Author(s): Jeffrey A. Tuhtan; Iring Wolf Kosters; Gert Toming; Shokoofeh Abbaszadeh; Stefan Hoerner

Linked Author(s): Wolf Iring Kösters, Shokoofeh Abbaszadeh, Stefan Hoerner, Jeffrey Tuhtan

Keywords: Fish mortality; Downstream passage; Turbines; Pumps; Sensors

Abstract: Hydropower turbines and pumping stations can be hazardous to fish during downstream passage, and blade strikes have been identified as a major cause of mortality. Laboratory blade strike studies using live fish indicate that the relative blade strike velocity is a key parameter when estimating strike-induced fish mortality. However, passive sensors used to record the physical conditions during downstream passage typically measure the pressure, linear acceleration, rate of rotation, and magnetic field intensity, and thus do not directly detect the relative blade strike velocity. To investigate how this limitation may be overcome, we tested a 10 cm long, 2.5 cm diameter cylindrical polycarbonate plastic sensor during a series of laboratory blade strikes ranging from 1 to 10 m/s and obtained the triaxial acceleration time series for each event at 2048 Hz. The novelty of this study is that we trained, validated and tested state-of-the art machine learning regression models on these time series to estimate the relative blade strike velocity. Our findings highlight that under laboratory strike conditions, the relative blade strike velocity can be adequately estimated, with mean average errors at all velocities of 0.81-0.89 m/s. This work is significant because it indicates that passive sensors can estimate laboratory relative blade strike velocities using accelerometer data; however, field validation of the proposed methods is needed to ensure reliable estimates in real-world conditions.

DOI: https://doi.org/10.64697/iahr.proc.ise2026.abs.267

Year: 2026