CN-119272435-B - Turbine blade cooling structure parameter recommendation method based on generative model

CN119272435BCN 119272435 BCN119272435 BCN 119272435BCN-119272435-B

Abstract

The application discloses a turbine blade cooling structure parameter recommendation method based on a generation model, which belongs to the technical field of turbine blade cooling design of aeroengines and comprises the steps of realizing that characteristic parameters generated by an encoder belong to a cooling structure parameter space by separating and training two parts of a decoder and an encoder in the generation model, ensuring the physical meaning of the generated parameters, and realizing multi-design target collaborative recommendation by using an innovative model framework of which one encoder is linked with a plurality of decoders and a training method of which a plurality of design targets are related by a loss function. The method and the device realize the direct output of the cooling structure parameters aiming at the given design indexes under different working conditions through the generated model, so as to achieve the aim of improving the cooling design efficiency of the turbine blade, and achieve the aim of continuously improving the recommending capacity of the model through incorporating data in each design activity into a training data set.

Inventors

LI DIKE
Tao Kaihang
QIU LU
ZHU JIANQIN
TAO ZHI

Assignees

天目山实验室

Dates

Publication Date: 20260505
Application Date: 20240928

Claims (8)

1. The turbine blade cooling structure parameter recommendation method based on the generative model is characterized by comprising the following steps of: Constructing a training data set of turbine blade cooling design indexes corresponding to different working conditions and structural parameters based on a numerical simulation method; constructing a plurality of decoder models based on a neural network architecture, training the decoder models through the training data set, and generating a plurality of trained decoder models; Constructing an encoder model based on a neural network architecture, training the encoder model through the training data set, and generating a trained encoder model; Connecting the output of the trained encoder model with the input of the trained plurality of decoder models to generate a turbine blade cooling structure parameter recommendation model; And generating cooling structure parameters of the design target under the current working condition based on the turbine blade cooling structure parameter recommendation model.
2. The turbine blade cooling structure parameter recommendation method based on the generative model of claim 1, wherein the operating conditions include, but are not limited to, main flow Reynolds number, turbulence level, density ratio, blowing ratio, and blade internal-external pressure difference; The structural parameters include, but are not limited to, impingement hole diameter, gas film hole incidence angle, aspect ratio, hole spacing, and row spacing.
3. The turbine blade cooling structure parameter recommendation method based on the generative model of claim 1, wherein the process of constructing a plurality of decoder models based on the neural network architecture comprises constructing a plurality of decoder models from working conditions and structural parameters to cold air flow and blade comprehensive cold efficiency design indexes based on the neural network architecture.
4. The turbine blade cooling structure parameter recommendation method based on a generative model of claim 1, wherein the decoder model has the expression: ; Wherein Re is the main flow Reynolds number, tu is the turbulence degree, DR is the density ratio, M is the blowing ratio, P is the pressure difference between the inside and outside of the blade, d i is the impact hole diameter, d f is the air film hole diameter, alpha is the air film hole incidence angle, l/d f is the length-diameter ratio, P/d f is the hole spacing, S/d f is the row spacing, G (X) is the constructed decoder model, and Y is the design index.
5. The turbine blade cooling structure parameter recommendation method based on the generative model as claimed in claim 1, wherein the working condition and the structural parameters of the encoder to the cold air flow and the blade comprehensive cold effect design index are the same as those of the decoder.
6. The turbine blade cooling structure parameter recommendation method based on a generative model of claim 1, wherein the expression of the loss function of the encoder is: ; wherein Y R is the cold air flow output by the decoder and the comprehensive cold effect performance of the blade, Y is a design index, and K represents the dimension of input data.
7. The generated model based turbine blade cooling structure parameter recommendation method of claim 1, wherein the process of connecting the output of the trained encoder model with the inputs of the trained plurality of decoder models comprises: And taking the trained encoder output as a hub to connect a plurality of decoder models in parallel, and forming a multi-design-target collaborative recommendation unified paradigm through a training method of associating a plurality of design targets by a loss function.
8. The turbine blade cooling structure parameter recommendation method based on a generative model as claimed in claim 7, wherein the multi-design objective collaborative recommendation unified paradigm is: ; where k is the total number of decoders, n i is the dimension of the design index in the ith decoder, The tensor format of the design indicator predictor for all dimensions in the ith decoder, Tensor format for design index target values for all dimensions in the ith decoder, For the design indicator predictor for the j-th dimension in the i-th decoder, Is the design index target value of the j dimension in the i-th decoder.

Description

Turbine blade cooling structure parameter recommendation method based on generative model Technical Field The invention belongs to the technical field of turbine blade cooling design of aeroengines, and particularly relates to a turbine blade cooling structure parameter recommendation method based on a generated model. Background In the current cooling design process of the turbine blade of the aeroengine, the initial cooling structure parameters are directly given mainly by means of manual experience, and then the cooling structure parameters are optimized by adopting an iterative optimization algorithm, so that the design indexes are achieved. In the face of future more advanced aeroengines, the working conditions of turbine blades are more severe, design indexes are more severe, structural layout is finer, the current 'empirical' initial design method is extremely likely to deviate from a feasible design area, time and difficulty of subsequent 'iterative' optimization are increased, each design activity is independent, and historical design results and data cannot be reused. Therefore, in order to improve the cooling design efficiency and design capacity of the turbine blades of the aero-engine in China, development of a high-efficiency and continuously evolving turbine blade cooling structure parameter recommendation method is needed. Disclosure of Invention In order to solve the technical problems, the invention provides a turbine blade cooling structure parameter recommendation method based on a generated model, so as to solve the problems in the prior art. In order to achieve the above object, the present invention provides a turbine blade cooling structure parameter recommendation method based on a generative model, including: Constructing a training data set of turbine blade cooling design indexes corresponding to different working conditions and structural parameters based on a numerical simulation method; constructing a plurality of decoder models based on a neural network architecture, training the decoder models through the training data set, and generating a plurality of trained decoder models; Constructing an encoder model based on a neural network architecture, training the encoder model through the training data set, and generating a trained encoder model; Connecting the output of the trained encoder model with the input of the trained plurality of decoder models to generate a turbine blade cooling structure parameter recommendation model; And generating cooling structure parameters of the design target under the current working condition based on the turbine blade cooling structure parameter recommendation model. Preferably, the operating conditions include, but are not limited to, main flow Reynolds number, turbulence, density ratio, blow-off ratio, and vane internal and external pressure differential; The structural parameters include, but are not limited to, impingement hole diameter, gas film hole incidence angle, aspect ratio, hole spacing, and row spacing. Preferably, the process of constructing the plurality of decoder models based on the neural network architecture comprises constructing the plurality of decoder models from working conditions and structural parameters to cold flow and blade comprehensive cold efficiency design indexes based on the neural network architecture. Preferably, the decoder model has the expression: Y=G(X)=G(Re,Tu,DR,M,ΔP,di,df,α,l/df,P/df,S/df,...); Re is the main flow Reynolds number, tu is the main flow Reynolds number, DR is the turbulence degree, DR is the density ratio, M is the blowing ratio, deltaP is the internal and external pressure difference of the blade, di is the diameter of the impact hole, df is the diameter of the air film hole, alpha is the incident angle of the air film hole, l/df is the length-diameter 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. Preferably, the working condition and structural parameters of the encoder, namely the cold air flow and the comprehensive cold effect design index of the blade are the same as those of the decoder. Preferably, the expression of the loss function of the encoder is: wherein YR is the performance of the cold air flow output by the decoder, the comprehensive cold effect of the blade and the like, Y is a design index, and K represents the dimension of input data. Preferably, the process of connecting the output of the trained encoder model with the input of the trained plurality of decoder models comprises: And taking the trained encoder output as a hub to connect a plurality of decoder models in parallel, and forming a multi-design-target collaborative recommendation unified paradigm through a training method of associating a plurality of design targets by a loss function. Preferably, the unified paradigm of multi-design-objective collaborative recommendation is: Where k is the total number of decoders, ni is the dime