Open AccessAn intelligent polynomial chaos expansion method based upon features selection
Author(s) -
Wei Zhang,
Qiang Wang,
Chao Yan
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1786/1/012046
Subject(s) - benchmark (surveying) , polynomial chaos , computer science , airfoil , dimension (graph theory) , curse of dimensionality , selection (genetic algorithm) , function (biology) , polynomial , mathematical optimization , chaos (operating system) , construct (python library) , artificial intelligence , algorithm , mathematics , engineering , geodesy , structural engineering , evolutionary biology , pure mathematics , biology , programming language , geography , computer security , mathematical analysis , statistics , monte carlo method
Polynomial chaos expansion (PCE) method is a common tool for uncertainty quantification (UQ) in fluid mechanics. However, there exists ‘Dimensional Curse’ when the parameters dimension is very high, and large samples are required to solve the PCE function. This would hinder the application of PCE in high dimensions. An intelligent PCE method based on the idea of features selection in machine learning is proposed in this paper. Therefore, only several important features will be selected to construct the PCE function, then fewer samples will be needed to solve the model, and it will be more efficient. Several benchmark functions and an RAE2822 airfoil flow case are utilized to verify the UQ capability of the intelligent PCE. It is proved to be more efficient than the original PCE, with nearly same accuracy.