Author(s): Farzad Asgari; Seyed Hossein Mohajeri; Mojataba Mehrain; Mohammad Javad Ostad Mirza Tehrani

Linked Author(s): Seyed Hossein Mohajeri

Keywords: Doppler Velocimeter; Velocity time-series; Flow measurement; Filtering; Fast Kernel Density Estimation

Abstract: Filtering spike contaminated velocity time series, recorded by Acoustic or Laser Doppler Velocimeter (ADV or LDV) in laboratory open-channel flow studies and field measurements, has always been challenging. Despite numerous conducted studies on the velocity time-series signal filtering methods, the importance of the number of spikes of invalid data on the performance of filtering algorithm has not been determined. Indeed, there is still a lack of comprehensive and updated study on the performance of despiking algorithms of the velocity signals. In the present study, a new developed software package for despiking Doppler Velocimeter Data has been introduced. The package is composed of various detection techniques such as Phase-Space Threshold (PST), Velocity Correlation Filter (VCF), Kernel Density Estimation (KDE). A new filtering technique, so called “Three Dimensional Fast Kernel Estimation (3D-fastKDE)” has been developed and employed to detect spikes in highly polluted signals. Implemented replacement algorithms include Last Good Values (LGV), and 12 points cubic polynomial interpolation (12pp). The performance and accuracy of detection and replacement techniques has been explored in this study using different experimental data sets.

DOI: https://doi.org/10.3850/IAHR-39WC2521711920221215

Year: 2022