Energy-based wind field reconstruction and extrema estimation for downburst events

In civil engineering, an accurate characterization of wind loads is fundamental to wind-resistant design and wind effect assessment on structures. Reliable estimation of wind field extremes remains challenging. Existing methods typically require large sample sizes and incur substantial time and economic costs, limiting their practical applicability. To address these issues, a concept of time-varying energy (abbreviated as TV energy) is proposed to guide the reconstruction of wind speed time histories of downburst events. First, field-measured nonstationary wind speed data during the downburst are utilized to examine how time-frequency characteristics of fluctuating wind affect structural responses, and to quantify the correlation of statistical characteristics in the time-frequency domain. Then, a lightweight algorithm combining a mathematical model and a machine learning is proposed for wind field reconstruction and extrema estimation of downburst fluctuating wind. The wind speed and the TV energy are first reconstructed by the Kriging-based sequence interpolation based on data at measurement points; the resulting reconstructed wind speed is referred to as primary wind speed. Then extrema model is proposed according to the extrema correlation between the TV energy and wind speed time series. The estimated expectation and variance are used for extrema estimation, adjusting the extrema and their occurrence time of the primary wind speed at unmeasured points. A 10-storey steel frame structure under field-measured downburst wind is employed to demonstrate the effectiveness of the proposed algorithm, thereby providing a more reliable method for nonstationary wind field reconstruction.