Bucket-by-bucket overshot wheel driven by a sluice gate, with a grindstone load, torque and efficiency curves.
GraphsTorque vs wheel speed: water (blue), load plus friction (red), live dot, predicted steady speed (yellow)
Power delivered by the water vs wheel speed (white dot: now)
Wheel speed over the last 60 seconds
The science
Sluice flow. The gate opening a under a pond head h acts as an orifice: Q = Cd w a √(2 g h) with Cd = 0.62 and channel width w = 0.94 m. In the flume the water leaves the lip at critical depth yc = (Q² / (g w²))1/3 and speed Q / (w yc).
Overshot wheel. Water falls into the bucket that sits under the lip (jet angle 11 deg from the top). Each of the 24 buckets keeps its own volume vi. A bucket moving faster than the jet (rim speed above the flume speed) catches only the fraction vjet/(ωR) of the flow. A bucket holds all its water until its mouth tilts past 95 deg from the top and then loses it linearly until it is empty at 172 deg (the capacity function). The torque is the sum of the weights acting at the water's centre radius rc = 2.3 m: τw = Σ ρ g vi rc sinφi. Slow wheels fill each bucket more, so τw falls as the speed rises, while the power τwω has a maximum.
Dynamics. I dω/dt = τw - τstone tanh(ω/0.06) - τfric tanh(ω/0.02) - cω, with I = 0.7 m R². A pawl stops the wheel from running backwards, and a wheel with τw below the breakaway torque stays stalled. The pit wheel (36 teeth) drives the 12-tooth pinion, so the grindstone turns three times faster than the wheel. Efficiency is stone power divided by ρ g Q H with the 5.5 m drop from the flume to the tail race.
Independent checks. (1) The time-stepped bucket model is run at fixed speeds and its mean torque is compared with a direct quadrature of one bucket's life. (2) The steady speed from a long ODE run is compared with the root of τw(ω) = τresist(ω). (3) The live volume and energy ledgers must close: water in = stored + spilled, and work from the water = work into the stone + friction + kinetic energy.
Honest simplifications. No impact torque from the jet, no water inertia in the buckets, a fixed capacity curve instead of full free-surface geometry, constant discharge coefficient, no gear or bearing losses beyond the friction terms, and a rigid wheel. The visible water in each bucket is a level box scaled to the computed volume; the particles are decoration driven by the computed flow.
What to try: open the gate fully with no load and watch the wheel race, then add load until it stalls; find the speed that gives maximum power on the right-hand graph; raise the mass to see slow start-up; hide half of the buckets and see how the filling changes.