Method, device, equipment, medium and product for optimizing combustion chamber of aero-engine

Abstract

This disclosure provides a method, apparatus, device, medium, and product for optimizing the combustor of an aero-engine. The method includes: determining target performance information corresponding to the combustor of the aero-engine; wherein the target performance information indicates the optimization objectives and constraints of the combustor; determining an optimization strategy between the target performance information and the configuration parameters of the combustor, and determining several initial configuration parameters based on the optimization strategy; iteratively calculating the initial configuration parameters using a target performance parameter prediction model to obtain specified configuration parameters; wherein the target performance parameter prediction model is a surrogate model that meets accuracy requirements; analyzing the specified configuration parameters to obtain the optimal specified configuration parameter, and determining this optimal specified configuration parameter as the target configuration parameter; evaluating the target configuration parameter based on the target performance information, and applying the target configuration parameter if the evaluation is successful.

Description

Technical Field

[0001] This disclosure generally relates to the field of aero-engine simulation technology, and in particular to a method, apparatus, equipment, medium and product for optimizing the combustion chamber of an aero-engine. Background Technology

[0002] Combustion chamber design for aero-engines is a core challenge in achieving high-performance, low-emission propulsion systems. Currently, the global aero-engine industry is accelerating its transformation towards digitalization and intelligence. However, deeply applying artificial intelligence to combustor design and development still faces challenges.

[0003] In related technologies, it is difficult to accurately build parameter association logic, resulting in insufficient rationality in the selection of configuration parameters. Furthermore, the models used for iterative calculations often fail to balance accuracy and efficiency. At the same time, the performance analysis and screening of specified configuration parameters after iteration lacks precision, and the evaluation and verification of target configuration parameters are difficult to fully meet the preset performance requirements. Consequently, the optimization efficiency is low, the results are inaccurate, and the high-performance design requirements of the combustion chamber cannot be efficiently met. Summary of the Invention

[0004] This disclosure addresses some deficiencies mentioned in the background art by providing a method, apparatus, device, medium, and product for optimizing the combustion chamber of an aero-engine.

[0005] In a first aspect, embodiments of this disclosure provide a method for optimizing the combustion chamber of an aero-engine, comprising:

[0006] Determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization objectives and constraints of the combustion chamber; An optimization strategy is determined between the target performance information and the configuration parameters of the combustion chamber, and several initial configuration parameters are determined based on the optimization strategy; The initial configuration parameters are iteratively calculated using a target performance parameter prediction model to obtain the specified configuration parameters; wherein, the target performance parameter prediction model is a surrogate model that meets the accuracy condition. The specified configuration parameters are analyzed to obtain the specified configuration parameters with optimal performance, and the specified configuration parameters with optimal performance are determined as the target configuration parameters; The target configuration parameters are evaluated based on the target performance information, and if the target configuration parameters pass the evaluation, the target configuration parameters are applied.

[0007] In one embodiment of the first aspect, before determining the optimization strategy between the target performance parameter and the configuration parameters of the combustion chamber, the target performance parameter prediction model is determined by: determining a training sample set; wherein the training sample set consists of several configuration parameters and specified performance parameters; The covariance matrix is ​​determined based on the training sample set using a one-dimensional active subspace; The training sample set is reduced in dimensionality based on the covariance matrix to obtain matrix eigenvalues ​​and eigenvectors. If the training sample set satisfies the dimensionality reduction condition based on the matrix eigenvalues, the eigenvectors are processed to obtain active variables; wherein, the active variables are used to indicate the influence of each configuration parameter in the training sample set on the specified performance parameter; If the first low-order surrogate model meets the simulation requirements based on the active variables, the first low-order surrogate model is determined as the initial performance parameter prediction model. Based on the specified performance parameters and the corresponding optimization objectives and constraints, the initial performance parameter prediction model is trained to obtain the target performance parameter prediction model.

[0008] In one embodiment of the first aspect, after obtaining the active variable, the method further includes: If the low-order surrogate model does not meet the simulation requirements based on the active variables, a high-order surrogate model is constructed using an adaptive high-precision modeling tool; wherein, the adaptive high-precision modeling tool includes adaptive Gaussian process regression and adaptive multilayer perceptron; If the higher-order surrogate model meets the simulation requirements based on the covariance matrix, the higher-order surrogate model is dimensionality reduced to obtain a second lower-order surrogate model, and the second lower-order surrogate model is determined as the initial performance parameter prediction model.

[0009] In one embodiment of the first aspect, the step of reducing the dimensionality of the high-order surrogate model to obtain a second low-order surrogate model, based on the covariance matrix and determining that the high-order surrogate model meets the simulation requirements, includes: If both the higher-order surrogate model obtained by the adaptive Gaussian process regression and the higher-order surrogate model obtained by the adaptive multilayer perceptron meet the simulation requirements, determine the accuracy of each higher-order surrogate model. The high-accuracy high-order proxy model is dimensionality reduced to obtain the second low-order proxy model.

[0010] In one embodiment of the first aspect, after constructing a high-order proxy model using an adaptive high-precision modeling tool, the method further includes: If the higher-order surrogate model does not meet the simulation requirements based on the covariance matrix, the training sample set is augmented using a data augmentation strategy to obtain an augmented data sample set. An enhanced high-order proxy model is constructed based on the enhanced data sample set using the adaptive high-precision modeling tool. If the enhanced high-order proxy model meets the simulation requirements, the enhanced high-order proxy model is dimensionality reduced to obtain a third low-order proxy model, and the third low-order proxy model is determined as the initial performance parameter prediction model.

[0011] In one embodiment of the first aspect, determining the training sample set includes: Determine the parameter range of the configuration parameters corresponding to the combustion chamber; Multiple initial configuration parameter sample sets are obtained by sampling based on the parameter range using the optimal Latin hypercube sampling method. Determine the overall potential energy of each of the initial configuration parameter sample sets, and determine the initial configuration parameter sample set corresponding to the minimum overall potential energy as the training configuration parameter sample set; Simulation calculations are performed on the training configuration parameter sample set to obtain the training performance parameter set corresponding to the training configuration parameter sample set; The training configuration parameter sample set and the training performance parameter set are determined as the training sample set.

[0012] In a second aspect, embodiments of this disclosure provide a combustion chamber optimization apparatus for an aero-engine, comprising: The first determining module is used to determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization target and constraints of the combustion chamber; The second determining module is used to determine an optimization strategy between the target performance information and the configuration parameters of the combustion chamber, and to determine several initial configuration parameters based on the optimization strategy. The calculation module is used to iteratively calculate the initial configuration parameters using a target performance parameter prediction model to obtain the specified configuration parameters; wherein the target performance parameter prediction model is a surrogate model that meets the accuracy condition. The analysis module is used to analyze the specified configuration parameters, obtain the specified configuration parameters with optimal performance, and determine the specified configuration parameters with optimal performance as the target configuration parameters; An evaluation module is used to evaluate the target configuration parameters based on the target performance information, and apply the target configuration parameters if the evaluation is passed.

[0013] In a third aspect, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to implement the steps of a combustion chamber optimization method for an aero-engine.

[0014] In a fourth aspect, a computer-readable storage medium is provided having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of a combustion chamber optimization method for an aero-engine.

[0015] In a fifth aspect, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of a combustion chamber optimization method for an aero-engine.

[0016] As will be described in detail below, a method, apparatus, device, medium, and product for optimizing the combustor of an aero-engine according to embodiments of this disclosure are disclosed. By first clarifying the target performance information of the combustor and accurately defining the optimization objectives and constraints, clear guidance is provided for subsequent parameter association and strategy formulation. This effectively improves the accuracy of establishing the association logic between target performance information and configuration parameters, thereby ensuring the rationality of the initial configuration parameter selection and avoiding the problem of blind parameter selection from the source of optimization.

[0017] A surrogate model that meets the accuracy requirements is used as the target performance parameter prediction model. The initial configuration parameters are iteratively calculated, which can take into account both the accuracy and efficiency of the iterative calculation by relying on the characteristics of the surrogate model, and can accurately output the specified configuration parameters through the iterative process. Then, the optimal specified configuration parameters are selected as the target configuration parameters through performance analysis. The evaluation and verification are completed in combination with the preset target performance information to ensure that the target configuration parameters fully meet the performance requirements and improve the accuracy and reliability of the optimization results (i.e., the target configuration parameters). Attached Figure Description

[0018] Figure 1 A flowchart illustrating a combustion chamber optimization method for an aero-engine provided in this embodiment of the disclosure; Figure 2 A schematic diagram of a first low-order proxy model for an aero-engine combustion chamber optimization method provided in this embodiment of the disclosure; Figure 3 An overall flowchart of a combustion chamber optimization method for an aero-engine provided in this disclosure embodiment; Figure 4 A schematic diagram of a combustion chamber optimization device for an aero-engine provided in an embodiment of this disclosure; Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present disclosure. Detailed Implementation

[0019] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present disclosure and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the drawings, not the entire structure.

[0020] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0021] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0022] Research has revealed that aero-engine combustor design is a core challenge in achieving high-performance, low-emission propulsion systems. Currently, the global aero-engine industry is accelerating its transformation towards digitalization and intelligence. However, deeply integrating artificial intelligence into combustor design and development still faces challenges.

[0023] In related technologies, it is difficult to accurately build parameter association logic, resulting in insufficient rationality in the selection of configuration parameters. Furthermore, the models used for iterative calculations often fail to balance accuracy and efficiency. At the same time, the performance analysis and screening of specified configuration parameters after iteration lacks precision, and the evaluation and verification of target configuration parameters are difficult to fully meet the preset performance requirements. Consequently, the optimization efficiency is low, the results are inaccurate, and the high-performance design requirements of the combustion chamber cannot be efficiently met.

[0024] Based on the above research, this disclosure provides a method for optimizing the combustor of an aero-engine. By first clarifying the target performance information of the combustor and accurately defining the optimization objectives and constraints, it provides clear guidance for subsequent parameter correlation and strategy formulation, effectively improving the accuracy of the construction of the correlation logic between target performance information and configuration parameters, thereby ensuring the rationality of the initial configuration parameter selection and avoiding the problem of blind parameter selection from the source of optimization.

[0025] A surrogate model that meets the accuracy requirements is used as the target performance parameter prediction model. The initial configuration parameters are iteratively calculated, which can take into account both the accuracy and efficiency of the iterative calculation by relying on the characteristics of the surrogate model, and can accurately output the specified configuration parameters through the iterative process. Then, the optimal specified configuration parameters are selected as the target configuration parameters through performance analysis. The evaluation and verification are completed in combination with the preset target performance information to ensure that the target configuration parameters fully meet the performance requirements and improve the accuracy and reliability of the optimization results (i.e., the target configuration parameters).

[0026] To facilitate understanding of this embodiment, a detailed description of the combustion chamber optimization method for an aero-engine disclosed in this disclosure will be provided first. The execution entity of the combustion chamber optimization method for an aero-engine provided in this disclosure is generally a computer device with certain computing capabilities, such as a terminal device, a server, or other processing devices. In some possible implementations, the combustion chamber optimization method for an aero-engine can be implemented by a processor calling computer-readable instructions stored in memory.

[0027] See Figure 1 The diagram shows a flowchart of a combustion chamber optimization method for an aero-engine provided in this embodiment of the present disclosure. The method includes steps S101 to S105, wherein: S101. Determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization objectives and constraints of the combustion chamber.

[0028] In the embodiments of this disclosure, target performance information corresponding to the combustor of the aero-engine to be simulated can be determined. The optimization objectives of the target performance information include, but are not limited to, exit temperature distribution coefficient, total pressure loss, and combustion efficiency; the constraints of the target performance information can limit the range of the combustor configuration parameters, ensuring that the optimization objectives do not exceed preset parameter values.

[0029] S102. Determine the optimization strategy between the target performance information and the configuration parameters of the combustion chamber, and determine several initial configuration parameters based on the optimization strategy.

[0030] In embodiments of this disclosure, several first configuration parameters can be randomly generated based on target performance information. These first configuration parameters can be determined using optimal Latin hypercube sampling.

[0031] Then, an optimization strategy can be determined and optimized based on the optimization objective to obtain the second configuration parameters. After determining the second configuration parameters, iterative calculations can be performed using a multi-objective genetic algorithm (NSGA-II) to obtain several initial configuration parameters.

[0032] S103. The initial configuration parameters are iteratively calculated using the target performance parameter prediction model to obtain the specified configuration parameters; wherein, the target performance parameter prediction model is a surrogate model that meets the accuracy condition.

[0033] In embodiments of this disclosure, initial configuration parameters can be input into a target performance parameter prediction model for iterative calculation to obtain prediction results for each initial configuration parameter. These prediction results indicate the potential target performance that each initial configuration parameter may produce during actual execution.

[0034] Here, the feasibility of each initial configuration parameter can be determined based on the prediction results, and the initial configuration parameters that meet the feasibility conditions can be determined as the specified configuration parameters.

[0035] Among them, the feasibility condition is the initial configuration parameter that meets the constraints in the target performance information.

[0036] Here, if the initial configuration parameters do not meet the feasibility conditions, the initial configuration parameters can be adjusted to obtain the third configuration parameters. The third configuration parameters are then re-inputted into the target performance parameter prediction model for iterative calculation. After iterating to the preset number of iterations, the third configuration parameters that meet the feasibility conditions are also determined as the specified configuration parameters.

[0037] S104. Analyze the specified configuration parameters to obtain the specified configuration parameters with the best performance, and determine the specified configuration parameters with the best performance as the target configuration parameters.

[0038] In embodiments of this disclosure, after determining the specified configuration parameters, non-dominated sorting and crowding calculations can be performed on the configuration parameters to determine the performance of each specified configuration parameter relative to the combustion chamber.

[0039] Then, the optimal configuration parameters can be determined as the target configuration parameters.

[0040] For example, the performance of each specified configuration parameter relative to the combustion chamber is determined as the Pareto optimal solution. Then, a secondary decision is made on the Pareto optimal solution, prioritizing the solution with the smallest OTDF (index D, OTDF=0.27), but physical feasibility needs to be checked: cone=38.5° (within the range [30°, 60°], easy to manufacture) and total pressure loss=2.9% (close to the upper limit of the constraint 2.8%, but acceptable for engineering) as the target configuration parameters.

[0041] S105. Evaluate the target configuration parameters based on the target performance information, and apply the target configuration parameters if the evaluation is successful.

[0042] In the embodiments of this disclosure, after the target configuration parameters are determined, the target configuration parameters can be simulated and calculated based on the target performance information using combustion chamber numerical values ​​to obtain the simulation results corresponding to the target configuration parameters.

[0043] Once the simulation results are confirmed to meet the performance requirements of the combustion chamber, the target configuration parameters are output for application in the actual processing.

[0044] In the embodiments of this disclosure, firstly, target performance information corresponding to the combustion chamber of the aero-engine is determined; wherein, the target performance information is used to indicate the optimization objectives and constraints of the combustion chamber; secondly, an optimization strategy between the target performance information and the configuration parameters of the combustion chamber is determined, and several initial configuration parameters are determined based on the optimization strategy; next, the initial configuration parameters are iteratively calculated using a target performance parameter prediction model to obtain specified configuration parameters; wherein, the target performance parameter prediction model is a surrogate model that meets accuracy conditions; next, the specified configuration parameters are analyzed to obtain the specified configuration parameters with optimal performance, and the specified configuration parameters with optimal performance are determined as the target configuration parameters; finally, the target configuration parameters are evaluated based on the target performance information, and if the target configuration parameters pass the evaluation, the target configuration parameters are applied.

[0045] In the above implementation, by first clarifying the target performance information of the combustion chamber and accurately defining the optimization goals and constraints, clear guidance is provided for subsequent parameter association and strategy formulation. This effectively improves the accuracy of building the association logic between target performance information and configuration parameters, thereby ensuring the rationality of the initial configuration parameter selection and avoiding the problem of blind parameter selection from the source of optimization.

[0046] A surrogate model that meets the accuracy requirements is used as the target performance parameter prediction model. The initial configuration parameters are iteratively calculated, which can take into account both the accuracy and efficiency of the iterative calculation by relying on the characteristics of the surrogate model, and can accurately output the specified configuration parameters through the iterative process. Then, the optimal specified configuration parameters are selected as the target configuration parameters through performance analysis. The evaluation and verification are completed in combination with the preset target performance information to ensure that the target configuration parameters fully meet the performance requirements and improve the accuracy and reliability of the optimization results (i.e., the target configuration parameters).

[0047] In an optional implementation, before determining the optimization strategy between the target performance parameters and the combustion chamber configuration parameters, a target performance parameter prediction model is determined as follows: First, determine the training sample set; the training sample set consists of several configuration parameters and specified performance parameters; Secondly, the covariance matrix is ​​determined based on the training sample set using a one-dimensional active subspace; Secondly, the training sample set is dimensionally reduced based on the covariance matrix to obtain matrix eigenvalues and eigenvectors; Secondly, when it is determined based on the matrix eigenvalues that the training sample set meets the dimensional reduction condition, the eigenvectors are processed to obtain active variables; among them, the active variables are used to indicate the influence of each configuration parameter in the training sample set on the specified performance parameter; Secondly, when it is determined based on the active variables that the first low-order surrogate model meets the simulation requirements, the first low-order surrogate model is determined as the initial performance parameter prediction model; Finally, based on the specified performance parameter and the corresponding optimization objective and constraint conditions of the specified performance parameter, the initial performance parameter prediction model is trained to obtain the target performance parameter prediction model.

[0048] In the embodiments of the present disclosure, the active subspace method based on global linear fitting can be used as the starting point for determining the process of the target performance parameter prediction model.

[0049] Here, the covariance matrix of the gradient vector can be confirmed through the one-dimensional active subspace. For example, for the one-dimensional active subspace f(x)≈β0+β T x, then xf≈β, and the covariance matrix C of the gradient vector of f(x) is obtained.

[0050] Among them, x is the construction parameter in the training sample set, β0 is the reference offset of the construction parameter, and β is the gradient vector.

[0051] After the covariance matrix is determined, the covariance matrix can be decomposed to obtain matrix eigenvalues and eigenvectors. For example, performing eigenvalue decomposition on matrix C, C =∫ββ T π(x)dx=ββ T =w Λ w T , to obtain matrix eigenvalues and eigenvectors.

[0052] Among them, Λ is the matrix eigenvalue, and w is the eigenvector.

[0053] Here, when the eigenvalue λ r is much larger than λ r+1 , that is, λ r λ r+1 (r < m), then the eigenvalue and eigenvector can be written as: ; W =[W 1W 2], where Λ 1=[λ 1,…,λ r ],(W 1)=[w 1, ,w r ] represents the direction of the low-dimensional subspace of the input model parameters.

[0054] Where m is the number of construction parameters.

[0055] For example, when m is 3, the covariance matrix C satisfies the following condition: .

[0056] Matrix eigenvalues ​​Λ Meets the following conditions: .

[0057] Feature vector w Meets the following conditions: .

[0058] Among them, the matrix eigenvalues ​​Λ Given λ1=0.0024, λ2=0.0003, and λ3=0.0001, satisfying λ1 λ2>λ3.

[0059] Here, after determining the eigenvalues ​​and eigenvectors of the matrix, the eigenvectors can be projected onto the subspace to obtain the active variables.

[0060] Among them, the active variable z meets the following conditions: ; w1 represents the weight and direction of the influence of configuration parameters on performance parameters, and x represents any configuration parameter vector in the training sample set.

[0061] Here, z = w1,1x2 + w1,1x2 + +w 1，i x i ; Among them, w 1，i Let I represent the weight and direction of the influence of the i-th configuration parameter on the performance parameters. The value of I ranges from 1 to m, where m is the number of configuration parameters.

[0062] Here, sensitivity analysis and physical insights can be performed on the training sample set based on the activity vector. Specifically, the activity direction vector components (i.e., w1 above) iThe magnitude and sign of the target quantity can be used to characterize the sensitivity of the target quantity to the input model parameters and predict the direction of parameter optimization.

[0063] Among them, the larger (smaller) the component value, the greater (smaller) the influence of the input model parameters on the target quantity; the positive (negative) component value indicates that a positive perturbation to the input model parameters will lead to an increase (decrease) in the target quantity.

[0064] Here, refer to Figure 2 The diagram shown is a schematic of a first low-order proxy model of a combustion chamber optimization method for an aero-engine provided in this embodiment of the present disclosure, wherein: Construct a linear low-order surrogate model (i.e., the first low-order surrogate model) on a one-dimensional activity space.

[0065] Then, the first prediction result can be obtained by making a prediction based on the training sample set using the first low-order proxy model.

[0066] Here, the accuracy of the first prediction result can be determined. If the accuracy of the first prediction result is greater than a preset accuracy threshold (for example, accuracy R2 > preset accuracy threshold 0.9), the first low-order proxy model is determined to meet the simulation requirements, and the first low-order proxy model is determined as the initial performance parameter prediction model.

[0067] In an optional implementation, after obtaining the active variables, the following steps are also included: First, when the low-order surrogate model determined based on active variables does not meet the simulation requirements, a high-order surrogate model is constructed using adaptive high-precision modeling tools; among which, the adaptive high-precision modeling tools include adaptive Gaussian process regression and adaptive multilayer perceptron. Then, based on the covariance matrix, the higher-order surrogate model is determined to meet the simulation requirements. The higher-order surrogate model is then dimensionality-reduced to obtain the second lower-order surrogate model, which is then used as the initial performance parameter prediction model.

[0068] In embodiments of this disclosure, a higher-order surrogate model can be constructed using adaptive Gaussian process regression (GPR): By employing Bayesian optimization and tree-based TPE hyperparameter optimization algorithms, the optimal kernel structure is automatically selected from kernel functions such as RBF, Matern, rational quadratic kernel, and the constructed combination forms based on the characteristics of the input data (i.e., the training sample set), and the hyperparameters are optimized simultaneously, thereby constructing a high-order surrogate model that combines high accuracy and robustness.

[0069] The Gaussian process is defined as follows: .

[0070] Where m(x) is the mean function and k(x, x') is the kernel function; The kernel function satisfies the following condition: .

[0071] Here, the combination of kernel functions can be implemented in the following way: .

[0072] Here, the gradient covariance matrix can be constructed through high-precision gradient solving (such as fourth-order central difference) to support the analysis of active subspaces.

[0073] Here, a high-order surrogate model can be constructed using an adaptive multilayer perceptron (MLP) by employing hierarchical sampling based on K-Means clustering instead of random sampling. This strategy ensures that both the training and validation sets maintain an overall structure and distribution similar to the original dataset by clustering in the input parameter space. This effectively avoids local data distribution deviations caused by random partitioning, making the results of hyperparameter optimization and performance evaluation more reliable and stable.

[0074] Subsequently, hyperparameter adaptive optimization based on the Optuna framework can be adopted. Starting from a simple network structure, the network depth and width are gradually increased through a conservative enhancement strategy. Combined with techniques such as Dropout and early stopping, overfitting is prevented, and finally the network with the highest accuracy and stability in this round is obtained. Then, the covariance matrix of the gradient is constructed by using the automatic differentiation method to quickly obtain high-precision gradient results, which greatly improves the quality and reliability of the active subspace analysis.

[0075] After determining the higher-order surrogate model, we can determine whether the higher-order surrogate model meets the simulation requirements based on the covariance matrix.

[0076] Here, if the accuracy of the higher-order surrogate model determined based on the covariance matrix is ​​greater than the preset accuracy threshold, it is determined that the higher-order surrogate model meets the simulation requirements. The higher-order surrogate model is then dimensionality-reduced to obtain the second lower-order surrogate model, which is then determined as the initial performance parameter prediction model.

[0077] In an optional implementation, after determining that the higher-order surrogate model meets the simulation requirements based on the covariance matrix, the higher-order surrogate model is dimensionality-reduced to obtain a second lower-order surrogate model. This process includes the following steps: First, under the condition that both the higher-order surrogate model obtained by adaptive Gaussian process regression and the higher-order surrogate model obtained by adaptive multilayer perceptron processing meet the simulation requirements, the accuracy of each higher-order surrogate model is determined. Secondly, the high-precision high-order proxy model is dimensionality reduced to obtain the second low-order proxy model.

[0078] In the embodiments of this disclosure, when both the higher-order surrogate model determined by adaptive Gaussian process regression and the higher-order surrogate model obtained by adaptive multilayer perceptron processing meet the simulation requirements, that is, when the accuracy of both the higher-order surrogate model determined by adaptive Gaussian process regression and the higher-order surrogate model obtained by adaptive multilayer perceptron processing is greater than a preset accuracy threshold.

[0079] Here, the accuracy of the high-order surrogate model obtained by adaptive Gaussian process regression and the high-order surrogate model obtained by adaptive multilayer perceptron can be compared, and the high-order surrogate model with higher accuracy can be selected as the second low-order surrogate model.

[0080] In an optional implementation, after constructing the high-order proxy model using an adaptive high-precision modeling tool, the following steps are also included: First, when the high-order surrogate model determined based on the covariance matrix does not meet the simulation requirements, the training sample set is augmented using a data augmentation strategy to obtain an augmented data sample set. Secondly, an enhanced high-order proxy model is constructed based on the enhanced data sample set using an adaptive high-precision modeling tool. Finally, if the enhanced higher-order surrogate model meets the simulation requirements, the enhanced higher-order surrogate model is dimensionality reduced to obtain the third lower-order surrogate model, which is then determined as the initial performance parameter prediction model.

[0081] In the embodiments of this disclosure, if the accuracy of both the higher-order surrogate model obtained by adaptive Gaussian process regression and the higher-order surrogate model obtained by adaptive multilayer perceptron processing is less than or equal to a preset accuracy threshold, it is determined that the higher-order surrogate model does not meet the simulation requirements.

[0082] Here, data augmentation strategies include: GPR posterior sampling, adaptive Gaussian noise, and KNN-guided interpolation.

[0083] In this context, GPR posterior sampling involves using a low-biased Sobol sequence to scatter points in the input space, jointly sampling the posterior function value and observation noise to generate statistically consistent candidate samples (x). , y After passing quality assessment, the data is included in the training set to ensure that the augmented data maintains consistency with the true conditional distribution in the sense of minimum mean square error.

[0084] The adaptive Gaussian noise is characterized by its noise intensity decreasing adaptively with the sample size. The adaptive Gaussian noise meets the following conditions: .

[0085] Among them, the coefficients 0.001 and 0.004 are calibration parameters for the data used. They need to be adjusted according to the data to ensure that the perturbation is weak when there are few samples and moderately enhanced when there are many samples. The samples after noise perturbation also need to be screened by RMSE quality assessment to suppress the risk of excessive smoothing or manifold shift.

[0086] In this method, KNN-guided interpolation is based on sampling weights λ from a Beta(2,2) distribution, with λ biased towards 0.5 to suppress extrapolation. For each sample, its K=5 nearest neighbors are taken to form a local convex hull, and λ-interpolation is performed within the hull to generate new samples. The interpolation results must be tested by the same quality assessment to ensure that the enhancement process follows a locally linear and non-discrete real manifold.

[0087] Afterwards, the higher-order surrogate model can be retrained using the augmented data sample set obtained through data augmentation strategies until the requirements for iterative training are met.

[0088] After meeting the iterative training requirements, and if the enhanced higher-order surrogate model meets the simulation requirements, the enhanced higher-order surrogate model is dimensionality-reduced to obtain the third lower-order surrogate model, which is then determined as the initial performance parameter prediction model.

[0089] If the enhanced higher-order surrogate model does not meet the simulation requirements, then based on the lower-order model and the sensitivity analysis results, find the sparse sample region in the parameter space (e.g., only 2 samples in the range of cone=0.8-1.0), return to resample to obtain the updated training sample set (supplementing several samples in a targeted manner), and continue to train the higher-order surrogate model until the accuracy of the higher-order surrogate model meets the requirements.

[0090] In one optional implementation, determining the training sample set specifically includes the following steps: First, determine the parameter range of the configuration parameters corresponding to the combustion chamber; Secondly, sampling is performed based on the parameter range using the optimal Latin hypercube sampling method to obtain multiple initial configuration parameter sample sets; Secondly, the overall potential energy of each initial configuration parameter sample set is determined, and the initial configuration parameter sample set corresponding to the minimum overall potential energy is determined as the training configuration parameter sample set; Secondly, simulation calculations are performed on the training configuration parameter sample set to obtain the training performance parameter set corresponding to the training configuration parameter sample set; Finally, the training configuration parameter sample set and the training performance parameter set are determined as the training sample set.

[0091] In the embodiments of this disclosure, the parameter range of the configuration parameters corresponding to the combustion chamber can be determined based on the combustion chamber reference configuration, combined with the combustion chamber design manual, literature and engineering experience.

[0092] For example, as shown in Table 1 below, an example of the parameter ranges of the configuration parameters and construction parameters corresponding to the combustion chamber is as follows: Table 1

[0093] Here, after determining the parameter range of the configuration parameters corresponding to the combustion chamber, the optimal Latin hypercube sampling method can be adopted and combined with the AE potential energy criterion to obtain a set of optimal combustion chamber configuration parameter samples with optimal spatial distribution, and a training configuration parameter sample set is obtained.

[0094] Here, the parameter range of the configuration parameters can be normalized, and each parameter range is converted to the interval [0,1] to eliminate the influence of the dimension.

[0095] After that, an m-dimensional normalized parameter space R formed by the configuration parameters m is sampled using the optimal Latin hypercube sampling method to generate an initial configuration parameter sample set.

[0096] After that, the initial configuration parameter sample set can be processed through the AE potential energy criterion to obtain the overall potential energy of the initial configuration parameter sample set.

[0097] Among them, the potential energy calculation formula meets the following conditions: .

[0098] Among them, ||x d -x j || is the Euclidean distance between the d-th initial configuration parameter sample and the j-th initial configuration parameter sample in the initial configuration parameter sample set. The value range of d is from 1 to k, the value range of j is from 1 to k, and k is the number of initial configuration parameter samples in the initial configuration parameter sample set.

[0099] After determining the overall potential energy corresponding to each initial configuration parameter sample set (each initial configuration parameter sample set corresponds to an overall potential energy), the initial configuration parameter sample set with the minimum overall potential energy can be determined, and this initial configuration parameter sample set is determined as the training performance parameter set.

[0100] For example, if U(X1) < U(X2), it is determined that the sample set X1 is better than X2, otherwise X2 is better. Among them, X1 and X2 are respectively the parameter sample sets of a single sampling (that is, the initial configuration parameter sample set), and finally the initial configuration parameter sample set with the minimum overall potential energy is output as the training performance parameter set.

[0101] See Figure 3 shown, which is the overall flowchart of an optimization method for the combustion chamber of an aeroengine provided by an embodiment of the present disclosure, where: S1. The optimal Latin hypercube sampling method is used to sample the parameters based on the range of the combustion chamber's structural parameters to obtain multiple initial configuration parameter sample sets.

[0102] S2. Determine the overall potential energy of each initial configuration parameter sample set, and determine the initial configuration parameter sample set corresponding to the minimum overall potential energy as the training configuration parameter sample set.

[0103] S3. Perform simulation calculations on the training configuration parameter sample set to obtain the training sample set.

[0104] S4. Determine whether the first low-order agent model meets the simulation requirements based on the training sample set.

[0105] S5. If the first low-order proxy model meets the simulation requirements, the first low-order proxy model is determined as the initial performance parameter prediction model.

[0106] S6. If the first low-order proxy model does not meet the simulation requirements, construct a high-order proxy model using an adaptive high-precision modeling tool.

[0107] S7. Determine whether the high-order agent model meets the simulation requirements based on the training sample set.

[0108] S8. If the high-order surrogate model meets the simulation requirements, the high-order surrogate model is reduced in dimensionality to obtain the second low-order surrogate model, and the second low-order surrogate model is determined as the initial performance parameter prediction model.

[0109] S9. If it is determined that the higher-order proxy model does not meet the simulation requirements, the training sample set is augmented using a data augmentation strategy to obtain an augmented data sample set.

[0110] S10. Based on the enhanced data sample set, construct the enhanced high-order proxy model using an adaptive high-precision modeling tool.

[0111] S11. Determine whether the enhanced higher-order proxy model meets the simulation requirements.

[0112] S12. If the enhanced high-order proxy model meets the simulation requirements, the enhanced high-order proxy model is dimensionality reduced to obtain the third low-order proxy model, and the third low-order proxy model is determined as the initial performance parameter prediction model.

[0113] S13. If the enhanced higher-order surrogate model does not meet the simulation requirements, then based on the lower-order model and the sensitivity analysis results, find the sparse sample region in the parameter space, and continue to train the higher-order surrogate model until the higher-order surrogate model meets the simulation requirements. Then, determine the higher-order surrogate model as the initial performance parameter prediction model.

[0114] S14. Train the initial performance parameter prediction model to obtain the target performance parameter prediction model.

[0115] S15. The target performance parameter prediction model is used to iteratively calculate the configuration parameters to be predicted (i.e., the initial configuration parameters mentioned above) to obtain the specified configuration parameters.

[0116] S16. Analyze the specified configuration parameters to obtain the specified configuration parameters with the best performance, and determine the specified configuration parameters with the best performance as the target configuration parameters.

[0117] S17. Evaluate the target configuration parameters based on the target performance information, and apply the target configuration parameters if the evaluation is successful.

[0118] In the actual implementation of this embodiment, the following technical effects are achieved: (1) Parameter sampling is performed by the optimal Latin hypercube sampling method, and the training configuration parameter sample set is determined by the overall potential energy screening. This can achieve uniform and efficient sampling in the full range of combustion chamber construction parameters, avoiding sample redundancy or uneven distribution. At the same time, the minimum overall potential energy screening ensures that the training sample set has the optimal initial characteristics, providing a high-quality data foundation for subsequent model construction and improving the reliability of model prediction from the source. Compared with traditional random sampling methods, it can significantly reduce the proportion of invalid samples and improve sampling efficiency and sample validity.

[0119] (2) By adopting a hierarchical model construction strategy of "low-order model priority verification + high-order model adaptive construction + dimensionality reduction", a dynamic balance between model accuracy and computational efficiency can be achieved. When the first low-order surrogate model meets the simulation requirements, it can be used directly to quickly obtain the initial performance parameter prediction model with low computational cost. When the low-order model does not meet the requirements, a high-order surrogate model is constructed through an adaptive high-precision modeling tool to ensure prediction accuracy. Then, it is transformed into a low-order model through dimensionality reduction. This not only retains the accuracy advantage of the high-order model, but also reduces the computational power consumption of subsequent iterations, thus solving the contradiction that traditional models are either not accurate enough or have low computational efficiency.

[0120] (3) For scenarios where the high-order surrogate model does not meet the simulation requirements, a data augmentation strategy and a sample sparse region completion strategy are adopted in sequence to form a closed-loop model optimization mechanism. The data augmentation strategy can expand the amount of data on the basis of existing training samples to alleviate the problem of insufficient model training under the constraint of small samples; the sample sparse region completion strategy combines the low-order model and sensitivity analysis to accurately locate the weak regions in the parameter space, supplement the training data in a targeted manner, avoid the prediction bias caused by the model due to incomplete sample coverage, and effectively improve the generalization ability and simulation reliability of the model.

[0121] (4) By obtaining a target performance parameter prediction model through multiple rounds of model verification, optimization and training, and then iteratively calculating the configuration parameters to be predicted based on this model, the correlation between combustion chamber configuration parameters and performance parameters can be accurately mapped, and the specified configuration parameters can be output efficiently. Subsequently, the optimal configuration parameters are screened through performance analysis and evaluated and verified in combination with the target performance information, which can ensure that the final determined target configuration parameters fully meet the preset optimization goals and constraints, improve the success rate of combustion chamber optimization design, and avoid the optimization results from being out of touch with actual engineering needs.

[0122] Based on the same inventive concept, this disclosure also provides an aero-engine combustor optimization device corresponding to the aero-engine combustor optimization method. Since the principle of the device in this disclosure for solving the problem is similar to the aero-engine combustor optimization method described above in this disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0123] Reference Figure 4 The diagram shown is a schematic of a combustion chamber optimization device for an aero-engine provided in this embodiment of the present disclosure, including: a first determining module 41, a second determining module 42, a calculation module 43, an analysis module 44, and an evaluation module 45; wherein: The first determining module is used to determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization target and constraints of the combustion chamber; The second determining module is used to determine an optimization strategy between the target performance information and the configuration parameters of the combustion chamber, and to determine several initial configuration parameters based on the optimization strategy. The calculation module is used to iteratively calculate the initial configuration parameters using a target performance parameter prediction model to obtain the specified configuration parameters; wherein the target performance parameter prediction model is a surrogate model that meets the accuracy condition. The analysis module is used to analyze the specified configuration parameters, obtain the specified configuration parameters with optimal performance, and determine the specified configuration parameters with optimal performance as the target configuration parameters; An evaluation module is used to evaluate the target configuration parameters based on the target performance information, and apply the target configuration parameters if the evaluation is passed.

[0124] This embodiment first clarifies the target performance information of the combustion chamber and accurately defines the optimization goals and constraints, providing clear guidance for subsequent parameter association and strategy formulation. This effectively improves the accuracy of building the association logic between target performance information and configuration parameters, thereby ensuring the rationality of the initial configuration parameter selection and avoiding the problem of blind parameter selection from the source of optimization.

[0125] A surrogate model that meets the accuracy requirements is used as the target performance parameter prediction model. The initial configuration parameters are iteratively calculated, which can take into account both the accuracy and efficiency of the iterative calculation by relying on the characteristics of the surrogate model, and can accurately output the specified configuration parameters through the iterative process. Then, the optimal specified configuration parameters are selected as the target configuration parameters through performance analysis. The evaluation and verification are completed in combination with the preset target performance information to ensure that the target configuration parameters fully meet the performance requirements and improve the accuracy and reliability of the optimization results (i.e., the target configuration parameters).

[0126] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.

[0127] Corresponding to Figure 1 In addition to the gas leakage monitoring method, this disclosure also provides an electronic device 500, such as... Figure 5 The diagram shown is a structural schematic of an electronic device 500 provided in an embodiment of this disclosure, including: The system includes a processor 51, a memory 52, and a bus 55. The memory 52 stores execution instructions and includes main memory 521 and external memory 522. The main memory 521, also called internal memory, temporarily stores the computational data in the processor 51, as well as data exchanged with external memory such as a hard disk. The processor 51 exchanges data with the external memory 522 through the main memory 521. When the electronic device 500 is running, the processor 51 communicates with the memory 52 through the bus 53, causing the processor 51 to execute the following instructions: Determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization objectives and constraints of the combustion chamber; An optimization strategy is determined between the target performance information and the configuration parameters of the combustion chamber, and several initial configuration parameters are determined based on the optimization strategy; The initial configuration parameters are iteratively calculated using a target performance parameter prediction model to obtain the specified configuration parameters; wherein, the target performance parameter prediction model is a surrogate model that meets the accuracy condition. The specified configuration parameters are analyzed to obtain the specified configuration parameters with optimal performance, and the specified configuration parameters with optimal performance are determined as the target configuration parameters; The target configuration parameters are evaluated based on the target performance information, and if the target configuration parameters pass the evaluation, the target configuration parameters are applied.

[0128] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0129] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0130] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0131] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0132] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0133] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0134] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for optimizing the combustion chamber of an aero-engine, characterized in that, include: Determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization objectives and constraints of the combustion chamber; An optimization strategy is determined between the target performance information and the configuration parameters of the combustion chamber, and several initial configuration parameters are determined based on the optimization strategy; The initial configuration parameters are iteratively calculated using a target performance parameter prediction model to obtain the specified configuration parameters; wherein, the target performance parameter prediction model is a surrogate model that meets the accuracy condition. The specified configuration parameters are analyzed to obtain the specified configuration parameters with optimal performance, and the specified configuration parameters with optimal performance are determined as the target configuration parameters; The target configuration parameters are evaluated based on the target performance information, and if the target configuration parameters pass the evaluation, the target configuration parameters are applied.

2. The method as described in claim 1, characterized in that, Before determining the optimization strategy between the target performance parameters and the combustion chamber configuration parameters, the target performance parameter prediction model is determined by: determining a training sample set; wherein the training sample set consists of several configuration parameters and specified performance parameters; The covariance matrix is ​​determined based on the training sample set using a one-dimensional active subspace; The training sample set is reduced in dimensionality based on the covariance matrix to obtain matrix eigenvalues ​​and eigenvectors. If the training sample set satisfies the dimensionality reduction condition based on the matrix eigenvalues, the eigenvectors are processed to obtain active variables; wherein, the active variables are used to indicate the influence of each configuration parameter in the training sample set on the specified performance parameter; If the first low-order surrogate model meets the simulation requirements based on the active variables, the first low-order surrogate model is determined as the initial performance parameter prediction model. Based on the specified performance parameters and the corresponding optimization objectives and constraints, the initial performance parameter prediction model is trained to obtain the target performance parameter prediction model.

3. The method as described in claim 2, characterized in that, After obtaining the active variables, the method further includes: If the low-order surrogate model does not meet the simulation requirements based on the active variables, a high-order surrogate model is constructed using an adaptive high-precision modeling tool; wherein, the adaptive high-precision modeling tool includes adaptive Gaussian process regression and adaptive multilayer perceptron; If the higher-order surrogate model meets the simulation requirements based on the covariance matrix, the higher-order surrogate model is dimensionality reduced to obtain a second lower-order surrogate model, and the second lower-order surrogate model is determined as the initial performance parameter prediction model.

4. The method as described in claim 3, characterized in that, When the higher-order surrogate model is determined to meet the simulation requirements based on the covariance matrix, the higher-order surrogate model is dimensionality-reduced to obtain a second lower-order surrogate model, including: If both the higher-order surrogate model obtained by the adaptive Gaussian process regression and the higher-order surrogate model obtained by the adaptive multilayer perceptron meet the simulation requirements, determine the accuracy of each higher-order surrogate model. The high-accuracy high-order proxy model is dimensionality reduced to obtain the second low-order proxy model.

5. The method as described in claim 3, characterized in that, After constructing a high-order proxy model using an adaptive high-precision modeling tool, the method further includes: If the higher-order surrogate model does not meet the simulation requirements based on the covariance matrix, the training sample set is augmented using a data augmentation strategy to obtain an augmented data sample set. An enhanced high-order proxy model is constructed based on the enhanced data sample set using the adaptive high-precision modeling tool. If the enhanced high-order proxy model meets the simulation requirements, the enhanced high-order proxy model is dimensionality reduced to obtain a third low-order proxy model, and the third low-order proxy model is determined as the initial performance parameter prediction model.

6. The method as described in claim 2, characterized in that, The determination of the training sample set includes: Determine the parameter range of the configuration parameters corresponding to the combustion chamber; Multiple initial configuration parameter sample sets are obtained by sampling based on the parameter range using the optimal Latin hypercube sampling method. Determine the overall potential energy of each of the initial configuration parameter sample sets, and determine the initial configuration parameter sample set corresponding to the minimum overall potential energy as the training configuration parameter sample set; Simulation calculations are performed on the training configuration parameter sample set to obtain the training performance parameter set corresponding to the training configuration parameter sample set; The training configuration parameter sample set and the training performance parameter set are determined as the training sample set.

7. A combustion chamber optimization device for an aero-engine, characterized in that, include: The first determining module is used to determine the target performance information corresponding to the combustion chamber of the aero-engine; wherein, the target performance information is used to indicate the optimization target and constraints of the combustion chamber; The second determining module is used to determine an optimization strategy between the target performance information and the configuration parameters of the combustion chamber, and to determine several initial configuration parameters based on the optimization strategy. The calculation module is used to iteratively calculate the initial configuration parameters using a target performance parameter prediction model to obtain the specified configuration parameters; wherein the target performance parameter prediction model is a surrogate model that meets the accuracy condition. The analysis module is used to analyze the specified configuration parameters, obtain the specified configuration parameters with optimal performance, and determine the specified configuration parameters with optimal performance as the target configuration parameters; An evaluation module is used to evaluate the target configuration parameters based on the target performance information, and apply the target configuration parameters if the evaluation is passed.

8. An electronic device 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 as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 6.

Images