Reducing SolutionTime for Hull CFD Analysis / Resistance Curve

Early this month I was in HSVA for an EU project. While I was there, I learned a way (a new way for me) to evaluate the resistance curve. Rather than measuring force for each velocity (along the resistance curve) they accelerated and decelerated the hull with constant acceleration.

In the acceleration region, because of the inertia effect, the force on the hull (Fx) is bigger than the actual resistance and it is the opposite in the deceleration region. Interpolating the effect all over the velocity domain eliminates the effect of acceleration and yields the actual curve for resistance.

This procedure can easily be implemented in CFD analysis.

For this study, I used a model scale hull form and evaluated the resistance between 0 m/s & 2.278 m/s. For the first part of this study, I did a few resistance simulations for comparison using the conventional way, for the second part I used the procedure described above. Considering the default procedure of Numeca Fine/Marine already includes an acceleration procedure for a regular resistance calculation, I used Fine/Marine for this study.

Mesh

CFD calculations were done using 1 million elements (created using Numeca Hexpress). A representative picture is given below;

Mesh Quality;

• Min. Orthogonality: 11.84 (deg)
• Max. Aspect Ratio: 190

First Part

Individual calculations were done for 0.976 m/s , 1.301 m/s and 1.789 m/s. Simulations were done according to the default procedure of Fine/Marine (using C-Wizard).

Second Part

The default procedure of C-Wizard is altered as below;

The acceleration and velocity plots for the hull are as below;

Results

A comparison between experimental results and individual analysis and acceleration and deceleration analysis is given below. The resistance values are omitted intentionally.

Unfortunately I could not find the time to do a mesh independence study. Hence comparing the error rate between two CFD methodologies and experimental results seem useful to me.

As one can see from the graphs above, the errors are similar between the two methodologies. I believe this indicates that the error is due to mesh density or case set-up (which is not the subject of this study).

Time Spent

Considering a resistance curve one should do at least 4 individual analyses. Time spent using this procedure (given below)?reduces the total time spent by half.

Thanks

I would like to thank Mr. Heinrich Streckwall, Mr. Jan Lassen and Mr. Tom Lücke from HSVA for taking the time to share their experience and broaden my vision.

Also thanks to Mr. Ahmet Soydan for post-processing the results.

Aim

The sole aim of this study is to learn from others' experiences. If you have any experience or insight regarding this subject please contact me.

This article was originally written and published by our CFD Engineer Alaz Talay.