|Institution:||University of Waterloo|
|Keywords:||Hydraulics; Turbulence; Sediment transport; Sediment mobility; Impulse; Force impulse; Despiking; ARMA; Autoregressive moving average models; Particle tracking|
|Full text PDF:||http://hdl.handle.net/10012/9060|
The link between hydrodynamics and morphology of river bedforms is an ongoing field of research. Existing studies have been unable to provide a complete understanding of the physical conditions required to initiate sediment movement, partially due to the complexities in accounting for turbulent fluctuations within a fluid. Flume experiments and velocity data filtering algorithms were completed to improve the methods available for use in future studies investigating the role of turbulence in particle mobility under non-uniform bed conditions. High resolution velocity profiling instruments, such as the Vectrino II™ manufactured by Nortek AS, have enabled a new generation of turbulence studies, but data filtering methods have not kept pace. A new method was developed that uses autoregressive moving average (ARMA) models for spike detection, replacement and classifying cell quality. A series of flume experiments were completed to test the concept of impulse under non-uniform bed conditions. Impulse is hypothesized to be a superior predictor of transport potential by accounting for the magnitude and duration of individual turbulent events. Six experiments were completed under low flow conditions with local bed slopes ranging from -2.7° to +2.7° relative to the overall bed to determine the spatial distribution of impulse events as well as the role of particle size and specific gravity. Another experiment was completed at high flow to test impulse with respect to the movement of a spherical particle placed at 42 sampling locations along the bed of the flume. Lastly, a Proof of Concept is presented for synchronized measurement of fluid velocities and image recordings of particles along the bed. Automated algorithms track a particle of interest and determine the precise time of movement for correlation with the velocity time series. Overall, ARMA models offer a promising approach for filtering velocimetric data. Kurtosis of the model residuals is revealed to be a robust cutoff parameter within the despiking algorithm. Flume experiments demonstrate that impulse events are strongest immediately downstream of the transition to zones of decelerating flows, and strengthen toward the sidewalls of the flume. A positive correlation is noted between impulse and particle mobility, but additional testing is recommended to determine the precise turbulent events required to move a range of particle sizes under varying hydraulic conditions. The developed Proof of Concept should facilitate this type of experimental study.