A method for recommending turbine blade cooling structure parameters based on generative models

Abstract

This invention discloses a method for recommending turbine blade cooling structure parameters based on a generative model, belonging to the field of aero-engine turbine blade cooling design technology. It includes: separating the decoder and encoder parts in the generative model training to ensure that the feature parameters generated by the encoder belong to the cooling structure parameter space, guaranteeing the physical meaning of the generated parameters; and achieving collaborative recommendation of multiple design objectives through an innovative model architecture linking one encoder to multiple decoders and a training method that associates multiple design objectives through a loss function. This application uses a generative model to directly output cooling structure parameters for given design indicators under different operating conditions, thereby improving the efficiency of turbine blade cooling design; and by incorporating data from each design activity into the training dataset, it continuously improves the model's recommendation capability.

Description

Technical Field

[0001] This invention belongs to the field of aero-engine turbine blade cooling design technology, and particularly relates to a method for recommending turbine blade cooling structure parameters based on a generative model. Background Technology

[0002] In the current design process of turbine blade cooling for aero-engines, the initial cooling structure parameters are mainly given directly based on human experience, and then iterative optimization algorithms are used to optimize the cooling structure parameters to achieve the design specifications.

[0003] Facing increasingly advanced aero-engines of the future, turbine blades will operate under more severe conditions, with more demanding design specifications and more intricate structural layouts. Current "empirical" initial design methods are highly likely to deviate significantly from the feasible design zone, increasing the time and difficulty of subsequent "iterative" optimization. Furthermore, since each design activity is independent, historical design results and data cannot be reused. Therefore, to improve the efficiency and capability of turbine blade cooling design in my country's aero-engines, it is urgent to develop an efficient and continuously evolving method for recommending turbine blade cooling structural parameters. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method for recommending turbine blade cooling structure parameters based on a generative model, thereby resolving the issues present in the prior art.

[0005] To achieve the above objectives, this invention provides a method for recommending turbine blade cooling structure parameters based on a generative model, comprising:

[0006] A training dataset of turbine blade cooling design indices corresponding to different operating conditions and structural parameters was constructed based on numerical simulation methods.

[0007] Multiple decoder models are constructed based on a neural network architecture, and the multiple decoder models are trained using the training dataset to generate multiple trained decoder models;

[0008] An encoder model is constructed based on a neural network architecture, and the encoder model is trained using the training dataset to generate a trained encoder model.

[0009] The output of the trained encoder model is connected to the input of the trained multiple decoder models to generate a recommended model for turbine blade cooling structure parameters.

[0010] Based on the recommended model of turbine blade cooling structure parameters, the cooling structure parameters of the design target under the current operating conditions are generated.

[0011] Preferably, the operating conditions include, but are not limited to: mainstream Reynolds number, turbulence intensity, density ratio, blowing ratio, and pressure difference between the inside and outside of the blades;

[0012] The structural parameters include, but are not limited to: impact hole diameter, air film hole diameter, air film hole incident angle, aspect ratio, hole spacing, and row spacing.

[0013] Preferably, the process of constructing multiple decoder models based on a neural network architecture includes: constructing multiple decoder models based on a neural network architecture, ranging from operating conditions and structural parameters to cooling air flow and blade integrated cooling efficiency design indicators.

[0014] Preferably, the expression of the decoder model is:

[0015] Y = G(X) = G(Re,Tu,DR,M,ΔP,d) i ,d f ,α,l / d f ,P / d f ,S / d f ,...);

[0016] Where Re is the mainstream Reynolds number, Tu is the turbulence intensity, DR is the density ratio, M is the blowing ratio, ΔP is the pressure difference between the inside and outside of the blade; di is the diameter of the impact hole, df is the diameter of the film cooling hole, α is the incident angle of the film cooling hole, l / df is the aspect ratio, P / df is the hole spacing, S / df is the row spacing, G(X) is the constructed decoder model, and Y is the design index.

[0017] Preferably, the operating conditions and structural parameters of the encoder, including the cooling air flow rate and the overall cooling efficiency design index of the blades, are the same as those of the decoder.

[0018] Preferably, the expression for the loss function of the encoder is:

[0019]

[0020] Where YR represents the output performance of the decoder, such as the airflow rate and the overall cooling efficiency of the blades, Y represents the design specifications, and K represents the dimension of the input data.

[0021] Preferably, the process of concatenating the output of the trained encoder model with the inputs of the trained multiple decoder models includes:

[0022] The trained encoder output is used as a hub to connect multiple decoder models in parallel, and the training methods of multiple design objectives are associated through a loss function to form a unified paradigm for collaborative recommendation of multiple design objectives.

[0023] Preferably, the unified paradigm for collaborative recommendation of multiple design objectives is:

[0024]

[0025] Where k is the total number of decoders, and ni is the dimension of the design metrics in the i-th decoder. Y is a tensor format of the predicted design metrics for all dimensions in the i-th decoder. i It is a tensor format for the design metric target values ​​of all dimensions in the i-th decoder. Let y be the predicted value of the design metric for the j-th dimension in the i-th decoder. ij Let be the target value of the design metric for the j-th dimension in the i-th decoder.

[0026] Compared with the prior art, the present invention has the following advantages and technical effects:

[0027] This invention innovatively proposes a physically meaningful generative artificial intelligence model with encoder-decoder separation training for the problem of recommending turbine blade cooling structure parameters;

[0028] This invention innovatively proposes a model architecture that links multiple decoders to an encoder and a training method that associates multiple design objectives through a loss function to address the collaborative recommendation problem of multiple design objectives for turbine blades. Attached Figure Description

[0029] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0030] Figure 1 This is a flowchart illustrating the method for recommending turbine blade cooling structure parameters based on a generative model, as described in an embodiment of the present invention. Detailed Implementation

[0031] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0033] Example 1

[0034] like Figure 1 As shown, this embodiment provides a method for recommending turbine blade cooling structure parameters based on a generative model, including:

[0035] A training dataset of turbine blade cooling design indices corresponding to different operating conditions and structural parameters was constructed based on numerical simulation methods.

[0036] Multiple decoder models are constructed based on a neural network architecture, and the multiple decoder models are trained using the training dataset to generate multiple trained decoder models;

[0037] An encoder model is constructed based on a neural network architecture, and the encoder model is trained using the training dataset to generate a trained encoder model.

[0038] The output of the trained encoder model is connected to the input of the trained multiple decoder models to generate a recommended model for turbine blade cooling structure parameters.

[0039] Based on the recommended model of turbine blade cooling structure parameters, the cooling structure parameters of the design target under the current operating conditions are generated, specifically:

[0040] Step 1: Building the Training Dataset

[0041] Based on existing numerical simulation methods, a training dataset is constructed to represent turbine blade cooling design indices such as airflow and overall blade cooling efficiency under different operating conditions and structural parameters.

[0042] Step 2: Decoder Model Construction and Training

[0043] A decoder model was constructed based on a neural network architecture, encompassing design parameters such as operating conditions and structural parameters, as well as airflow and overall blade cooling efficiency. This decoder model was then trained using the dataset constructed in step ①. Operating parameters included those commonly used in engineering practice, such as the prevailing Reynolds number (Re), turbulence intensity (Tu), density ratio (DR), airflow ratio (M), and pressure difference between the inside and outside of the blade (ΔP). Geometric parameters included the impact hole diameter (d). i ), air film pore diameter (d) f ), air film aperture incident angle (α), aspect ratio (l / d) f Hole spacing (P / d) f ), row spacing (S / d) f The number of parameters can be increased by expanding the dataset, and this application does not limit the types and number of parameters. The constructed decoder model G(X) is shown in equation (1).

[0044] Y = G(X) = G(Re,Tu,DR,M,ΔP,d) i ,d f ,α,l / df,P / df,S / df,...)(1)

[0045] Step 3: Encoder Model Construction and Training

[0046] An encoder model is constructed based on a neural network architecture, which integrates design parameters such as operating conditions, airflow, and blade cooling efficiency into structural parameters. The output of the encoder model is linked to the input of the decoder model trained in step ②, and the dataset constructed in step ① is used to train the encoder model. The operating parameters, structural parameters, and design parameters of the encoder model are completely consistent with those of the decoder model trained in step 2. The constructed encoder model F(Y) is shown in equation (2), and its function is to compress the high-dimensional parameter space containing design parameters Y such as airflow and blade cooling efficiency into the low-dimensional structural parameter space X.

[0047] X=F(Y)(2)

[0048] To achieve the design goals as much as possible, the encoder's training objective is to minimize the cool air flow rate, overall blade cooling efficiency, and other performance parameters (Y) output by the decoder. R The difference between the target performance Y and the encoder input is generally calculated using the mean square error of the two, as shown in equation (3).

[0049]

[0050] Where K represents the dimension of the input data. After training, the encoder output is the cooling structure parameters designed for the current operating conditions.

[0051] For turbine blade cooling structure design, it is generally required to simultaneously meet design indicators from multiple disciplines such as cooling air flow, wall temperature, strength and life. However, traditional autoencoders typically have one encoder and one decoder, lacking a model architecture and training method that allows multiple decoders to coexist. In response, this application proposes a model architecture that uses the encoder output as a hub to connect multiple decoders in parallel and a training method that associates multiple design objectives through a loss function, forming a unified paradigm for collaborative recommendation of multiple design objectives, as shown in Equation (4).

[0052]

[0053] Where k is the total number of decoders, and ni is the dimension of the design metrics in the i-th decoder. Y is a tensor format of the predicted design metrics for all dimensions in the i-th decoder. i It is a tensor format for the design metric target values ​​of all dimensions in the i-th decoder. Let y be the predicted value of the design metric for the j-th dimension in the i-th decoder. ij Let be the target value of the design metric for the j-th dimension in the i-th decoder. The values ​​of each term in this loss function are reduced throughout the training process of the autoencoder model, thereby achieving the correlation and satisfaction of multiple design objectives.

[0054] Step 4: Recommended Cooling Structure Parameters

[0055] During the application phase, based on the input operating conditions and design parameters such as target cooling airflow and blade overall cooling efficiency, the encoder model trained in step ③ is invoked to directly output the corresponding cooling structure parameters and complete the recommendation.

[0056] Step 5: Collection and Reuse of Historical Design Activity Data

[0057] The data from each design activity is incorporated into the training dataset built in step one, thereby continuously improving the capabilities of the decoder and encoder models built in steps two and three.

[0058] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for recommending turbine blade cooling structure parameters based on a generative model, characterized in that, Includes the following steps: A training dataset of turbine blade cooling design indices corresponding to different operating conditions and structural parameters was constructed based on numerical simulation methods. Multiple decoder models are constructed based on a neural network architecture, and the multiple decoder models are trained using the training dataset to generate multiple trained decoder models; An encoder model is constructed based on a neural network architecture, and the encoder model is trained using the training dataset to generate a trained encoder model. The output of the trained encoder model is connected to the input of the trained multiple decoder models to generate a recommended model for turbine blade cooling structure parameters. Based on the recommended model of turbine blade cooling structure parameters, the cooling structure parameters of the design target under the current operating conditions are generated.

2. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 1, characterized in that, The operating conditions include, but are not limited to: mainstream Reynolds number, turbulence intensity, density ratio, blowing ratio, and pressure difference between the inside and outside of the blades; The structural parameters include, but are not limited to: impact hole diameter, air film hole diameter, air film hole incident angle, aspect ratio, hole spacing, and row spacing.

3. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 1, characterized in that, The process of constructing multiple decoder models based on a neural network architecture includes: constructing multiple decoder models based on a neural network architecture, ranging from operating conditions and structural parameters to cooling air flow and blade integrated cooling efficiency design indicators.

4. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 1, characterized in that, The expression for the decoder model is: ； Where Re is the mainstream Reynolds number, Tu is the turbulence intensity, DR is the density ratio, M is the blowing ratio, and ∆P is the pressure difference between the inside and outside of the blade; d i For the diameter of the impact hole, d f Where α is the diameter of the air film aperture, l / d is the incident angle of the air film aperture, and α is the diameter of the air film aperture. f The aspect ratio, P / d f Hole spacing, S / d f Let G(X) be the row spacing, G(X) be the constructed decoder model, and Y be the design metric.

5. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 1, characterized in that, The operating conditions and structural parameters of the encoder, as well as the design indicators for the overall cooling efficiency of the blades, are the same as those of the decoder.

6. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 1, characterized in that, The expression for the loss function of the encoder is: ； Among them, Y R Y represents the cool air flow rate and the overall cooling efficiency of the blades output by the decoder, and K represents the dimension of the input data.

7. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 1, characterized in that, The process of concatenating the output of the trained encoder model with the inputs of the trained multiple decoder models includes: The trained encoder output is used as a hub to connect multiple decoder models in parallel, and the training methods of multiple design objectives are associated through a loss function to form a unified paradigm for collaborative recommendation of multiple design objectives.

8. The method for recommending turbine blade cooling structure parameters based on a generative model according to claim 7, characterized in that, The unified paradigm for collaborative recommendation with multiple design objectives is as follows: ； Where k is the total number of decoders, n i Design the dimensions of the metrics in the i-th decoder. Let be the tensor format of the predicted values ​​of the design metrics for all dimensions in the i-th decoder. Let be the tensor format of the design metric target values ​​for all dimensions in the i-th decoder. Let j be the predicted design metric value for the j-th dimension in the i-th decoder. Let be the target value of the design metric for the j-th dimension in the i-th decoder.

Images