Tuned Mass Damper expected load reduction

I have been working for a period trying to reduce the loads on Hywind OC3 using a Tuned Mass Damper with the FAST TMD module. I see from the TMD Module references that Lackner et al found a 10% reduction in the load for their setup and I also found a poster (“Can Tuned Mass Dampers help reduce wind turbine vibration”) where they show up to 45% reduction with multiple Tuned Mass Dampers tuned to different frequencies.

I am working on a novel Tuned-Mass-Damping-like damping system where the damper introduce no gravity and I can have huge reaction masses. I have thus removed the gravity from the source code (a_G_O (3) =0) and recompiled the TMD-module for OpenFAST. In similar dry applications you would expect above 90% reduction of stress for such mass ratios that I am using, but I can barely see a change when tuning based on free decay response and previous optimization for Tuned Mass Dampers. I am mostly focusing on the moments generated in the tower base and extract statistics from both time and frequency domain data for runs at 2500 s with stationary wind and turbulent waves at 0/45/90 degrees.

FAST gets instable (see below) when I go above about 4 000 000 kg as a reaction mass attached to the tower base (CompTTMD) using the gravity free damper. Could this be due to numerical issues such as accumulator overflows or division by zero? I am using double point precision.

Does anyone have experience with the Tuned Mass Damper module in FAST and what can be expected of load/stress reduction? Anyone with experience for Hywind OC3?

[code] Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 5, Blade = 1
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 6, Blade = 1
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 7, Blade = 1
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 7, Blade = 2
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 8, Blade = 2
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 9, Blade = 2
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 17, Blade = 3
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 18, Blade = 3
Warning: Turning off Unsteady Aerodynamics due to high angle-of-attack. BladeNode = 19, Blade = 3

FAST_Solution:FAST_AdvanceStates:AD_UpdateStates:BEMT_UpdateStates(node 9, blade
2):BEMT_UnCoupledSolve:There is no valid value of phi for these operating conditions: Vx =
5.5459, Vy = -49.58, rlocal = 28.14, theta = 0.13665, geometric phi = 3.0302
FAST_Solution:CalcOutputs_And_SolveForInputs:SolveOption2:AD_CalcOutput:BEMT_CalcOutput(node 16,
blade 1):Compute_UA_AirfoilCoefs:UA_CalcOutput:Mach number exceeds 1.0. Equations cannot be
evaluated.[/code]

Dear Even,

I’m I understanding correctly that the solution works well for small TMD mass but becomes numerically unstable (leading eventually to large system deflection and an error regarding high Mach number) as you increase the mass of the TMD?

I could foresee this being a problem because of the implicit coupling between ElastoDyn and ServoDyn/TMD and the approach taken to couple these modules in the FAST driver (glue) code. That is, the motions (including accelerations) of the tower and nacelle are passed from ElastoDyn to SerovDyn/TMD and the TMD reaction loads (forces/moments) are passed from ServoDyn/TMD to ElastoDyn. The coupling approach between ElastoDyn and ServoDyn/TMD implemented in the FAST driver (glue) code works well for the example models I’ve seen, but these models have relative small TMD mass compared to the mass in ElastoDyn (and thus, the TMD reaction loads output are only minimally impacted by the acceleration input). As the TMD mass is increased, the TMD reaction loads will be more strongly impacted by the acceleration input and coupling approach would likely need to be changed (similar to how we’ve implemented implicit input-output coupling between ElastoDyn and BeamDyn and between ElastoDyn, HydroDyn, and SubDyn). But modifying the coupling approach would require a fairly extensive change to the FAST glue code.

Best regards,