The genetic algorithm-radial basis function neural network to quickly predict aerodynamic performance of compressors
Authors:
- 13 August 2020
- journal article
- research article
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering
- Vol. 235 (5) , 537-552
Abstract
In this paper, compressor aerodynamic performance has been predicted based on throughflow theory, combined with a surrogate model, which is a combination of the Genetic Algorithm (GA) and generalized Radial Basis function (RBF) neural network. And the predicting results have been compared with those from the traditional models and spanwise mixing model, which still widely be used to predict the aerodynamic performance. We first predicted the deviation angle and total-pressure loss coefficient (TPLC) by the surrogate model, and then using these two intermediate variables connected the model with throughflow theory. The pressure ratio and efficiency, representing the compressors’ total performance parameters, are predicted and compared with experimental data. In order to increase the accuracy of prediction, a data augmentation method based on the piecewise cubic Hermite interpolation (PCHIP) algorithm is introduced to enlarge the training database. At the same time, considering the vast differences of deviation angle and loss in different working conditions as well as aerodynamic and geometric differences of rotor and stator, the database and the network should be split into six components based on the choke, the normal and the stall conditions as well as rotor and stator. Then, the performance curves of pressure ratio and efficiency can be determined by an iteration process. The predicting results are compared with experimental data, which shows that the surrogate model matches experiments much better than those from the traditional models and spanwise mixing model.Keywords
Funding Information
- National Science and Technology Major Project (017-II-0001-0013)
- National Natural Science Foundation of China (51676162)
This publication has 26 references indexed in Scilit:
- Secondary Flows, Endwall Effects, and Stall Detection in Axial Compressor DesignJournal of Turbomachinery, 2015
- A New Loss and Deviation Model for Axial Compressor Inlet Guide VanesJournal of Turbomachinery, 2014
- Development and validation of a new universal through flow method for axial compressorsProceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy, 2010
- Inlet Boundary Layer Effects in an Axial Compressor Rotor: Part II—Throughflow EffectsJournal of Engineering for Gas Turbines and Power, 1985
- Spanwise Mixing in Axial-Flow TurbomachinesJournal of Engineering for Power, 1982
- Rotor Wake Mixing Effects Downstream of a Compressor RotorJournal of Engineering for Power, 1982
- Structure and Decay Characteristics of Turbulence in the Near and Far-Wake of a Moderately Loaded Compressor Rotor-BladeJournal of Engineering for Power, 1981
- Loss Sources and Magnitudes in Axial-Flow CompressorsJournal of Engineering for Power, 1976
- The Off-Design Analysis of Axial-Flow CompressorsJournal of Engineering for Power, 1967
- Shock Losses in Transonic Compressor Blade RowsJournal of Engineering for Power, 1961