Structural surface wind pressure prediction method and system based on multi-feature gaussian process regression

Abstract

This invention provides a method and system for predicting wind pressure on structural surfaces based on multi-feature Gaussian process regression, belonging to the field of wind engineering and structural safety technology. Firstly, by constructing a spatiotemporal multidimensional feature set including the mean of the wind pressure coefficient, the standard deviation of the wind pressure coefficient, and the first-order slope of wind pressure variation over time, the invention elevates the simple spatial interpolation problem to a spatiotemporal dynamic regression task. This enables a more accurate establishment of the nonlinear mapping relationship between interpolated features and actual wind pressure values. Secondly, through iterative K-means clustering for physical partitioning, the invention globally and evenly restores the details of wind pressure distribution on the structural surface. Finally, the multi-feature Gaussian process regression model effectively captures the high-frequency pulsation characteristics and extreme value mutations of the wind pressure signal, overcoming the shortcomings of traditional models that predict too smoothly. This achieves accurate assessment of the overall stress state of large, tall structures with extremely low sensor deployment costs, possessing high engineering practical value.

Images