Modified kernel density-based algorithm for despiking acoustic Doppler velocimeter data

• Published in: Measurement : journal of the International Measurement Confederation 2022-11, Vol.204, p.112043, Article 112043
• Main Authors: Chen, Yue, Yang, Wenjun, Lin, Haili, Li, Bin, Jing, Siyu
• Format: Article
• Language: English
• Subjects: Acoustic Doppler velocimeter; Interpolation strategy; Spike detection; Turbulence measurement; Velocity data post-processing
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2022.112043

•Distribution characteristics are useful in de-spiking velocity data.•The proposed filtering method preserves valid data to the greatest extend.•Power spectra calculated from altered fluctuating velocity signals changes subtly.•Consideration should be taken in choosing the suitable spike-replacement strategy. Acoustic doppler velocimeter (ADV) has attracted significant attention, especially the research on its data post-processing, which is a vitial step before calculating turbulence statistics. This work processes the acoustic correlation velocimeter (ACV) time series, which have an identical characteristics with ADV data, sampled in open-channel flows utilizing current algorithms combined with a proposed modified kernel density-based algorithm (mkde) that preserves valid data to a larger extent. Trials investigating several interpolation strategies among various velocity series with different data quality demonstrate that without any replacement, the power spectrum density of data series with up to 22% spikes satisfies the Kolmogorov −5/3 law in an inertial subrange. Moreover, in highly contaminated velocity series presenting more consecutive spikes, linear and the last valid data interpolation turned out to be more robust. This paper shed light on dealing with outliers in acoustic Doppler velocimeter data.