My first ideas for a sediment supply mechanism revolved around a hopper of dry plastic sand, pulled/pushed into the stream by an auger mechanism turned via a cordless drill. This ended up being a little too ungainly. It was difficult to keep the drill going (charge in the batteries) and it was noisy. I also suspect the auger was grinding the material into finer pieces.
A view inside the Sediment Supply System from Matt Kuchta on Vimeo.
So the next idea was to use a "conveyer belt" to deliver the sediment. Cobbled from LEGOs, this used an electric motor to turn a large rubber band. The rubber band would then convey the sand into the model.
CIMG0858 from Matt Kuchta on Vimeo.
This video also shows the power supply that we used to provide a constant DC voltage. We put guards along the sides of the belt in our final version (along with some miniature operators), which my research student dubbed the "Auto-Sed."
This view shows some modifications we made to the stream channel (more on that later) and how we hooked up the the Auto-Sed. We set the sediment to dump into a funnel, wherein the water was also going. This solved a problem related to the dry plastic not completely getting wet due to surface tension (you end up with lots of little float-y bits down near the drain and less sand in the river)
Graph of sediment discharge related to power supply voltage. Most of the time, anything more than 3V tended to overwhelm the stream and pile up near the funnel. Results were remarkably consistent (data points consist of ~3 averaged measurements).
LEGO Auto-Sed mark 2.5 from Matt Kuchta on Vimeo.
Here's the Auto-Sed Mk 2.5. At this point, only some of our research used this device - we had plenty of other fish to fry, but apart from the large number of moving parts, the design held up rather well. The rubber band doesn't appear to be showing too much wear - but they're pretty standard "portfolio size" that I can get at our local art supply store.