Open AccessGEOMETRIC UNCERTAINTY ANALYSIS OF A CENTRIFUGAL COMPRESSOR USING KRIGING MODEL
Author(s) -
Shujiang Li,
Min Huang,
Xuejun Liu
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1043/5/052041
Subject(s) - centrifugal compressor , kriging , gas compressor , aerodynamics , latin hypercube sampling , turbomachinery , computer science , monte carlo method , control theory (sociology) , engineering , mechanical engineering , mathematics , aerospace engineering , control (management) , artificial intelligence , statistics , machine learning
Centrifugal compressor, which is an important turbomachinery, plays a significant role in the gas turbine engine. Various uncertain factors in the manufacturing process can lead to the geometric variations, which often cause noticeable variations in compressor performance. In this paper, a surrogate-based uncertainty analysis method is used to investigate the effect of geometric uncertainty on the aerodynamic performance of a centrifugal compressor. Firstly, a scaled-up version of NASA DDA’s 404-III centrifugal compressor is chosen as the research object. Considering the structural features of the compressor, a parameterized model with 38 independent control parameters is established to define the geometry of the compressor, enabling rapid modification of the geometry. Then, a widely-used simulation method based on computational fluid dynamics is employed to predict the aerodynamic performance at a series of specified geometries generated using Latin Hypercube Design. Next, based on the sample data, two Kriging models are built to emulating the complex functional relationship between the control parameters and the performance of compressor. Finally, a Kriging model-based Monte Carlo simulation method is used to perform the geometric uncertainty analysis. After specifying a reasonable probability distribution for each control parameter, the statistical distributions of the compressor performance are obtained and analyzed. The analysis results yield new insights for compressor designers.