Steady State Convergence
An important consideration when using any CFD tool to analyze vessel performance is whether a simulation is "converged." Simulation convergence can have different meanings depending on the context. For steady state simulations like the resistance and self-propelled simulations that are incorporated into the Standard version of Orca3D Marine CFD (OMCFD), it is important to consider whether the simulation result has converged with time to a steady state solution. The Marine template in OMCFD automatically selects the duration of the simulation based on various parameters including the hull type and speed. However, it is very important for the user to ensure that the model has reached a steady state by the end of the simulation. This is analogous to running a towing tank resistance experiment and deciding at what point in a given run the model behavior has steadied out sufficiently to record measurements. As described in the Orca3D Help File topic on Convergence of the Simulation, the most effective way to know if a CFD analysis has converged to a steady state is to monitor the Heave, Pitch, and Resistance (or EHP) of Resistance runs, and the Heave, Pitch, Resistance, and Velocity of Powered runs. When you see that the values are changing very little with time, the solution can be considered converged. If these values are changing significantly you should run the simulation for a longer period of time. Below is an exsample of a well-converged solution.
Sometimes you may find that the model Pitch and Heave are oscillating at the end of the simulation. The example below has a small oscillation in Heave and Pitch which is clearer in the second image where we have zoomed into the Heave time history. However, you can see that the magnitude of the oscillation is quite small (about 3mm for a full-scale vessel) and that the average is not changing very much. So from a practical perspective one could consider this simulation to have converged values of Heave and Pitch. If you see that Pitch and Heave oscillation magnitude is significant (and/or if the average value is changing significantly), you may need to run your simulation longer, and for planing hulls you may need to consider that your vessel is porpoising. Keep in mind that, by default, both Heave and Pitch have damping applied, and an investigation into porpoising requires that you set these damping values to 0.
For self-propelled simulations, the Forward Velocity of the vessel is not fixed as it is in a resistance simulation, but is varying until a force balance is reached between thrust and drag. Therefore the time history of Forward Velocity should also be examined in order to determine convergence of the solution. In this plot of Forward Velocity from a Powered analysis, the model has not yet converged at the end of 4s because the velocity is still climbing. Sometimes you may need to zoom into the time history near the end to determine if the velocity is changing significantly. The second plot below shows another useful time history which is the net surge force (the difference between the thrust and drag). At steady state conditions you would expect this net force to be near zero, but in the plot below it is still around 400 lb which is significant for this 32-foot model.
Numerical Convergence
Another very important consideration for simulation convergence has to do with the numerical convergence of the solution. The finite volume solver used in OMCFD discretizes both space and time in order to solve the RANS system of equations. While the Marine template attempts to produce accurate performance predictions with default settings, it is the responsibility of the user to confirm this is the case. This is especially true for grid convergence where it is important to ensure that the mesh discretization of the fluid volume has sufficiently resolved the full flow field. To do this, a traditional grid convergence study should be conducted, usually at a typical or representative speed (the "design" speed) for the design. Below are the results of a grid convergence study for a stepped planing hull showing a plot of total resistance and heave at the CG as functions of grid size (in millions of cells). Several interesting conclusions are evident. First, while the experimental data showed that the flow experienced ventilation behind the transverse step, the CFD simulations did not show ventilation occurring below a cell count of about 2 million cells (corresponding to the OMCFD "Normal" grid size setting). In general we do NOT recommend using grid sizes less than Normal with OMCFD simuations. Next, while the heave of the model was relatively consistent for grid sizes of 2 million cells and above, the resistance continued to decline with increasing number of cells. A detailed review of the simulation results revealed that while ventilation did occur at around 2 million cells, the ventilation and reattachment pattern changed significantly for increasing cell count. Presumably this is responsible for the decreasing resistance. For the purpose of making design decisions, one might conclude that the "Fine" grid size corresponding to about 5.6 million cells sufficiently resolves the flow field and could be used for additional simulations. This case study is a very good example showing the importance of conducting a grid convergence study for your own design.
A second area to consider regarding numerical convergence is the time step used in the solution. As is the case for the grid size, the Marine template used by OMCFD automatically selects an appropriate time step (and iteration count) for each specific simulation. In general we have found that the selected time step is sufficient for steady state simulations (resistance and self-propelled in calm water) without any specific need to look at shorter time steps. However, for more advanced simulations that do not result in a steady state behavior, such as porpoising studies, simulations in waves, or standard maneuvers like zig-zag maneuvers, this deserves closer attention. As the simulation becomes more dynamic in nature, shorter time steps (and larger iteration counts per time step) will generally be necessary to accurately capture the model and flow behavior. As shown below, OMCFD provides a mechanism to do this in the Marine module settings for the Numerical Option. Dynamic simulations should use one of the "(Transient)" options, usually starting with "Medium (Transient)" at first. You should notice that changing that option automatically changes the size of the time step as well as the number of iterations per time step (along with some internal numerical settings). Ideally for a dynamic simulation (such as a simulation in waves) you would run with Aggressive (Transient), Medium (Transient), and Conservative (Transient) and compare the resulting performance predictions. However in practice, the available project time often limit consideration of many variables so it is recommended that you use the most "conservative" setting that the project can afford.
There are of course other types of numerical convergence that can be considered, such as variation of the size of the fluid domain and its effect on simulation results. We have generally found these not to be necessary in typical problem cases, as the Marine template has the logic incorporated into it to select an appropriate domain size.