| Abstract | A series of experiments was conducted using a laboratory flume within the University of Ottawa cold room to measure accumulation of frazil ice and ice jamming for open channel flows over various bed materials and different water depths. The water surface level was monitored using four ultrasonic devices mounted above the experimental flume, while the amount of ice accumulation was determined through image processing techniques. To implement the image processing, a deep learning semantic segmentation technique capable of identifying surface ice was employed. To calculate the volume of frazil accumulation in each experiment, the obtained surface area during the residual stage, when the water temperature stabilized slightly below the freezing point, was multiplied by the submerged ice thickness. The submerged ice thickness was calculated using a time-based polynomial function, which approximately fits the measured ice lower levels for various experiments. For rough and fully turbulent flows, no surface skim ice was observed, but frazil ice accumulated at the flume’s end. The resulting surface ice propagated upstream at a rate of 0.6 cm/min after the supercooling stage. In contrast, in tests with lower turbulence and higher water depth, a combination of frazil and border skim ice was observed, with maximum ice cover progression rates of 2 cm/min after supercooling was reached. |
|---|
