Testing boat structure with FEA

Discussion in 'Boat Design' started by EddieGreen, Feb 13, 2021.

Tags:
1. Joined: Feb 2021
Posts: 16
Likes: 2, Points: 3
Location: Nice

EddieGreenJunior Member

Understood.
My question re. to the springs set-up, doesnt relate to a spring itself (F=-k*x is enough ) but to the FEM boundary conditions. In my mind (and this is exacerbated by the use of the convenient "plug and play" a.m. solution !).
I.e. either we are talking of a load or a restraint, not a mix e.g. 0 or 1, black or white but not grey. To me a restraint was removing (totally) some degrees of freedom for a selected node, edge or surface to move in one (or several) direction(s).
A spring "restraint" is more subtle and hence less straigthforward e.g. this is like glueing the soles of your shoes to the ground (to use your example) with PU elastic foam instead of superglue. My understanding is that the analysed model cannot move for the FEA concept to work i.e. the 6 degrees of freedom need to be removed and moreover as far as possible from the investigated piece of structure.

In the interesting Delft paper mentioned by @fastwave , with below some excerpts as it covers my initial questions :
So here again i would welcome some clarification.

It says for instance restraining the V1 chainplates to move into the V1 shroud direction, but also the "rig represented by forces"

To analyse the boat structure taking a model of e.g. a hull, primary stucture & appendages, my first inclination would have been to add these forces (calculation or rule of thumb to start with),

but when these a.m. restraints are applied, which loads are used ?
Thanks

2. Joined: Aug 2002
Posts: 15,156
Likes: 914, Points: 123, Legacy Rep: 2031
Location: Milwaukee, WI

gonzoSenior Member

Not at all. You never remove degrees of freedom unless that is what happens in real life. Completely constraining on all axis will give you wrong results. A constraint is a representation of how the actual body behaves. Constraining 6 axis is used to analyze a particular small area. For example, a bracket that will have a cable pulling from it can be constrained on 6 axis to the plate it is welded, to calculated the deformation on the hole a shackle is attached to. Considering a whole boat as an absolutely rigid structure will generate nonsense results.

3. Joined: Oct 2008
Posts: 6,846
Likes: 801, Points: 113, Legacy Rep: 2488
Location: Japan

Im not sure what a.m. means....but anyway.
Take the rigging, you cited. What is attached to it... the sails!
What is the max sea state you wish to design the boat to operate in...that relates to a max wind speed.
From that wind speed and the using worse case max sail area = a force!
That's it.

You decide what force is to be applied as you decide what is the load case and hence what you are analysing!

Then you don't understand how FE works.

Incorrect.
Unless the model is fully restrained, as previously noted, you get an infinite number of possible solutions. All FEM need to be restrained to prevent rigid body motion.

4. Joined: Feb 2021
Posts: 16
Likes: 2, Points: 3
Location: Nice

EddieGreenJunior Member

a.m. : above mentioned ? or already mentioned ? (I am not the one who has put "hell" in hello) sorry for my dubious sense of humor !
Yes this is my initial thought as previously mentioned . In this Delft paper (assuming i read it correctly) they are applying restraints ??
while i would have used the loads/forces coming from the rigging instead

Last edited: Feb 17, 2021
5. Joined: Aug 2002
Posts: 15,156
Likes: 914, Points: 123, Legacy Rep: 2031
Location: Milwaukee, WI

gonzoSenior Member

If you model a boat as a rigid object and don't allow for deformation caused by the applied forces the results will be wrong. That is what happened on the example of the boat that demasted twice. The designer did not take into account that the forces from the rigging change the shape of the hull. They did, as you recommend, use the sheer and keel as rigid bodies. Choosing what parts of the structure to completely constraint and which to partially constrain is the only way to get sensible answers.

6. Joined: Oct 2008
Posts: 6,846
Likes: 801, Points: 113, Legacy Rep: 2488
Location: Japan

This statement is again incorrect and indicates you have no knowledge of FEM and how it works.

It would help if you didn’t comment on things you have no knowledge of, otherwise this shall deteriorate into the usual pissing contest as you dig your bottomless pit of misdirection as it does with Tansl because you are both incapable of conceding the subject being discussed is beyond your knowledge. And then the thread is destroyed.

FEM uses matrix theory to solve the vast number of equations. The generalised form of the final equations:

{F} = [K] {d}

{F} = global nodal; forces, [K] is the global/total stiffness matrix and {d} is the unknown element nodal degrees of freedom.

When you use matrix theory the stiffness matrix [K] is singular as the final determinant result is zero. Very simplistically when a matrix = 0 in the solution phase you invert the matrix, which means there are an infinite number of solutions.

With each boundary condition that is applied this forces the model to remain in place – i.e. no rigid body motion. Which then means the matrix can be solved for an exact solution rather than an infinite number of them when the matrix is singular.

Rigid body motion means there is no exact solution to the matrix and the FEM literally flies off the screen as it is unrestrained in space – the man holding the rope that is pulled away – in the example analogy above. He is pulled away as nothing is holding him, thus the man is in rigid body motion.

Thus any 3D model can exhibit 6 degrees of freedom; three transnational and three rotational.
However, where any one of these degrees of freedom is left unrestrained it leads to an incomplete solution. When this occurs, rigid body motion occurs.

7. Joined: Oct 2008
Posts: 6,846
Likes: 801, Points: 113, Legacy Rep: 2488
Location: Japan

Yes.
How and where you apply restraints or boundary conditions is the skill and understanding of the FEM user.
There is no 'play book' for this, other than understanding the implications of what not to do, so you obtain meaningful results with confidence!
As noted way way back... you learn by first starting off with simple models of known exact solutions and apply simple sensitivity conditions to understand the implications of subtle changes to known and exact solutions that should not occur when you use theory of structural analysis when performing simple and exact known solutions to problems.

8. Joined: Aug 2002
Posts: 15,156
Likes: 914, Points: 123, Legacy Rep: 2031
Location: Milwaukee, WI

gonzoSenior Member

Firstly, take it easy with the insults and patronizing. Let's keep this focused on FEA. The pissing contest was started by you.

There have to be constraints for FEA to work. However, only some parts or areas have constraints. Bodies are not rigid and FEA is used to show deformations. In your example, the man's shoes can be constrained. When the rope pulls, the tension will create stresses in the body. The body will start bending towards the rope, until the man is laying on his face. Where is the disagreement in that the man's body should not be completely constrained, but only parts of it? For example, the sole of the shoes to the ground, and the hand to the rope.

9. Joined: Oct 2008
Posts: 6,846
Likes: 801, Points: 113, Legacy Rep: 2488
Location: Japan

And there we go again...
Please stop posting on topics/subjects you do not understand.

10. Joined: Sep 2001
Posts: 627
Likes: 120, Points: 53, Legacy Rep: 110
Location: Bordeaux France

an2reirfifty boat designs

Hello there I wish you anew a wonderful week from France ! Thank you for your postings and for the feedback!
In my opinion all and every of your feedbacks here are very useful to the proper understanding of th testing with FEA software
and I do give all your feedback thought and attention
As previouly mentioned I did about 200 FEA simulations with multiple software in the last 300 days
Previous to that I had the honour to serve with Universal Race Technology Composite a UK carbon motorsport company where I did hundreds of FEA
designing and testing the carbon chassis and high load parts of the carbon structure automotive .
I'd like to point to couple of aspects:
1. I think there is a new generation of FEA software like Scan and Solve for Rhino, Autodesk Fusion 360 but as well the Bureau Veritas FEA software
These new software were created in order to make mechanical simulation more affordable by companies and more practical to understand and use than the previous ; the most interesting being the "event simulation" in Autodesk Fusion
Some of these software work on the cloud and from my perspective this is very useful
2. One toolkit that is being used in EU for structural design in compliance with international regulations is the Bureau Veritas standards and the Bureau Veritas software Compose IT
3. Bureau VEritas does as well offer its own FEA software and does suggest the integration of the validation software Compose IT with the Bureau Veritas FEA software and as well BV does explain how to do that
4. I am doing myself work towards the integration of the Scan and Solve FEA testing with validation by the Compose IT software from Bureau Veritas
5. When doing FEA testing with Scan and Solve for Rhino it is very important to :
5. a pay attention the the accurate and correct input of the materials mechanical properties and especially when inputing fiberglass composites by density ; modulus of elasticity, poisson ratio, Yield Strenght, tensile strenght compressive strenght
5.b more specifically when simulating composite fiberglass there are two ways to set the fiber orientation either by one axe and the laminate thickness or by one curve and surface the laminate thickness being perpendicular to that surface
5.c The software computes the displacement of the parts on a uniform background grid of finite elements. If 2 or more parts cross an element, the displacements of the parts in that element will be connected. If they parts touch in an element they will be effectively treated as if bonded. If the parts cross an element with some space between, the parts will be treated as connected with a mixture of "air" and some fraction of the materials comprising the parts. This rarely makes sense physically (simulation isn't that smart), so you want to be aware of it to avoid these situations.
5.d Pairs of parts separated by a gap smaller than the element size will be treated as connected across the gap. The element size is shown in the report under Settings-Basis Functions.
5.e Deflection scaling takes the deflection predicted by the problem you've set up and multiplies those deflections by the deflection scale to improve their visibility. Normally deflections are so small (tiny fractions of mm for small parts) that they are not visible. Using the deflection slider allows you to understand how your shape is deflecting under load.
If deflections are visible without scaling or at low scale factors, then the linear elasticity assumptions (small deflections) built into the simulation are probably being violated and the simulation results will be of questionable accuracy.

Have a woderful day

Last edited: Feb 18, 2021
11. Joined: Feb 2021
Posts: 16
Likes: 2, Points: 3
Location: Nice

EddieGreenJunior Member

This is exactly where i am !

Let's take a couple of practical examples to try and conclude this thread... to see whether i am setting things up correctly ... or not ?!

EXAMPLE #1
===========
1/Let's assume i want to check the scantling and shape of some main web frames around the mast section of a metallic sailboat (easier with an isotropic material to start with) , together
with the scantlings and location of the chainplates. A rather classic enquiry i would think.

I want to carry a more thorough analysis than just isolating one single frame (where some basic rules of thumb could be used), as other frames linked to this main
frame through hull plating, longitudinal stringers, keelson, girders, etc, are certainly contributing to the geometry strength.

As a first step, let's take a truncated portion of the hull between 2 sections with say 5 frames : 2 on each side of the main frame supporting the chain plates. I am adding the rider bars (T structure)
to complete the frames, all the longitudinal stringers and girders, plus the hull plating between these 2 sections.

For this transverse analysis i am applying loads along the V1/D1 directions and the compression from the mast. Underneath there is a post running from the deck to the keelson in order to sustain the compression from the mast on deck.

There are certainly additional loads, e.g. could also add the hydrostatic pressure along the immersed hull surface if significant enough, but let's try to keep things as simple as possible, at least for the 1st run.

Now comes my favorite part with the restraints to be applied for the FEM analysis to converge. How about restraining the keel or at least the keel-hull interface for this specific study,
as it is a rather strong area and far apart from the chainplates ?

Would you put all the restraints there or would you also add some restraints (what kind i.e. along which axis ? using xyz : x longitudinal, y transverse, z: vertical)
e.g. at the ends of this truncated hull ? (not really advised if i remember some earlier contribution)

EXAMPLE #2
==========
2/ Assuming i have now completed the whole hull model with its primary structure (e.g. frames & stringers, stem bar, keelson & main girders, bulkheads) plus appendages.
I could test grounding (e.g. with a load x times the displacement), slamming (with e.g. a pressure proportional to the square of the velocity as a first step) effects on this hull structure...

but let's say i just want to start with the impact of the rigging loads on the structure (to the V1/D1 loads and mast compression, i would for instance add loads along the stay and backstay, say the rigging is without runners to simplify)

Now if i want to look at some longitudinal resistance (vs. my tranverse #1 exercise), namely the stress on the deck girders/plating etc where would you advise to locate the restraints ?
Unlike the grounding scenario i am not going to use the sheer line, i could use the keelson, appendages, or even a full length chine close to the water line, ...
What do you think ? If you can specify which constraint type, namely along xyz or a specific direction this would be even more helpful.

My intent is to perform these 2 analyses as learning exercises, using SnS and cross checking with Simscale, as soon as i have clear ideas re. the initial set-up.
Do not need "ready to go" recipes, but would welcome ideas (as precise as possible) or even some practical hints with hopefully some rationale behind to improve my understanding.

I appreciate your patience and as said earlier value all inputs. At a time restaurants/bars/pubs are closed in several of our areas/states/countries, forums are given us a touch of social life

12. Joined: Feb 2021
Posts: 16
Likes: 2, Points: 3
Location: Nice

EddieGreenJunior Member

Hello Andrei ! Lucky you, good wines in your region !

Re the Delft paper mentioned by Fastwave above and the explanations re. their better way of restraining a boat, do they really suggest to apply restraints along the rigging (as you know with SnS, it's either a load or a restraint)
or they do the constraints/restraining by applying the right forces/loads along the rigging ? let's take for instance for my better understanding what is applied at the V1/chainplate interface ?
Many thanks again for your help,
Ed

an2reir likes this.
13. Joined: Aug 2002
Posts: 15,156
Likes: 914, Points: 123, Legacy Rep: 2031
Location: Milwaukee, WI

gonzoSenior Member

You are a self claimed expert. How about you explain it to me?

14. Joined: Oct 2008
Posts: 6,846
Likes: 801, Points: 113, Legacy Rep: 2488
Location: Japan

Please state where I said I am a self proclaimed expert?

I have done there, HERE and HERE and HERE and HERE, but it appears you have either not bothered to read them or just have not understood them.

So, in the context of FEA - what does rigid body motion mean to you?

15. Joined: Oct 2008
Posts: 6,846
Likes: 801, Points: 113, Legacy Rep: 2488
Location: Japan