Turbomachinery aerodynamic performance-blade load optimization method based on machine learning

Abstract

The invention discloses a turbomachinery aerodynamic performance-blade load optimization method based on machine learning, and the method comprises the steps: determining a turbomachinery working fluid, carrying out the parameterization of a turbomachinery, obtaining an optimization process input variable and an optimization target, and determining the empirical design space of the input variable; performing Bayesian optimization sampling on the turbomachinery in the empirical design space of the input variable according to the optimization target, selecting a working fluid in the optimization sampling process, calculating to obtain an optimization target value, and storing all Bayesian optimization sampling data; constructing a Unet-CNN neural network, and carrying out network training; performing random sampling on optimization process input variables in an empirical design space, constructing a geometric model to perform unsteady CFD calculation, performing post-processing to obtain a high-performance test set and a low-performance test set of the Unet-CNN neural network, and testing the Unet-CNN neural network; and enabling the Unet-CNN neural network passing the test to be used for turbomachinery optimization, and obtaining the optimal turbomachinery structure. According to the method, the cost and time consumption for constructing the proxy model can be greatly reduced.

Description

technical field [0001] The invention belongs to the field of energy and power, and in particular relates to a machine learning-based aerodynamic performance-blade load optimization method of a turbomachinery. Background technique [0002] With the development of industrialization, climate change and the continuous deepening of human awareness of environmental protection, the efficient and clean utilization of energy has become a research hotspot for institutions and researchers around the world. my country's long-term goals of "carbon peak" and "carbon neutrality" put forward higher requirements for energy conservation and emission reduction in my country's industrial field. Turbomachinery is the core component of the power cycle, and its aerodynamic performance will directly affect the power and efficiency of the cycle system. At the same time, the three-dimensional flow inside the turbomachinery has strong unsteady characteristics, which causes periodic pressure fluctuati...

AI Technical Summary

Problems solved by technology

However, the direct combination of CFD solution and optimization algorithm has poor adaptability. When the optimization algorithm is replaced, CFD calculation needs to be performed again, which greatly increases the calculation cost and time consumption.

Although the method of using the proxy model can easily replace the optimization algorithm, it initially requires a large number of CFD calculation operating points to build a global high-precision proxy model, which also takes a long time

In addition, most traditional surrogate models are only constructed for optimization objectives, and it is difficult to obtain turbomachinery flow field information during the optimization process

[0004] To sum up, traditional turbomachinery optimization methods have disadvantages such as heavy workload, long design cycle, poor adaptability, and poor physical interpretation. It is urgent to develop an efficient, accurate, and interpretable turbomachinery optimization method.

Images