Hi, so it is a little more complex than that. If we were to assume no load on the turbine axis, you would be looking for a moment in the Y direction to converge around 0. But most turbine blades drive something, so you need to know the resistance torque, and you should instead converge on that value.
Also, it’s quite hard to get an exact RPM that matches your force, you would be better off testing say 10 points, and interpolating the location the RPM curve crosses the torque line.
Sometimes fluctuations are inevitable, but other times they are caused by bad convergence, and 9/10 of bad convergence cases are caused by bad mesh. In this case, you have a very coarse setting and get a mesh of around 1.2 million. If I were tackling this, I would look at where in my mesh I get high cell density, and defeature or simplify them.
Best,
Darren