Experiments and numerical simulations of the dynamics of an R.O.V. thruster during maneuvering

Abstract

Propeller dynamics have typically been ignored in controller design, lumped into the category of 'unmodeled dynamics.' This is acceptable for propellers operating at constant speed in relatively uniform flows. Operational parameters of small remotely operated vehicles and autonomous underwater vehicles require a great deal of transient operation of the propellers. This and the small mass of the vehicles make the dynamics of the propellers a significant factor in vehicle control. Expanding roles of these vehicles require improved control and therefore improved understanding of the dynamics of the thrusters during maneuvering. In this thesis, the dynamics of maneuvering thrusters were explored through numerical simulation and experiments. Vortex lattice propeller code developed for use with nonuniform inflow was adapted to incorporate varying propeller speed and inflow velocity. Test runs were made using a three bladed propeller. Experiments were preformed on a thruster from the ROV Jason using the water tunnel at the Massachusetts Institute of Technology. The thruster incorporated a ducted three bladed propeller. Runs were made using step changes in shaft velocity as well as sinusoidal perturbations on top of steady state velocities. Runs were also made incorporating fully reversing propeller operation. Experiments were done with and without the duct in place. The numerical simulation and experimental results showed that accelerating propeller angular velocity created higher thrust values than steady state propeller operation at the corresponding instantaneous shaft velocity. Decelerating angular velocities created lower thrust values. This is attributed to a lag in the local flow velocity due to the momentum of the fluid. For the case of the accelerating propeller, the angle of attack at the blade is higher, resulting in higher lift force and greater thrust. Errors in the numerical code at low advance coefficients prevented direct comparison of numerical code results to experimental results.