Kriging based sequence interpolation and probability distribution correction for gaussian wind field data reconstruction

Abstract

Data reconstruction is an important research topic for missing data recovery and data supplement. Spatial interpolation is often used for data reconstruction. The interpolation for time series is usually conducted at each time point by common methods, which is sometimes inefficient. Additionally, statistical characteristics recovering may not be taken into account together during interpolation, which affects the utilization of data in structural design and safety assessment. In this paper, supposing that the data of a potential observation point are completely missing (no historical observation information at all), Kriging based Sequence Interpolation (KSI) combined with probability distribution correction is proposed for data reconstruction. KSI is proposed for two purposes. First, the primary interpolation results can be obtained by KSI, namely global interpolation once. Second, the Probability Density Function (PDF) of missing points can be reconstructed by KSI and subsequently provide a reference to correct the primary interpolation results. The proposed method is verified by both the simulated and field monitoring data, and the results show that the calculation efficiency is greatly improved compared with common methods. The fusion of data reconstruction and PDF reconstruction is novel and the effectiveness is verified via the structural dynamic response analysis. The RMS of structural displacement responses induced by the reconstructed wind load is almost coincide with that by the actual wind load.