Interesting results. Just two comments:
- 1) The negative mean displacement is expected. The tower displacement in FAST is defined as positive in the nominally downwind direction, but the overhanging weight of an upwind rotor will cause the tower to displace slightly upwind in the absence of aerodynamic loads.
2) For the simulation with only the second fore-aft bending mode enabled, a tower-top displacement of 0.5 m is very large (normally I would expect much less excitation/displacement of the second modes than the first), so, I would guess the nonlinearities in the response are likely the result of geometric nonlinear terms (e.g. the axial shortening) that are modeled in FAST. The geometric nonlinear terms will have a smaller effect for smaller displacements.