conditions, and avoids the long parameter tuning process of traditional methods.
Smart Images
Figure CN122014423A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of engine optimization technology, and in particular to a parameter optimization method and apparatus based on distributed artificial intelligence hybrid modeling. Background Technology
[0002] In aero-engine systems, how to construct a distributed artificial intelligence (AI) architecture to handle complex input-output relationships at the component level, prediction, and system-level coupling optimization problems? Specifically, this involves using AI agents to learn nonlinear input-output mapping relationships in some key components that cannot be accurately described by traditional formulas. At the same time, it involves establishing parameter optimization algorithms that can handle the complex coupling effects of preceding components on subsequent components, and achieving global parameter optimization under various operating conditions. This ensures that an optimal parameter combination that can adapt to multiple operating conditions is found, thereby solving the technical bottlenecks faced by traditional engine modeling methods in handling nonlinear coupling relationships between components and multi-condition parameter optimization. Summary of the Invention
[0003] In view of this, this disclosure proposes a parameter optimization method and apparatus based on distributed artificial intelligence hybrid modeling.
[0004] According to one aspect of this disclosure, a parameter optimization method based on distributed artificial intelligence hybrid modeling is provided. The method includes: acquiring a target optimization task for a target engine system, the target engine system including multiple components; determining a target adjustment direction for the target engine system based on the target optimization task; inputting the target adjustment direction into an optimization model to obtain an optimization result, the optimization result including parameter adjustment amounts of components required to meet the target adjustment direction, wherein the optimization model includes multiple distributed artificial intelligence agents, each distributed AI agent corresponding to one or more components, each distributed AI agent being used to optimize the parameters of the corresponding component, and the neural network layer in the distributed AI agent corresponding to an AI model or a mechanism model.
[0005] Thus, this distributed artificial intelligence hybrid modeling parameter optimization method deeply integrates neural network layers with mechanistic models, resulting in an optimized model that combines data fitting ability with physical interpretability; distributed parallel computing can shorten the optimization cycle under complex operating conditions; and reinforcement learning training strategies are used to quickly map the target adjustment direction in response to the dynamic changes of the target optimization task. This distributed optimization paradigm, which deeply couples physical mechanisms with artificial intelligence, redefines the control architecture of aero-engines.
[0006] In one possible implementation, the method further includes at least one of the following: corresponding each component in the target engine system that moves at the same rotational speed to the same distributed AI body; corresponding each component in the target engine system that is on the same energy conversion path to the same distributed AI body; corresponding each component in the target engine system that is on the same fault propagation range to the same distributed AI body; and corresponding each component in the target engine system that is synchronously controlled to the same distributed AI body.
[0007] In this way, the above four division methods can simultaneously satisfy physical correlation and control logic independence, reflect the coupling relationship of components in mechanical structure, and ensure that the AI can independently complete the closed-loop control of perception-decision-execution.
[0008] In one possible implementation, the method further includes: constructing a corresponding operating condition neural network for each operating state of the target engine system, each operating condition neural network including multiple component layers, each component layer corresponding to different components, and the optimization model including the operating condition neural network for each operating state.
[0009] This design not only fully preserves the physical connections and energy transfer characteristics between engine components, but also lays a solid mathematical foundation for the unified expression and collaborative optimization of heterogeneous models. Each operating condition neural network achieves accurate modeling of physical laws under specific operating conditions through parameter coupling at the component layer. The operating condition neural network shares feature extraction layers through transfer learning, reducing the need for new operating condition data.
[0010] In one possible implementation, the method further includes: determining each training sample based on the historical operating data of each component of the target engine system to form a training sample set, wherein each training sample includes operating parameters input to the component from the outside and test result data of the component operation; using the training sample set to train the initial optimization model until the trained optimization model meets the preset conditions, and outputting the trained optimization model.
[0011] In this way, by constructing a training sample set and training an optimization model using historical operating data, a data-driven intelligent foundation is provided for engine parameter optimization. This data-driven training method enables the optimization model to truly have the ability to "learn from experience".
[0012] In one possible implementation, the optimization model is trained using the training sample set until the trained optimization model meets preset conditions, and the trained optimization model is output. This includes: determining the maximum value term for each working state based on the maximum value of the output of each component layer in the corresponding working state neural network, and determining the output values of all component layers in the corresponding working state neural network, and determining the weighted sum term for the corresponding working state based on all output values and a preset dynamic balance coefficient; determining the first loss for each working state based on the maximum value term and the weighted sum term, and using the sum of the first losses for all working states as the loss function value used during optimization model training; and outputting the currently trained optimization model if the loss function value is less than a preset threshold.
[0013] In this way, by constructing a multi-condition loss function by fusing the maximum value term and the weighted sum term, the optimization model training achieves a dual-objective optimization of "extreme constraint-global balance". The maximum value of the output of each component layer is used as a constraint term to ensure that the training process strictly follows the physical limit boundary. The dynamic balance coefficient realizes differentiated weighting of the contribution of different component layers, avoiding the "local optimum trap". The composite loss function of the maximum value term and the weighted sum term makes the gradient direction clearer, and the composite loss function reduces the overall convergence time and significantly reduces the consumption of computational resources.
[0014] In one possible implementation, the preset condition further includes that the number of iterations meets a preset iteration threshold.
[0015] According to another aspect of this disclosure, a parameter optimization device based on distributed artificial intelligence hybrid modeling is provided. The device includes: an acquisition module for acquiring a target optimization task for a target engine system, the target engine system including multiple components; a determination module for determining a target adjustment direction for the target engine system based on the target optimization task; and an optimization module for inputting the target adjustment direction into an optimization model to obtain an optimization result, the optimization result including parameter adjustment amounts of components required to satisfy the target adjustment direction. The optimization model includes multiple distributed artificial intelligence (AI) entities, each distributed AI entity corresponding to one or more components, each distributed AI entity being used to optimize the parameters of the corresponding component, and the neural network layers in the distributed AI entities corresponding to AI models or mechanistic models.
[0016] In this way, by deeply integrating neural network layers with mechanistic models, the optimized model has both data fitting ability and physical interpretability; distributed parallel computing can shorten the optimization cycle under complex operating conditions; and reinforcement learning training strategies can be used to quickly map the target adjustment direction in response to the dynamic changes of the target optimization task. This distributed optimization paradigm, which deeply couples physical mechanisms with artificial intelligence, redefines the control architecture of aero-engines.
[0017] In one possible implementation, the device further includes a partitioning module for performing at least one of the following: mapping components in the target engine system that move at the same rotational speed to the same distributed AI body; mapping components in the target engine system that are on the same energy conversion path to the same distributed AI body; mapping components in the target engine system that are on the same fault propagation range to the same distributed AI body; and mapping components in the target engine system that are synchronously controlled to the same distributed AI body.
[0018] In this way, the above four division methods can simultaneously satisfy physical correlation and control logic independence, reflect the coupling relationship of components in mechanical structure, and ensure that the AI can independently complete the closed-loop control of perception-decision-execution.
[0019] In one possible implementation, the device further includes a construction module for: constructing a corresponding operating condition neural network for each operating state of the target engine system, each operating condition neural network including multiple component layers, each component layer corresponding to different components, and the optimization model including the operating condition neural network for each operating state.
[0020] This design not only fully preserves the physical connections and energy transfer characteristics between engine components, but also lays a solid mathematical foundation for the unified expression and collaborative optimization of heterogeneous models. Each operating condition neural network achieves accurate modeling of physical laws under specific operating conditions through parameter coupling at the component layer. The operating condition neural network shares feature extraction layers through transfer learning, reducing the need for new operating condition data.
[0021] In one possible implementation, the device further includes a training module, configured to: determine each training sample based on historical operating data of each component of the target engine system, forming a training sample set, wherein each training sample includes operating parameters input to the component from the outside and test result data of the component operation; train the initial optimization model using the training sample set until the trained optimization model meets preset conditions, and output the trained optimization model.
[0022] In this way, by constructing a training sample set and training an optimization model using historical operating data, a data-driven intelligent foundation is provided for engine parameter optimization. This data-driven training method enables the optimization model to truly have the ability to "learn from experience".
[0023] In one possible implementation, the optimization model is trained using the training sample set until the trained optimization model meets preset conditions, and the trained optimization model is output. This includes: determining the maximum value term for each working state based on the maximum value of the output of each component layer in the corresponding working state neural network, and determining the output values of all component layers in the corresponding working state neural network, and determining the weighted sum term for the corresponding working state based on all output values and a preset dynamic balance coefficient; determining the first loss for each working state based on the maximum value term and the weighted sum term, and using the sum of the first losses for all working states as the loss function value used during optimization model training; and outputting the currently trained optimization model if the loss function value is less than a preset threshold.
[0024] In this way, by constructing a multi-condition loss function by fusing the maximum value term and the weighted sum term, the optimization model training achieves a dual-objective optimization of "extreme constraint-global balance". The maximum value of the output of each component layer is used as a constraint term to ensure that the training process strictly follows the physical limit boundary. The dynamic balance coefficient realizes differentiated weighting of the contribution of different component layers, avoiding the "local optimum trap". The composite loss function of the maximum value term and the weighted sum term makes the gradient direction clearer, and the composite loss function reduces the overall convergence time and significantly reduces the consumption of computational resources.
[0025] In one possible implementation, the preset condition further includes that the number of iterations meets a preset iteration threshold.
[0026] According to another aspect of this disclosure, a parameter optimization apparatus based on distributed artificial intelligence hybrid modeling is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.
[0027] According to another aspect of this disclosure, a non-volatile computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described method.
[0028] According to another aspect of this disclosure, a computer program product is provided, including a computer program or a non-volatile computer-readable storage medium carrying the computer program, wherein the computer program, when executed by a processor, implements the steps of the above-described method.
[0029] This disclosure obtains a target optimization task for a target engine system, which includes multiple components. Based on the target optimization task, a target adjustment direction for the target engine system is determined. This target adjustment direction is input into an optimization model to obtain optimization results. The optimization results include parameter adjustments for the components required to meet the target adjustment direction. The optimization model includes multiple distributed artificial intelligence (AI) agents, each corresponding to one or more components. Each distributed AI agent optimizes the parameters of its corresponding component. The neural network layers in the distributed AI agents correspond to AI models or mechanistic models. By deeply integrating the neural network layers with the mechanistic model, the optimization model possesses both data fitting capabilities and physical interpretability. Distributed parallel computing shortens the optimization cycle for complex operating conditions. For the dynamic changes in the target optimization task, a reinforcement learning training strategy is used to achieve rapid mapping of the target adjustment direction, effectively solving the modeling problem of complex coupling relationships between engine components. Through the deep integration of physical prior knowledge and neural networks, it can handle component characteristics with analytical formulas and learn complex nonlinear mapping relationships, significantly improving the efficiency and accuracy of engine modeling and providing an efficient and reliable solution for the design optimization of aero-engines.
[0030] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description
[0031] The accompanying drawings, which are included in and form part of this specification, illustrate exemplary embodiments, features, and aspects of this disclosure together with the specification and serve to explain the principles of this disclosure.
[0032] Figure 1 A flowchart illustrating the parameter optimization method provided in an embodiment of this disclosure is shown.
[0033] Figure 2 A schematic diagram of the parameter optimization method provided in the embodiments of this disclosure is shown.
[0034] Figure 3 A schematic diagram of a neural network provided in an embodiment of this disclosure is shown.
[0035] Figure 4 A block diagram of the parameter optimization apparatus provided in an embodiment of this disclosure is shown. Detailed Implementation
[0036] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.
[0037] As used herein, the terms “comprising,” “including,” “having,” or variations thereof are open-ended and include one or more of the stated features, integrals, elements, steps, components, or functions, but do not exclude the presence or addition of one or more other features, integrals, elements, steps, components, functions, or groups thereof.
[0038] When an element is referred to as “connected,” “coupled,” “responding,” or a variation thereof relative to another element, it may be directly connected, coupled, or responding to another element, or there may be an intermediate element present.
[0039] Although the terms first, second, third, etc., may be used herein to describe various elements / operations, these elements / operations should not be limited by these terms. These terms are only used to distinguish one element / operation from another. Therefore, without departing from the teachings of the inventive concept, a first element / operation in some embodiments may be referred to as a second element / operation in other embodiments.
[0040] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.
[0041] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.
[0042] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant regions.
[0043] To address the aforementioned technical issues, this disclosure provides a parameter optimization method based on distributed artificial intelligence hybrid modeling. By employing a distributed AI architecture, it avoids the need for extensive data and lengthy parameter tuning required by traditional methods. This method shortens initial training to within 30 minutes and incremental training to within 2 minutes. Simultaneously, it achieves a global optimization effect where a single set of parameters adapts to multiple operating conditions, eliminating the need for individual parameter adjustments for different conditions. This effectively solves the modeling challenge of complex coupling relationships between engine components. Through the deep integration of physical prior knowledge and neural networks, it can handle component characteristics with analytical formulas and learn complex nonlinear mapping relationships, significantly improving the efficiency and accuracy of engine modeling and providing an efficient and reliable solution for aero-engine design optimization.
[0044] Now combined Figure 1 and Figure 3 This section provides an illustrative explanation of the parameter optimization method. This method can be executed on a server or terminal. Figure 1 As shown, this method may include the following steps S101 to S103.
[0045] Step S101: Obtain the target optimization task for the target engine system, which includes multiple components.
[0046] The target engine system is one that is currently of interest. An engine system may include multiple components such as a crankshaft and connecting rod mechanism, valve train, fuel supply mechanism, cooling system, lubrication system, and starting mechanism.
[0047] Step S102: Determine the target adjustment direction for the target engine system based on the target optimization task.
[0048] The target adjustment direction indicates a general direction for engine adjustment. For example, maximum thrust or minimum fuel consumption.
[0049] Step S103: Input the target adjustment direction into the optimization model to obtain the optimization result. The optimization result includes the parameter adjustment amount of the component required to meet the target adjustment direction. The optimization model includes multiple distributed artificial intelligence (AI) entities, each distributed AI entity corresponds to one or more components, and each distributed AI entity is used to optimize the parameters of the corresponding component. The neural network layer in the distributed AI entity corresponds to an AI model or a mechanism model.
[0050] like Figure 2As shown, the number of distributed AI agents can be set according to actual conditions. Mechanism models are mathematical models constructed based on scientific principles such as physics and chemistry. They explain phenomena or predict behavior by describing the system's internal mechanisms (such as the laws of conservation of mass / energy / momentum), and the parameters of mechanistic models have clear physical meanings. AI models are a type of artificial intelligence model with a large number of parameters, constructed from artificial neural networks.
[0051] This distributed AI hybrid modeling parameter optimization method deeply integrates neural network layers with mechanistic models, resulting in an optimized model that combines data fitting capabilities with physical interpretability. For differentiable components such as compressors, the AI agent can employ Long Short-Term Memory (LSTM) networks to capture the nonlinear characteristics of the flow-pressure ratio, keeping the fitting error within the expected range. Distributed parallel computing shortens the optimization cycle for complex operating conditions. For dynamic changes in the target optimization task (such as switching from "maximum thrust" to "minimum fuel consumption"), reinforcement learning training strategies enable rapid mapping of the target adjustment direction. This distributed optimization paradigm, deeply coupling physical mechanisms with artificial intelligence, redefines the control architecture of aero-engines.
[0052] The method may further include at least one of the following: Assigning components moving at the same speed in the target engine system to the same distributed AI body; for example, integrating components moving at the same speed, such as the high-pressure turbine rotor and compressor, into an AI body, and achieving "synchronous perception-synchronous decision-making" of the mechanical system through closed-loop fusion of rotor dynamics model and real-time vibration data; Assigning components in the same energy conversion path in the target engine system to the same distributed AI body; dividing AI bodies according to energy conversion paths ensures optimal efficiency at each stage. Taking a steam turbine as an example, in the steam expansion process from the high-pressure cylinder to the low-pressure cylinder, each stage of rotor blades and corresponding nozzles constitute an energy conversion unit. Dividing adjacent 2-3 stages of components into one AI body enables adaptive adjustment of inter-stage pressure matching; Assigning components moving at the same speed in the target engine system to the same distributed AI body. In a system, components within the same fault propagation range correspond to the same distributed AI body. For example, the lubrication system of an aero-engine includes an oil tank, pump, oil filter, and bearings. Faults (such as lubricating oil leakage) are usually confined to a closed loop, making them suitable as independent AI bodies. Thus, based on fault mode and effects analysis, grouping components whose faults do not propagate across boundaries into the same AI body can improve the system's fault tolerance. Similarly, components in the target engine system that are synchronously controlled correspond to the same distributed AI body. For example, in a turbocharger system, the mechanical connection between the turbine and compressor requires them to be controlled synchronously. However, the timing requirements for compressor surge suppression (millisecond level) and turbine blade life management (minute level) differ greatly. Dividing the system into corresponding AI bodies based on the time sensitivity of the control tasks can balance system response speed and computing resources.
[0053] In this way, the above four division methods can simultaneously satisfy physical correlation and control logic independence, reflect the coupling relationship of components in mechanical structure, and ensure that the AI can independently complete the closed-loop control of perception-decision-execution.
[0054] The method may further include: constructing a corresponding operating condition neural network for each operating state or condition of the target engine system, wherein one AI entity corresponds to one or more operating condition neural networks, each operating condition neural network includes multiple component layers, each component layer corresponds to different components, and the optimization model includes the operating condition neural network for each operating state.
[0055] Specifically, the engine system is reconstructed as a multi-layer neural network structure, where each engine component corresponds to a network layer, such as... Figure 3 As shown, the output of the previous component layer and the parameters of the current component layer together determine the output of the current component layer. Simultaneously, an independent neural network is constructed for each operating condition, and a unified mathematical expression is used. Establish a complete optimization model, in which, This indicates the initial position and status information of each component, with components numbered 1, 2, ... , Indicates component layer parameters, , Indicates the first The component layer parameters of each component. Under this architecture, the functions of each layer... It can be flexibly configured as a mechanistic model or an AI model according to actual needs, forming a hybrid modeling architecture. For the non-differentiable component layer, a Gaussian model is used as a surrogate model, and an input-output mapping is constructed based on historical data. The output formula of each component layer can be expressed as follows: .in, Indicates the first The output of each component layer Indicates the first Functions of each component layer Indicates the first The output of each component layer Indicates the first The network parameters of each component layer are presented. This method constructs a distributed AI hybrid modeling architecture for aero-engines, reconstructing the traditional engine system into a multi-layer neural network structure. Combined with a dual optimization mechanism of neural network optimization and Bayesian optimization, it achieves deep integration of the mechanistic model and the AI model. Each engine component corresponds to a network layer, maintaining physical connections while laying the foundation for unified optimization of heterogeneous models.
[0056] This design not only fully preserves the physical connections and energy transfer characteristics between engine components, but also lays a solid mathematical foundation for the unified expression and collaborative optimization of heterogeneous models. Each operating condition neural network achieves accurate modeling of physical laws under specific operating conditions through parameter coupling at the component layer. The operating condition neural network shares feature extraction layers through transfer learning, reducing the need for new operating condition data.
[0057] The method may further include: determining training samples based on historical operating data of each component of the target engine system to form a training sample set, wherein each training sample includes externally input operating parameters to the component and experimental result data of component operation; training an initial optimization model using the training sample set until the trained optimization model meets preset conditions, and outputting the trained optimization model. In this way, constructing a training sample set and training an optimization model using historical operating data provides a data-driven intelligent foundation for engine parameter optimization. This data-driven training method enables the optimization model to truly possess the ability to "learn from experience."
[0058] The process of training the optimization model using the training sample set until the trained optimization model meets preset conditions and outputting the trained optimization model includes: determining the maximum value term for each working state based on the maximum value output of each component layer in the corresponding working condition neural network, for example, as described below. Furthermore, it determines the output values of all component layers in the corresponding working condition neural network, and determines the weighted sum term under the corresponding working condition based on all output values and preset dynamic balance coefficients, such as the following... The first loss for each working state is determined based on the maximum value and weighted sum term, and the sum of the first losses for all working states is used as the loss function value for training the optimization model, as shown in the following text. If the loss function value is less than a preset threshold, the currently trained optimized model is output. In this way, by fusing the maximum value term and the weighted sum term to construct a multi-condition loss function, the optimized model training achieves a dual-objective optimization of "extreme constraint-global balance." The maximum value output of each component layer is used as a constraint term to ensure that the training process strictly follows the physical limit boundary; the dynamic balance coefficient achieves differentiated weighting of the contributions of different component layers, avoiding the "local optimum trap"; the composite loss function of the maximum value term and the weighted sum term makes the gradient direction clearer, and the composite loss function reduces the overall convergence time, significantly reducing computational resource consumption.
[0059] To address the complexity of multi-condition optimization, embodiments of this disclosure construct a unified loss function that balances critical performance with overall performance. This loss function employs a dual constraint mechanism. A maximum value term ensures that key components maintain excellent performance under various operating conditions, avoiding local performance bottlenecks. Simultaneously, a weighted summation term guarantees the overall system performance balance, preventing over-optimization of some components at the expense of others. Finally, the average output value across multiple operating conditions can be used as the final loss function output. This design achieves a fundamental shift from single-condition local optimization to multi-condition global optimization. This method introduces an adaptive parameter adjustment mechanism, using dynamic balancing coefficients to automate the optimization process. This mechanism automatically adjusts the weighting relationship between the outputs of key components and the overall system output according to the optimization progress. In the early stages of optimization, it prioritizes rapid improvement of overall performance, while gradually strengthening the fine-tuning of key components in the later stages, ensuring both good convergence stability and high-quality optimization solutions throughout the optimization process.
[0060] In addition, the preset conditions may also include the number of iterations meeting a preset iteration threshold. That is, when the number of iterations exceeds the preset number, training stops and the model is output.
[0061] For example, taking the compressor component as an example, the data input to the optimization model may include n-speed, p25 inlet pressure, T25 inlet temperature, alpha2 guide vane angle, p3 outlet pressure, Wbleed bleed air flow rate, and War water-air ratio. External operable data includes: compressor speed-converted flow rate scaling factor, compressor speed-pressure ratio characteristic scaling factor, compressor speed-efficiency characteristic scaling factor, other characteristic modification factors, and compressor. The specific calculation process of the AI body can be found in Table 1. The optimization results output by the optimization model may include T3 outlet temperature, pwrCmp compressor power, TBld2LPT bleed air temperature, and w21 compressor flow rate.
[0062] Table 1. Optimization process of the optimization model for compressor components.
[0063] To effectively address the technical challenge of coexisting differentiable and non-differentiable components in hybrid modeling, this method designs a unique alternating optimization strategy. This strategy decomposes the optimization process into two coordinated stages based on the differentiability of the components. In the differentiable component optimization stage, the network parameters of all non-differentiable components are fixed, and a multi-condition backpropagation algorithm is used to efficiently optimize the gradients of differentiable components, including the differentiable physical model, fully leveraging the advantages of gradient information in parameter space navigation. In the non-differentiable component optimization stage, the network parameters of differentiable components are fixed, and a Gaussian process-based Bayesian optimization method is used to intelligently optimize the non-differentiable components without gradients, effectively balancing exploration and utilization by constructing a surrogate model.
[0064] During each model iteration, this method also trains the initial optimized model based on preset first and second optimization strategies until the trained optimized model meets preset conditions, thus obtaining the trained optimized model. Specifically, the first optimization strategy fixes the network parameters of the component layers of all non-differentiable components and uses a multi-working-state backpropagation algorithm to perform gradient optimization on the differentiable components. Specifically, the second optimization strategy fixes the network parameters of the component layers of all differentiable components and uses a Bayesian optimization algorithm based on Gaussian processes to perform gradient-free optimization on the non-differentiable components.
[0065] The two optimization phases are organically coordinated through an interactive iterative mechanism. In each iteration, the system alternately optimizes differentiable and non-differentiable components. The optimization progress is judged by monitoring the convergence characteristics of the multi-condition loss function (see below). The optimization process terminates when the loss change between consecutive iterations is less than a preset threshold or the maximum number of iterations is reached. This alternating optimization mechanism not only makes full use of the optimization characteristics of different types of components, but also ensures that the entire hybrid system can converge to the true global optimum through global coordination.
[0066] This disclosure discloses an embodiment that reconstructs the engine system into a neural network, achieving an organic integration of the mechanistic model and the AI model, and employs an alternating optimization strategy to solve the unified optimization problem of heterogeneous components. This method overcomes the limitations of traditional single modeling paradigms, coordinating the processing of different types of model components within a unified computational framework, and providing a novel technical path for multi-condition optimization of complex engineering systems.
[0067] This disclosure also provides a parameter optimization device based on distributed artificial intelligence hybrid modeling. The device includes: an acquisition module for acquiring a target optimization task for a target engine system, the target engine system including multiple components; a determination module for determining a target adjustment direction for the target engine system based on the target optimization task; and an optimization module for inputting the target adjustment direction into an optimization model to obtain an optimization result. The optimization result includes parameter adjustment amounts for components required to meet the target adjustment direction. The optimization model includes multiple distributed artificial intelligence (AI) entities, each corresponding to one or more components. Each distributed AI entity is used to optimize the parameters of the corresponding component. The neural network layer in the distributed AI entity corresponds to an AI model or a mechanism model.
[0068] In this way, by deeply integrating neural network layers with mechanistic models, the optimized model has both data fitting ability and physical interpretability; distributed parallel computing can shorten the optimization cycle under complex operating conditions; and reinforcement learning training strategies can be used to quickly map the target adjustment direction in response to the dynamic changes of the target optimization task. This distributed optimization paradigm, which deeply couples physical mechanisms with artificial intelligence, redefines the control architecture of aero-engines.
[0069] In one possible implementation, the device further includes a partitioning module for performing at least one of the following: mapping components in the target engine system that move at the same rotational speed to the same distributed AI body; mapping components in the target engine system that are on the same energy conversion path to the same distributed AI body; mapping components in the target engine system that are on the same fault propagation range to the same distributed AI body; and mapping components in the target engine system that are synchronously controlled to the same distributed AI body.
[0070] In this way, the above four division methods can simultaneously satisfy physical correlation and control logic independence, reflect the coupling relationship of components in mechanical structure, and ensure that the AI can independently complete the closed-loop control of perception-decision-execution.
[0071] In one possible implementation, the device further includes a construction module for: constructing a corresponding operating condition neural network for each operating state of the target engine system, each operating condition neural network including multiple component layers, each component layer corresponding to different components, and the optimization model including the operating condition neural network for each operating state.
[0072] This design not only fully preserves the physical connections and energy transfer characteristics between engine components, but also lays a solid mathematical foundation for the unified expression and collaborative optimization of heterogeneous models. Each operating condition neural network achieves accurate modeling of physical laws under specific operating conditions through parameter coupling at the component layer. The operating condition neural network shares feature extraction layers through transfer learning, reducing the need for new operating condition data.
[0073] In one possible implementation, the device further includes a training module, configured to: determine each training sample based on historical operating data of each component of the target engine system, forming a training sample set, wherein each training sample includes operating parameters input to the component from the outside and test result data of the component operation; train the initial optimization model using the training sample set until the trained optimization model meets preset conditions, and output the trained optimization model.
[0074] In this way, by constructing a training sample set and training an optimization model using historical operating data, a data-driven intelligent foundation is provided for engine parameter optimization. This data-driven training method enables the optimization model to truly have the ability to "learn from experience".
[0075] In one possible implementation, the optimization model is trained using the training sample set until the trained optimization model meets preset conditions, and the trained optimization model is output. This includes: determining the maximum value term for each working state based on the maximum value of the output of each component layer in the corresponding working state neural network, and determining the output values of all component layers in the corresponding working state neural network, and determining the weighted sum term for the corresponding working state based on all output values and a preset dynamic balance coefficient; determining the first loss for each working state based on the maximum value term and the weighted sum term, and using the sum of the first losses for all working states as the loss function value used during optimization model training; and outputting the currently trained optimization model if the loss function value is less than a preset threshold.
[0076] In this way, by constructing a multi-condition loss function by fusing the maximum value term and the weighted sum term, the optimization model training achieves a dual-objective optimization of "extreme constraint-global balance". The maximum value of the output of each component layer is used as a constraint term to ensure that the training process strictly follows the physical limit boundary. The dynamic balance coefficient realizes differentiated weighting of the contribution of different component layers, avoiding the "local optimum trap". The composite loss function of the maximum value term and the weighted sum term makes the gradient direction clearer, and the composite loss function reduces the overall convergence time and significantly reduces the consumption of computational resources.
[0077] In one possible implementation, the preset condition further includes that the number of iterations meets a preset iteration threshold.
[0078] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.
[0079] This disclosure also provides a parameter optimization device based on distributed artificial intelligence hybrid modeling, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.
[0080] This disclosure also provides a non-volatile computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.
[0081] This disclosure also provides a computer program product, including a computer program or a non-volatile computer-readable storage medium carrying the computer program, wherein the computer program, when executed by a processor, implements the steps of the above method.
[0082] Figure 4A block diagram of a parameter optimization apparatus provided in an embodiment of this disclosure is shown. For example, apparatus 1900 may be provided as a server or terminal device. (Refer to...) Figure 4 The apparatus 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions, such as application programs, that can be executed by the processing component 1922. The application programs stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 1922 is configured to execute instructions to perform the methods described above.
[0083] Device 1900 may also include a power supply component 1926 configured to perform power management of device 1900, a wired or wireless network interface 1950 configured to connect device 1900 to a network, and an input / output interface 1958 (I / O interface). Device 1900 can operate on an operating system, such as Windows Server, stored in memory 1932. TM macOS X TM Unix TM Linux TM FreeBSD TM Or similar.
[0084] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions that can be executed by a processing component 1922 of the device 1900 to perform the above-described method.
[0085] Computer-readable storage media can be tangible devices capable of holding and storing programs / instructions used by instruction execution devices. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.
[0086] The computer program (or computer-readable program instructions) described herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage medium in the respective computing / processing device.
[0087] The computer program (or computer program instructions) used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions to implement various aspects of this disclosure.
[0088] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.
[0089] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.
[0090] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.
[0091] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0092] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.
Claims
1. A parameter optimization method based on distributed artificial intelligence hybrid modeling, characterized in that, The method includes: Obtain a target optimization task for a target engine system, which includes multiple components; Based on the target optimization task, the target adjustment direction for the target engine system is determined; The target adjustment direction is input into the optimization model to obtain the optimization result. The optimization result includes the parameter adjustment amount of the component required to meet the target adjustment direction. The optimization model includes multiple distributed artificial intelligence (AI) agents, each distributed AI agent corresponds to one or more components, and each distributed AI agent is used to optimize the parameters of the corresponding component. The neural network layer in the distributed AI agent corresponds to an AI model or a mechanism model.
2. The method according to claim 1, characterized in that, The method further includes at least one of the following: The components of the target engine system that move at the same speed are associated with the same distributed AI body; The components in the target engine system that are on the same energy conversion path are mapped to the same distributed AI entity; The components in the target engine system that are within the same fault propagation range are associated with the same distributed AI entity; The components of the target engine system that are synchronously controlled are mapped to the same distributed AI entity.
3. The method according to claim 1, characterized in that, The method further includes: A corresponding operating condition neural network is constructed for each operating state of the target engine system. Each operating condition neural network includes multiple component layers, and each component layer corresponds to a different component. The optimization model includes the operating condition neural network for each operating state.
4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: Each training sample is determined based on the historical operating data of each component of the target engine system, forming a training sample set. Each training sample includes the operating parameters input to the component from the outside and the test result data of the component operation. The initial optimization model is trained using the training sample set until the trained optimization model meets the preset conditions, and the trained optimization model is output.
5. The method according to claim 4, characterized in that, The optimization model is trained using the training sample set until it meets preset conditions, and the trained optimization model is output, including: In each working state, the maximum value term in the corresponding working state is determined based on the maximum value of the output of each component layer in the corresponding working state neural network, and the output value of all component layers in the corresponding working state neural network is determined, and the weighted sum term in the corresponding working state is determined based on all output values and the preset dynamic balance coefficient. The first loss for each working state is determined based on the maximum value and weighted sum for each working state, and the sum of the first losses for all working states is used as the loss function value used when training the optimization model. If the loss function value is less than a preset threshold, the currently trained optimized model is output.
6. The method according to claim 4, characterized in that, The preset conditions also include that the number of iterations meets a preset iteration threshold.
7. A parameter optimization device based on distributed artificial intelligence hybrid modeling, characterized in that, The device includes: An acquisition module is used to acquire a target optimization task for a target engine system, the target engine system comprising multiple components; The determination module is used to determine the target adjustment direction for the target engine system based on the target optimization task. An optimization module is used to input the target adjustment direction into an optimization model to obtain an optimization result. The optimization result includes the parameter adjustment amount of the components required to meet the target adjustment direction. The optimization model includes multiple distributed artificial intelligence (AI) entities, each corresponding to one or more components. Each distributed AI entity is used to optimize the parameters of the corresponding component. The neural network layer in the distributed AI entity corresponds to an AI model or a mechanism model.
8. A parameter optimization device based on distributed artificial intelligence hybrid modeling, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.
9. A non-volatile computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program, or a non-volatile computer-readable storage medium carrying a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.