[sreenath1994]

Respected Community Members,

I am a novice trying to apply CFD for studying water on deck problem of a ship running in high waves. The problem at hand is to compute the time history of the roll moments caused by the water flowing across the deck. I got some results for the unsteady simulation.

I am trying to do the grid convergence study now for the purpose of verification. My question is, which variable should I use for the grid refinement study. The quantity that I am interested in is the time history of roll moment. In some of the papers which I have read, they talk about the convergence of an integral quantity (eg. drag) or a local flow variable. But in my case the simulation is transient, the moments caused by the deck water flow continuously change.

How should I proceed with the verification of my results ?

For reference I am putting a time history of the deck water moments and an animation of the simulation.

Animation Link https://drive.google.com/file/d/1UuF...ew?usp=sharing

Thank you for your support.

[flotus1]

Using the quantity you are actually interested in is the best way to do it in my opinion.
For a transient multi-phase simulation, you won't get the perfectly neat kind of convergence you may have come across in literature.
A plot with time as the x-axis, and several graphs of the deck water moments obtained with different time steps or cell sizes should be sufficient.

[LuckyTran]

Always start with your sought after parameter(s) because it is irrefutably the quantity you are interested in.

If you're already aware of the grid convergence index (or anything similar to it) you can compare the result at every time-step and calculate the index at every time-step. This gives you error bars at every point in the plot and then you can say whatever you want regarding the mesh sensitivity.

Note that you have to do a very similar exercise to show time-step sensitivity, i.e. convergence with respect to physical time-step size deltat

[flotus1]

Depending on how meaningful the initial conditions are, and how sensitive the simulation is with respect to these initial conditions, you might have to omit the initial transient phase from convergence studies.
Then you only compare the results once a statistically steady state is reached. Time-shifting results obtained with different grids/time step sizes to get a better comparison is allowed here.

[sreenath1994]

Thank you very much @Alex and @Lucky, sir for the valuable suggestions. I now got an idea as to how to proceed.

[FMDenaro]

My first idea is not a suggestion but a question: what formulation are you using ? Depending on that also the variables you solve have a different meaning to be considered.

[sreenath1994]

@Denaro sir, thank you for the question. Segregated flow model is used, VOF multiphase model for capturing the free surface, K-Epsilon turbulence, wall functions are used and overset grid is used for the hull motion. I attaching a list of all the physical models used in the simulation. https://drive.google.com/file/d/1Vss...ew?usp=sharing

The wave generated inside the domain is forced at the boundaries to the theoretical solutions to avoid reflections. As a first step, in the above simulations the ship is moved using the experimental motion data from the gyro and the resulting flow across the deck is simulated. In the next step the motions will also be computed along with the deck water flow.

I am very sorry, I am not fully aware of the theory behind all the models. I only have the basic knowledge of CFD.

[HankDrokes]

Originally Posted by sreenath1994

@Denaro sir, thank you for the question. Segregated flow model is used, VOF multiphase model for capturing the free surface, K-Epsilon turbulence, wall functions are used and overset grid is used for the hull motion. I attaching a list of all the physical models used in the simulation. https://drive.google.com/file/d/1Vss...ew?usp=sharing

The wave generated inside the domain is forced at the boundaries to the theoretical solutions to avoid reflections. As a first step, in the above simulations the ship is moved using the experimental motion data from the gyro and the resulting flow across the deck is simulated. In the next step the motions will also be computed along with the deck water flow.

I am very sorry, I am not fully aware of the theory behind all the models. I only have the basic knowledge of CFD.

Very interesting. Hope you will share your results with us.