Super interesting thread Paul!

How much does the stationary solver's output differ from the transient, i.e. if you were to model an SJR now would the output be comparable to the previous runs, or would it make sense to rerun everything with the stationary solver?

Either way I would love to see how an SJR actually performs, in this digital wind tunnel. The wingsail results were hardly surprising but the SJR has been somewhat of an enigma. We already got a cool surprise (for me at least) with the bad tack being reversed with flat vs. cambered. I'm curious to see how a split sail handles the turbulent mast in between the two panels and how the flow reattaches to the aft panel on the different tacks (and AoA). And of course the theoretical L/D compared to other shapes.

And, since this thread hasn't been updated in a while, does it mean that you got busy with other things and if so, can we expect the experimentation to resume at some point?

Finally, what is the software called? I don't think you mentioned it anywhere. Is it public or is it something proprietary that you have privileged access to?

Hi,

I was able to continue some of my CFD calculations for junk rig profiles (2D). In the meantime, I finetuned the solver and let it run long enough.

David, your wingsail was next. I took your drawing as you posted it, and made a profile out of it. The results (at 10 kn) show what is expected:

Recently, I stepped back from comparing the different junk rig types to each other with my 2D simulations. In my view, there are too many variables involved which I don't have at hand: how much camber is there in reality? How is the real profile shaped? Is it the same as designed? How much camber? Is flat really flat? If not, how much camber? The list goes on...

Instead, I focus on variables which can be determined, for instance: mast position, position of max camber, etc... Let's see how that goes.

Also, after the wingsail simulations I switched from the transient solver I used to a stationary solver. This one only gives good results until the flow is fully detached, but is way, way, way faster.

Paul

Anonymous wrote:

Failure and Progress

 David, I did not let simulations run with your provided wingsail profile yet. I want to ensure solid results for the trivial test cases before commencing to more interesting geometries.

Bonjour

I would also be interested on a perticular wingsail profile.

I've already tried a demonstrator of a junkwing. The next step, under slow motion progress is to develop a prototype for Mingming.

I used a very simple profile : I started with a  classical Clarck Y profile. I only kept the extrado. I made a symetrical profile with two extrado. Then I rotated the two extrado until thry crossed at 70% of the profile.It produced a wishbone forward with two tails.

The wishbone part is the forward part of the profile. The tails are the positions, on each tack, of the flexible part of the profile in order to have an assymetrical profile.

To hold the flexible part in position, I use a double sheating system with two points of sheating on each side of the battens.

Eric

Failure and Progress

Short update from the simulation area: Unfortunately, I discovered that some of those simulation results I posted here in the last weeks were taken from not yet converged runs. My fault! It is mainly an issue for AoA’s > 8°. Reason for this: with higher AoA comes more flow separation/ detachment, which creates huge eddies behind the profile. Those simply take longer to resolve numerically than attached flow. It is a pure simulation issue and has nothing to do with real world physics.

Anyways, I let the simulations rerun up to 30 s (instead of 12-15 s as before). See the results below:

Sorry for the graph being a bit cramped. Basically, the data with “30s” in its name is from the rerun simulations.

Good thing is, the big picture doesn’t change. However, there are still some significant deviations. I have to bite the apple and let future simulations run long enough – and double check their convergence.

It also appears that slightly higher AoA’s would be of interest, too. Let’s say up to 20-30° – specially for the flat cut profile, the maximum of the Cl/Cd-line is not yet reached.

Also, after some good nights sleep and some head-scratching, I came to realize what some of you already suggested: For comparison, the Cl/Cd-plot might be not the right choice. What is most interesting for close hauled performance is mainly (or only?) Cl. I’ll go and make some Cl/AoA-plots, next time.

One last thing: A recent conversation with Prof. Graf from the Kiel “Yacht Research Unit” suggested that it might be sufficient to use a steady-state solver. Until now, I used a transient solver (that’s why I talk about computed seconds, i.e. have my simulation run to 30 s – which is not the time it takes to calculate that simulation). If that would speedup things while still being accurate, I’ll go for that! I'll do some comparison runs, then lets see! More hard work for my laptop to come…

Cheers,

Paul

PS: David, I did not let simulations run with your provided wingsail profile yet. I want to ensure solid results for the trivial test cases before commencing to more interesting geometries.

".. the line was always there, it is just so thin that you could not see it in that view ...I just highlighted it for you..."

"I can imagine that this sounds wierd...  "   

No, perfectly clear explanation - thanks.