Interpolation of wind pressures using Gaussian process regression

By Xingliang Ma, Fuyou Xu, Bo Chen. Journal of Wind Engineering and Industrial Aerodynamics, 2019

49 citations recorded by the source corpus.

Abstract This study focuses on using limited wind tunnel test records to interpolate surface wind pressures where no measuring taps are arranged or under wind directions that are not measured. Gaussian process regression (GPR), a machine learning method, is first introduced to investigate the issue. For the interpolations of lower-order moments, time histories, higher-order moments, and wind pressure cumulative distribution functions (CDF), several novel GPR-based approaches are proposed. The newly proposed time history interpolation approach can capture the slight variation of wind pressure field with time, owing to the invocation of the time-varying GPR models. Interpolations of skewness, kurtosis, and wind pressure CDFs are attempted for the first time. Taking the surface pressures of a rectangular flat-roof low-rise building as an example, all the proposed approaches are comprehensively discussed, and their interpolation accuracy and superiority are demonstrated by comparing the interpolation results with the measured data and with the outputs from current state-of-the-art approaches.

Subfields: Building and Construction, Environmental Engineering, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/2f72d72eb71a950be8fe45eedefc064785cc033b