Pressure fluctuations in the Hull example
Hi, I was going through the hull case in examples/InletOutlet/07_CurrentHull as I aim to simulate a case similar to the one provided. When I output pressures in the post-processing, I notice that there are fluctuations as shown in the screenshots below. Are these fluctuations mainly due to the boundary conditions employed here, together with the size of the domain? Or are these due to something else entirely? I'm mainly concerned with how these fluctuations can have an impact on the force coefficients.
Top side:
Bottom side:
Comments
@taffo
Hi @taffo could I have an update on this? Thank you.
Dear,
Even in varying the dimensions of the tank and its depth, I could not retrieve the classic wave angle on moving bodies at the surface of water. You can check the bow wave angle, on the test case
I remark also that, even when increasing the tank's width and length, this bow wave tends to develop with a amplitude far greater than what we expect from comparison with VOF model we have made to cross check. It may be a numerical error due to BCs.
To go further, one would need a BC that we could use to simulate infinite water depth, to check with classic analytical formulas for moving bodies at surface.
For now, I would say that this test case if very "peeky", and cannot simulate practical cases for now. The computed wave drag is far what we can get from simple analytical models. (when not using the geometry nor the proportion of the test case, because it would lead to wrong forces computations. You should use a correct tank geometry, in relation with the Froude Number.)
Interesting points Alan! Could you touch on which analytical models you would use to do to get the drag force output?
And how would you determine the most fitting size of the tank (other than the infinite depth point)?
This case should return much better results in DualSPHysics if it was setup using mDBC, but that can be difficult to do for now.
Kind regards
Dear,
To be precise, the analytical models I use most frequently are all obtained by regression of tank test campaigns. There is no real analytical model that can be used to precisely predict the drag force curves of any given hull shape. The hull, in the test case example, is a planning hull, whose drag can be well approximated by Stavinsky's method.
Other usefull basic formulas are given in the recommendations of the International Towing Tank Conference https://ittc.info/ You will find specific formulas applicable to numerous hull shapes.
Again, I use the ITTC recommendations for the tank size, but only as reference. I recall down below the more prominent ones. To make a fair trade between my computationnal power and the precision of the results, I limit the lenght of the tank I use in DSP to 10 times the hull lenght for Frounde number <0.4. For greater Frounde number, I choose a lenght of the tank equal to 5 times the hull lenght. For practical use, the drag force computed by this method is mitigate by the analytical computations done aside.
"Most tests of types (a) and (b) are carried out in a towing tank with a length of 35 times the ship model length and more. A mean value for the model length to tank width ratio (L/b) is 0.47 for stationary straight line tests, and somewhat smaller (0.42) for harmonic tests."
"Tests in deep water should be performed with a depth to draft ratio that is large enough to be free from shallow water effects. Referring toIMO (MSC/Cir 644), a minimum value of h/T = 4 is considered as acceptable. This figure, which accounts for practical issues of full scale trials,must be considered as a strict minimum for deep water model tests. The critical speed is defined as (gh)1/2. In deep water the test speed should be below 50% of the critical speed. For shallow water tests (h/T < 4) the depth should be scaled correctly; this may impose a re-striction on the maximum draft. At very small h/T, the vertical variations of the tank bottom should be less than 10% of the under keel clearance, which may determine the minimum draft."
I have not been through the mDBC. I understand that this configuration will give access to slip conditions, which is great, because it would lead to great improvements. Though it is not implemented yet, isn't it ?
For now, I will try to set up a case with mDBC, to see the gain of avoiding the GAP and better pressure calculations at boundaries. Thanks for the advice.
Thanks for the detailed explanation!
I am happy to hear that your "analytical solution" was in fact experiment data, I thought I had missed a key point in my studies..
In regards to what I do know about, mDBC, then yes, indeed when it becomes possible to use mDBC and set surface velocity to zero, i.e. no-slip condition, DualSPHysics as a software will improve dramatically. Now it will be possible to actually simulate the flows much more physically accurate, instead of the default "pseudo no-slip". I am looking forward to it my self! :-)
But still, indeed using mDBC for the hull shape should improve results! It will take some time, it will be painful, but it should be doable for the hull.
Kind regards
Hi,
What you're seeing is the effect of the boundary conditions used in this case. While robust in avoiding particle penetration and give an overall flow resemblance to the real case, DBCs are not suitable for physically-meaningful results, ESPECIALLY when pressure is the main target.
The other option you have is mDBC, which comes with its difficulties (this is not a simple geometry and you have to guarantee full kernel support for the fluid in the vicinity of all of the boundary section). With careful analysis and investigation, you may find a resolution that gives you some improvements, however know that this problem is begging for variable resolution to be properly resolved, and VR is not available in DualSPHyiscs (yet!)
A small comment on the Kelvin angle: I was able to retrieve the right angle when studying this type of problem some time ago, it's a matter of having a much larger domain. What was used here was merely a small test tank to show how the hull responded when the Chrono engine was activated.
Best,
Angelo
The other part of Variable Resolution, is multi-GPU support.. I hope you could write "(yet!)" to that too hehe..
But really interesting (and surprising!) to hear that DBC were never meant to be used for physical results and glad that we have a better method.
Kind regards