[supermanks]

Hello all,

I am currently running simulations for a rigid-particle driven in a channel-flow (pressure-driven, laminar, 2D) using the 6DOF solver and dynamic-mesh capabilities in Fluent. I have been using the absolute frame of reference (laboratory frame) thus far, and to capture the long-term motion of the particle, I have had to use large channel-lengths, and this is time-consuming. Now I am testing out a non-inertial frame-approach where my reference frame moves with the particle ONLY in the x-direction, while the y-direction and the rotational direction remain in the absolute frame. Here, x-direction is along the channel, y-direction is perpendicular to the flow/channel-length, and the rotational axis points perpendicular to the channel-plane according to the right-hand rule. While so shifting the frame to the particle, there is a source-term for the x-momentum equation, which reads "-rho*a", where "rho" is the fluid density and "a" is the particle's instantaneous acceleration. In addition, the channel-walls are now moving backwards each time-step equivalent to what would have been the instantaneous horizontal velocity of the particle in the absolute frame-case. I implemented both the source-term and the moving wall conditions using relevant UDFs (DEFINE_SOURCE, DEFINE_PROFILE, and, DEFINE_EXECUTE_AT_END) and I performed all basic checks to see if they were indeed being reflected in the solution process. The left and right boundaries are periodic with an imposed pressure-gradient in both the reference-frame cases. The initial condition is a steady-state flow-field solved for by fixing the particle, and I later release it for the time-evolution of trajectories/velocities. However, I see that the trajectories are quite different for the absolute vs. reference frame cases, although the trends of the particle rising-up are reflected well in both the cases. I do not see where this difference could be arising because physically these cases should be equivalent according to what I'd thought. The only idea I can think of is the order in which I impose the source-term/wall-velocity. At the first time-step, the source-term and the wall-velocity are taken to be zero, which means I'm neglecting any acceleration the particle might incur due to the large drag/lift from the steady-state solution. These quantities are passed to the solver only after the first time-step, so that is probably introducing a noticeable difference in the trajectories, and at the same time, the effect of the first time-step is not so large that the predicted trends in the motion are severely violated. I am currently working on this idea and hopefully that solves that my problem, but I'd appreciate any help/pointers/ideas that you can share about the discrepancy in trajectories, seeing as I've not been able to find many similar threads on the forum.

Thanks a lot!