About RAFT Solution Process

[hymdrj]

Hello,
I am studying the RAFT-based WEIS (Level 1) program. And I have some questions.

• Evaluate structure mass and hydrostatic characteristics (WISDEM)
• Model mooring system and solve unloaded equilibrium position
• Evaluate linear hydrodynamic coefficients, including excitation at select sea states
• Evaluate aerodynamic coefficients at select wind speeds
• Apply mean loads and solve for mean offset position
• Reevaluate aerodynamic coefficients to account for tilt
• Compute rotor response coefficients (including effect of control) • Linearize viscous drag excitation and damping at given sea state
• Solve for system response
• Iterate over previous three steps until response convergence

In the RAFT calculation process when we talk about convergence, does it refer to checking whether the differences in the numerically computed values converge while iterating to find the solution of an equation? Also, should this difference (relative error) be within 0.01?

Furthermore, I am currently reviewing the VOLTURN example of RAFT. If I want to change or add the brace positions in VOLTURN for the calculations, which part of RAFT and WISDEM should I modify?

Any advice would be appreciated.

Thank you for reading.

[giannikosk]

Hello @hymdrj , I am no part of the team working on raft, but i have worked a bit with raft as a standalone module (not as part of wisdem) and i think i can answer your questions.

Regarding your first question and the aspect of convergence, solution changes through the iterations because of the linearization of the viscous drag terms, those being the viscous damping (linearized through a Borgman linearization procedure) and the drag excitation force. If you check the code for the solve dynamics function i think you will understand what i am talking about. In the Borgman linearization you need to apply the standard deviation of the relative velocities in the terms of the linearized drag. However, since this relative velocity response is part of the solution that you are looking for in the first place, you cannot know at the same time the standard deviation of the response to be solved for and the response. Because of this you are doing this iterative solution until the response converges, in which case the standard deviation used for the damping terms, is the same as the standard deviation of the response you are looking for. This iterative process also utilizes an under-relaxation factor approach when combining the previous and current solutions in the convergence procedure, which you can modify to have a faster convergence if you wish to.

Regarding the Volturn example, if you wish to change parts of the geometry, this should be done through the input .yaml file that you are feeding to raft. There in the stations section you can modify various parameters of the members (and your pontoons) including shape size location and added mass and damping coefficients.