Simulation for ALL—Ease the Stress

SVEN ERIKSSON: Welcome to this morning session in the simulation track. And the title of my presentation is "Simulation For All, Ease The Stress." And my intention is to try to show on this session that our tools in simulation is pretty much easy to use. They are correct. They are nice. They are fast. They are safe. And I'm going to show both our points and I'm going to make quite a demonstration as well.

And my name is Sven Eriksson. I'm working at the company Symetri. We work all over the Nordics and in UK. We have a simulation team that covers all of those countries. And I'm leading that team.

My personal background is that I'm master of science in navel architecture. I've been doing simulations, primarily FEA simulations, since 1990 as a consultant since 1993 or so. So I've seen quite a lot of different stuff and quite many different programs. Welcome if you want to listen to simulation. And I've been also working quite a lot with Fluid Flow CFD analysis and thermal management. So it's kind of the whole mechanical area.

But today it will all be focused on the strength area and the tools Inventor Professional and Nastran in-CAD. And my agenda for today is, first, a little bit about simulation. What is it? Why? Well, the author is, of course, always-- not really, but and how do you do it. I will show a little about how other companies have used the software successfully. I will show a little demonstration of the Inventor Professional, or Inventor Stress, Inventor Simulation, or whatever you want to call it. I'm going to show Nastran in-CAD, some different functions in that. And hopefully you can catch up some workflows and see a little bit of tips and tricks. It's just one hour, so I can't make a complete and deep going session.

So if we start with this simulation for all stuff, many think I don't need simulation or my products are too simple or it might be too difficult. Just for predicting failure. And some find it expensive. And all of this is, of course, yes and no. Some myths. Something is as it was 20 years ago when many of us started. 30 years ago maybe. Simulation was difficult. I mean, it still is difficult. It's not just to click the green button and what you get is correct. You need to look at it a bit more. But I'll go through a bit more here. Today much is very much different, I'd say.

Let's start with this I don't need simulation. Many say that. And the products don't need a simulation. But even if a product is simple by simulation, you more easily understand how your problem works. Why is it deflecting as it is? How does the load paths look? By simulating the product, you can easily be more efficient in your design work. You can make the product better. You can choose better material. You can make a more environmentally friendly material. You can look at stuff that you can't really handle in a hand calculation.

And I'm going to show a short film of simulation. This is not what I call my demo. This is just a quick one about Fusion. Show what can be done in Fusion and for you to see how easy it can be to pick up a problem, simplify it a little, add loads, and watch the results. And Fusion is an easy and cheap product. Easy to use. And it was earlier our cheapest product. Not anymore. It's still cheap, but we have the other ones for free. So they are, of course, even a little bit cheaper. So you can evaluate very much. This is just a simple linear static stress simulation. But you can do so much more inside Fusion. But it will be nothing more about Fusion today. Not in my session. It will be a lot of Fusion out there.

Some people think of simulation just for predicting failure. And many use it for predicting failure. If you calculate an aircraft, you might just want to see that you have everything in place. If you calculate a building, it might be enough to make it not falling down. But to improve your design, to make it light or to make it more efficient, to make it handle all the loads, in this case it might be very important to have it light. Because when you deliver the product to the customer, if you can make it 20 kilos lighter, maybe it can be handled by one person instead of having two. So shipments will be cheaper.

So you can do quite a lot. And as I mentioned in the beginning, it can give understanding. I'm not sure how many of you-- we can back a little bit to that. How many of you have been working with simulations earlier? Most of you. Many of that has worked for a long, long time as well, I guess. I know some at least. Yeah.

Anyhow, when I started my career, or before at school, we used at the Royal Institute of Technology when we calculate our ships, we had some rubber models. It was ships built in rubber plates glued together. And you can take the ship, this size roughly, and you can bend it and twist and get the full feeling of what is happening. I mean, if you look at this piece of steel and machine [INAUDIBLE], for example, you can't see anything. However you load it, it will not deflect that.

So we can see it in the [INAUDIBLE]. You can't see the stresses. You can't see anything when it's loaded. If you have a huge structure or an aircraft wing, you see that it bends. But you can't understand the stress levels and exactly the load parts. With this former rubber model that we used 30 years ago and with FEA, you can see and understand everything. You see the reflection. You see if you scale everything. You see what's happening. You see the mechanisms. You understand what's going on. And that is not-- I mean, to understand it and to work with a model, that is far from rocket science. It's just to do it. You will see it a little bit in the demo if you haven't seen it before.

The expensive part. Of course, FEA software can be expensive. We have competitive systems that costs $2,000 a month or more. Inventor Professional or Inventor Stimulation, it's for free. You also have this shape generator included. We have Inventor Fusion, it has a cost. It's not a high cost. And we have Nastran in-CAD. Not for everyone yet, but everyone that swaps to the new collection has that for free as well. So the FEA part of the simulation tools, they are really cheap, but still good.

A little bit of the capabilities of the software. If we look at the inventor, we have linear stress analysis. And still, this is what many think that inventor do. Single part linear stress simulations. Anyone that still thinks that's what it can do in the Inventor? I hear it every week. Customer saying it's just for one part simulation. It can't be more wrong. Several years back, we can move a little bit forward, normal modes eigenfrequencies is also one. But if we take the third one, assemblies and contacts.

Several years back, I worked with a [INAUDIBLE]. It was a model that consists of hundreds of parts in the FEA model. I had like 50 contact pairs. It was millions of degrees of freedom several years back. I solved it, and I knew that when I had the CAD model ready, I knew that two hours from now, I will have a result. It wasn't any worries. I knew that it was just to start the simulation, and I got the results. So it handles big problems. It's fast. It's accurate. But it's linear. So we can't do any nonlinear stuff on that one. And then the shape generator, as you saw.

If you want to go more advanced, we have Nastran in-CAD. And we have the first three we have in Nastran in the same way. You can do it in a little bit more detail. But it's pretty much the same. But we also have a lot more linear stuff. We have dynamics. We have pre-stress bolts, for example. We have composites, if anyone wants to do that. We have linear buckling and we have thermal stress. And then we have most of the advanced mode dynamics that you probably want.

Everything that is commonly used in a vibration test, for example. And we have most of the nonlinear stuff that you use. You have a rubber material. You have plastic curves. You have large displacements. I'd say you have most. You have drop tests. You have nonlinear buckling. So almost anything I think you need. There are stuff you shouldn't do. You shouldn't model a car and run it into the wall. That kind of simulation is a little bit too complex for this software if you want to make it in a very efficient way. You can make it, but it will probably take a little bit more time than you want.

I'm going to show a few examples that are the typical cases when people start using Nastran instead of Inventor. And then the first one is to have more options, to have some more and better meshing options to the model line elements. For example, pre-stressed beam connections, very fast, easy, and accurate in Nastran.

Buckling. And this is the case I'm going to demo. If you have a thin wall structure, buckling is typically a problem. You need to handle the stability problem. Thermal stresses. I have a customer example about that one. Fatigue. If you want to calculate fatigue within the FEA software, you can make it inside Nastran in-CAD. If you want to make a drop test, just take something, drop it to the ground, make it pretty much automatically. It works fine. And then the nonlinear and the many use of that. When you pass the yield limit, the rubber material, large displacements.

And now to some customer examples. I'm trying to cover most of the areas. We have in frame design an escalator made by customer. This one was built up with a frame generator. The main structure of the escalator. They moved from frame design to frame analysis. And automatically, the model turns from just the line or the beams to a beam FEA model where you have beam segments, you have a rigid links. All the black marks here is a rigid link that connects the beam ends.

As the beams-- they doesn't physically meet. They end away from each other. And then load is supplied and a simulation is made. Well, it's like an ordinary FEA analysis. You get stresses. You get deflections. In this one, you also get beam diagrams. You get moment diagrams. You get share diagrams. Any kind of beam output that you might want to see. And also, of course, reaction forces.

And this one, the top part was made separately. And if you want to see that a little bit more, you can join our session at 1 o'clock this afternoon. It will be a short demo on that one in the frame analysis. That is mainly about frame design, that session. So it will be five minutes, 10 minutes of simulation. And in this case, we could check deflections as well. Deflection stresses. Pretty easy to do.

This is a trailer. It's a welded trailer. The customer wanted to check for fatigue to see if it would last the number of cycles they wanted or the service life of at least 10 years, hopefully 20 years. And this was completely done inside Inventor. So I wouldn't say it's a big model, but it's not a very small model. It's a medium size model. Maybe it's 50 parts in this one and maybe it's five contact pats, five to 10 contact parts.

And it's, as I said before, it's fast. It's robust. You get reactions. You get stresses. You have the possibility to make refinements to extract all the stresses you need to make your proper fatigue analysis. You get all the stress components. We'll come to that in the demo. And of course, deflections.

The case before was in Inventor, as I said. And this is [INAUDIBLE] in Nastran in-CAD. And the reason this case is analyzed in Nastran in-CAD is that we wanted to use beam elements. We wanted to have bolted connections. We wanted to have rods that it's hanged in. So this is what's the model. The corners were modelled a little bit more explicit. And they had some springs attached to the corners to control it [INAUDIBLE]. And it's pretty easy.

As I said before, getting the understanding on what's going on in the problem. If you look at the picture to the right, you see the twisting and the size of the twisting compared to every other deflection. And that makes it a lot easier to understand what's really going on. If you look from the top, you can see that the top gets quite skewed as well. So use the results. Exaggerate the formations.

Look at the model in many different ways and you realize how much more understanding you get from FEA than just picking up the stress level. I think it's a misuse of FEA software to just pick out the deflection value or a stress value or tell stress values. Use it for understanding. Use it to find load parts. See where the stresses are really low, et cetera. And this is the corresponding stress plots.

One more customer case. This is a plasticity analysis. It's a high internal pressure, in this case, and the customer wants to see how big the remaining deflection will be. And as it's a thin plate model or a thin plate design, it is analyzed using shell elements. That can be done easily in both the Inventor and in Nastran in-CAD. It's easy to make a solid model when you're do in the CAD and then swapping to mid plane or just creating surfaces on the model.

And as it's high load, you want to see the plastic deformation. Of course, nonlinear material curve is needed. The simulation is run. You get a deflection under load, and you get deflection after the load is removed. So this is the remaining deflection in this case. And that is a simulation. Of course, it takes a little bit more time to run this. In this case, I'm not sure if it's 10 or more iterations that is needed to get the final results. But it is pretty straightforward to make that kind of simulation.

Vibrations. Not that everything is exposed to vibrations, but very often construction needs to handle vibrations mainly due to transportation. If you have a big machine that should be transported from place A to place B, it might go by car, by truck, by ship. It will be exited and for that reason, it needs to handle those cases. And many customers of us, they design their stuff, they take it to a testing facility, and they make a shaking test. And if it works, of course, it's fine. But they still don't know the margin. It might be 1% in margin.

If it doesn't work, then you have a broken piece. You might have only tested one case and you were supposed to test six directions. And you go home. You don't have a clue what has happened. So if you make the vibration tests inside, in this case, Nastran in-CAD, you get the proper results directly. In this case, there are some eigenfrequencies that will shake more. And you see both deflections and stresses. The customer could make some modifications and move everything to a safe ground.

[INAUDIBLE] it sounds like it's too much noise. Is it OK with you or is it just for me it sounds like that? OK, thanks.

Thermal stress. In this case, it's not the same brake disk as in the first example shown. But if you have a brake disk, it will be warm. That's the complete meaning of it. They take the kinetic energy and makes heat of it. And to simulate the stresses inside a brake disk, when you break, you need to make a transient simulation to calculate the stresses. You put all the heat under the time it takes to stop.

And that transient simulation is done inside Nastran, in this case. You get the temperature distribution that you can watch and look if it's realistic, if it's too warm, and then you might need to redesign it, make it a little bit heavier to handle the energy. And then you take it to the next step. Lift the temperature results into the stress solver. Put it as a load, the temperature load, and then you make a FEA simulation. You get the displacement and you get the stresses. And from that, you can evaluate if it will last or not.

So that was most of my PowerPoint. And my intention now is to swap to Inventor instead. I have some words to say first. And in this case, I'm going to check some typical examples what you want to look into. Stresses at service load is a typical example. You might want to see stress at ultimate load as well. Buckling in this kind of problem to see that it will not collapse. It might happen way before you reach the yield limit.

Fatigue. I mean, this machine, this crane will be lifting logs day in and day out. It will be hundreds of thousands lifts. Maybe not hundreds, but let's say 100,000 lifts. And they will make sure that it doesn't break in the welds during that time. Can that be done in Inventor, all of this? No. We can't make any buckling in Inventor. We can't make any plastic deformation in Inventor. We can't make a nonlinear static stress analysis. If we look at what happens at the ultimate load end and it might start to collapse but maybe it will not collapse fully.

Fatigue. I'd say we can make it in Inventor. I'd prefer to make it in Inventor or the static part of Nastran in-CAD. But it can be done with a fatigue module as well. But it's the handbook way I'm going to show in this case. And this is the problem I'm going to work with.

So now it's inside Inventor. I'm running on a laptop. Pretty standard. No, not pretty standard, but it's a laptop. So what do we do then? We have the model. Everything is now-- the simulation part is inside environments. So here we have dynamic simulation. I'm not going to show at all. It's a really powerful tool. We have stress analysis I'm going to show today. We have frame analysis. I'm going to show it in the afternoon. And we have Nastran in-CAD. I'm going to show a bit today.

So what we do now that we just enter stress analysis and we need to create a study? Here we have lots of choices. We can make some optimizations, some convergence tests. We can choose between a static analysis that is the default. And we have the possibility to make a modal analysis. I'm just going to run for the static analysis, in this case. I keep the rest as default. And we get the free coming up here. We have everything is empty for the moment, except from the parts.

So what we do. As I'd say in most FEA systems that I work with and the standard methodology, work from left to right. And then when it comes to the run or simulate button, you are properly finished. So we follow that path this time as well. In this case, it's a linear analysis. We have [INAUDIBLE] material already inside Inventor. So we skip that. We are not going to look. I'm going to make it the easiest way now in Inventor. So we skip the thin bodice and mid surface and that part. I'm going to work with solids.

So we come to the constraints. How to attach it to the environment. And to be honest, this is the part that you should pay attention to when you work with your own problem. I'd say that at least 50% of the failure is done in boundary conditions. Often it goes higher. Often you can make any kind of failure on that one. So beware of that. Try to attach it at positions where you are not interested in the results. Attach it quite far away from interesting areas. In this case, it's primarily the outer part here that we are worried about. So I'm going to attach it here way in the back.

So what I do is I click on Fix and I choose my surfaces that I want to fix. So I take all these three surfaces. And I can choose components as well in this case. So I can hold it in just one direction. I can put similar conditions instead, if I want. But in this case, this is good. So I click OK on that. And you see I get the [INAUDIBLE] as well.

Then I'm going to add some forces. I put the load to this surface. I use vector components. And this I can type in a little bit like I want. I can type minus 20 kilonewtons if I want to. And I'm going to have it forced downwards as well. So I type minus 4,000. That makes it neutral right away.

And then I have, if we look at it, I have a somewhat downward pointing load. And then I have some more loads. I have up in this-- on these bearings, I have the load from the hydraulics. It's 20,000 newtons in that way and 2,000 upwards. Apply. And then we have about the lower two where we put a load as well. And I think I used 15,000 upwards and 30,000. That was what I had.

OK, so this is my boundary conditions and loads. And, of course, we need some gravity. And if you have a line pointing in the right direction, you can easily just choose the standard and use some line. But in this case, it's not going that way. So instead, I say it's minus [INAUDIBLE]. And I get the marker on that one as well. So we can see that we have it. In this case, we don't need to go to contact, because it's automatically done, as long as we stick the bonded ones.

We can right away click on Mesh. Depending on how we have done your mesh settings. We have some general mesh settings here that can be modified. You have under stress analysis settings here. You can choose your mesh settings. And I'd say you probably want to change them a little from the default settings.

So that is done already. And if I mesh this now, I get some warnings, because the Inventor stress thinks that this is a thin plate bottom. And I agree. It's definitely a thin bottom. But I still want to run it as a solid. And the reason for that is that in the end, I want to check it for fatigue and I want to check it in depth into the world. So I want to stick to a solid. And depending on how you want to look at it, it can work fine.

So I have done everything [INAUDIBLE] here and I'm about to click to simulate. So we go to that one and click Run. And I get the warning again. It's a thin bottom. And still I get results. And as you see, it doesn't take a long time. It doesn't have to take a long time. It is a pretty small problem. But five seconds, you don't get a coffee on that one. And here we see we have just below 200 megapascals.

So if we have a standard construction steel, it would probably be good enough. It would be a factor too in safety for the design load. And that would probably be good enough. Maybe not for a lifting device. But as mentioned before, we are going to look at more. Yeah, can you turn down the volume a little?

So we continue on this one. And check the same thing in Nastran in-CAD. So I finished this one, and we go to Environments, and we move away to Autodesk Nastran in-CAD here. And we have a new tree. Pretty empty tree, because we don't have anything more than the parts. But as we already have done the model, we can just click here Import Stress Analysis. And we populate it with all the loads from Inventor stress. We have the fixed constraint. Well, that's what we get.

So what I want to start with now is I want to add new material. So I select material from the Autodesk library. I use the standard constructions build 355. And I click on that one. And I can delete the old one. So now I have one material.

And what I'm going to do now is I'm going to make a thin plate body out of this one. And I make it by doing offset surface function. So I click on that one and I choose the plates I want to use for this mid plane part. So it will be the six sides of the arm itself. I give the offset or I could give the thickness. In this case, the arm is 1.2 millimeters or intended to be 1.2 millimeters. So I have an offset of 0.6. Click OK.

I got a warning. What did I do wrong? [INAUDIBLE] made something wrong here. OK. I click on that one, that one, that one. The typical demo problem. Set surfaces. My dialog box is gone. I make a restart on this one. I'm not sure [INAUDIBLE] graphics card combined with the monitor or something. Let's try again. Go to Autodesk Nastran in-CAD. I have my model. Take this one.

That's more how it should look. OK, so now you see I have got-- the transparent surfaces do have a mid plane defined on them. They don't automatically have any elements yet. But it's defined to be mid plane. I continue with the other ones. They are not really thin, but I want to run a thin plate on everyone. So I take all the four surfaces. And it's two reasons. One is I want to show it and the other one is that if you connect the shell element to the solid, it will be a hinge that you may not like to have. Or hopefully it will not [INAUDIBLE]. So you need some kind of a shell on the surface as well. We take that last one as well.

So now we have our model. Everything is defined. We can look at each surface here. We have our defined material on it. We have done the idealization by doing the offset surfaces. And we continue with the constraints and loads. Because we have defined constraints and load before, but that was on the solid surfaces. And as you see on this model, we don't have any surfaces left where we had the loads. So we need to go through all the constraints again.

And we start with the fixed constraints. And we just need to pick the lines that we-- sorry, lines. We need to pick the lines that we are going to use for the boundary conditions. So I pick the end lines here. It does not understand if I put the boundary conditional load on a surface and that was modified to a line, it disappears. So now I get that one. And I continue with the forces.

This is this outer one. Put that one there. You see a small marker. And we continue with a force number two. [INAUDIBLE] this one. [INAUDIBLE] that one. And what you need to remember now, in Nastran, you need to mark total force. If you don't, you'll get the full load on each. So if we mark that one, it works fine. And then we take the last one. And I forgot this almost every second time. So now we have all our loads and boundary conditions set to this new model as well.

And as we now have plates instead of solids, we need to define contacts. And we do that by clicking Solver to use the automatic function. And as we have an offset here, we choose offset bonded as contact and we use the maximum activation distance. I put it a bit more than the [INAUDIBLE]. And so I use eight millimeters on this one. It could be less, but this works. We continue and choose the mesh settings, the model checks, and suggests.

And as this beam is 80 millimeters thick, I think having one element on that distance is a bit too little. So I make it a little bit coarse. But I define it as 20 millimeters. I have parabolic. I have continuous. And I click OK. And I get the mesh. A nice looking mesh. A little bit too coarse to have it as a final mesh in the job. But I think it's good enough for this demo. So we have the mesh. We have the loads. We have the material. Everything should be ready. And we click Run.

And this is also pretty fast. It's a bit faster as we have shell elements. We have a little bit higher stress. It was 180 before. It's 8%, 9% higher. But I assume it's because we had solid elements on that one. So this is the linear stress analysis inside Nastran. What we should do now is that we need to swap the analysis type. So I go to this one and choose Edit. I modify this from linear static to linear buckling.

I also go to Options and I choose the number of modes I want to check. It can be done in many ways. You can choose many modes, and you can choose within which bounds you want to see the backing modes. In this case, I say I want to see four just as a figure. And I click OK. And I can choose which representation level I want to be. And if I have done a special FEA level, I can use that one. So this is what's done. We have four levels, four modes that we are going to look at. And we click Run again.

And it is, I'd say, it's fast. It's not a problem in waiting time in these kinds of problem. We have a solution time of roughly 10 seconds on this one. And we get the shape. And what you should think of when you run a linear dynamic-- sorry, a linear buckling analysis is that the values that you get from the buckling simulation, they are not a correct value. You only calculate the modes. In this case, you have eigenvalue number one at 0.82. That means that at 82% of the nominal load, this thing will buckle.

So if you're loaded with a full load, it will start buckling. Not as you intended, maybe. But even though this stress is pretty close to what we calculated in the linear stress, this does only say if we look at displacement, you see we get displacement of one. We have a level displacement of a one. And then we get the stress corresponding to that one millimeter.

So the stress we get, it has nothing to do with reality. You can compare it within the model. So you can say if I light the probe and if I look here, if I have 1.8 here and [INAUDIBLE]. You can't use the values. Only look at the modes, the mode number, in which levels. So what I'm saying here is it will buckle at 0.82. And that is a problem. You can't make a lifting device that buckles below the limit load or the intended load.

So what we do here is that we go right into the model and offset the surface number one. That was our arm. So what we do is we go into that one and test it with another dimension. So we change 1.2 to two millimeters instead, just as a rough guess. It should be the thickness up to three, the exponent of three that you increase the buckling stability of. So this would make it if someone is quick and had to divide by 1.2 up to three. Let's see who's the fastest. Five seconds.

So it's easy to work with modifications here, to check what happens if we move now from 1.2 to two. And what I get from this one is that we move from 0.82 to 3.75. So in a linear backward analysis, we have a safety factor of 3.75. And that is about minimum what you should have in this kind of simulation or design, because it is probably welded. You will have pre-stresses from welding that might give or probably will give some compression stresses that will initiate buckling. And it will probably have some predefined buckling. So you need to have a bit of a margin. But we have been able to show now that you easily can make a linear buckling analysis. And that was my intention.

I'm not sure if I have time to-- I'll make a quick one on the nonlinear static as well. I think I will have the time. I swap back to 1.2 millimeters and show quickly how you can make a nonlinear static analysis as well. So we just go here and we go down to nonlinear static. And we choose OK on that one. We go to the nonlinear setup. Edit that one. And we need to define how many increments, in how many steps will we divide the loading path. And we say 10 and we want to see the intermediate results. So we see each step and we click Run again.

So what we do now is we step the load and make a nonlinear analysis. And we will see results from each time step now. So after 20%, we have this level. And it keeps increasing now. So we can follow the analysis. We can stop it if we see that it buckles too early. We want to stop it. So this is a way to combine a nonlinear static analysis with regard to deflection and at the same time see buckling.

And if we also swap the material to nonlinear material data, then we can make a plastic analysis at the same time as well. And as we are close to the full load now, we see it starts having a little bit harder to converge. It has done 90% so far. But it is reducing the time steps and it's [INAUDIBLE] a bit. So now it's complete. And this is what we now get from the nonlinear static analysis.

And now we can look at the stresses. So what we see now is that at the full load, at buckling, we have a maximum stress-- maximum [INAUDIBLE] stress of the shell elements. So just below 300 megapascals. And it's not even above the yield limit. So this is a valid result. So what we see is that it didn't collapse at 0.82. It started buckling, but it didn't collapse. Still you need to make it stronger.

This was the last piece of Nastran. I think we can make it also. We go back to the stress analysis. And what I'm going to do is I'm going to look at the problem here, but adding-- let's see that I-- yes, I have everything left. I want to refine in this area where I expect the highest risk of buckling to occur.

So what I'm going to do is the standard procedure, I'd say, for this kind of problem. I make a little bit of a refinement area. So I create the CAD sketch. I project this line. I add a little line in this direction. It's not important how it looks. I end it. I make a little bit of [INAUDIBLE]. So I'll make this one. Let's make it four millimeters. It's not very exact.

The reason I choose four is that the weld route will probably be about that. Probably will have a weld size of three millimeters of this one, and that makes the weld route come about 4 millimeters out. So I want to be able to look at the stresses at this position. So I make it like that. I go to Split. I split. I use this line split to split that surface. And I'll split the surface behind. So I split the two surface, front and back over the plate. And I apply.

Maybe I was too fast. I probably failed to make-- I think I failed to make these ends meet. I guess I made like that. Like that instead. So now we see that surface split. We go back. I'm sorry for the delay here.

So what I'm going to do now is I'm going to make a mesh refinement by this bottom. So I choose that surface and that surface. Those two banana shaped surfaces. I'd say that within those, I want to have a mesh size of two millimeters. That makes us get two elements over from the weld to the weld route. And that's what you always need. You need to have at least two elements of the weld to get correct results.

Update Mesh. It takes a little bit longer, as we'll get-- it's a little bit harder for it to mesh. So this is how it will look instead. I should add that one as well. So that piece should be added as well to get a little bit nicer mesh. So now we can start simulating this problem. And still it complains that I have thin elements. But in this specific area, I will have a pretty nice mesh anyhow.

So, while this is calculating. Oh sorry, it was too fast. So what I have now is I have a little bit more correct stress calculated here. I have a little bit too coarse mesh here. But we can go into the mesh and look-- sorry, into the result and look a little bit more deeply into it. And if we light the probe and look here two elements out, this is about where we have the weld route. And we have about 200 megapascals in that area.

And if we assume-- that is the [INAUDIBLE] stress. I forgot to mention about in Inventor, as in Nastran in-CAD, you have almost all the outputs that you might want to have. You have these five as the standard, easy to get alternatives. But you have all the stress components. You have all the displacements components. You have all the strain components. And you have all the contact pressure components. So you have pretty much all the results output that you usually want to have.

And in this case, I want to see the third principle stress, as we have a compressive load here. I want to look at the third principle stress. And if we light up the probe again and [INAUDIBLE] here. If you see, look at the bottom, I didn't say it, but in bottom [INAUDIBLE] when probing, you have the level listed. So we have 205 megapascals, roughly, in this area.

So what we do is that we note that stress. And in this case, we make it easy and we assume the opposite load is 0. So if we load, we get 205. And if we unload, we get 0. So that's how far I use the FEA in this kind of stress and fatigue analysis. So I'll leave the demo. Not the demo. But I'll leave the FEA part of the demo and continue hearing my PowerPoint instead. And I go to that slide. This is how I would like to handle the fatigue [INAUDIBLE]. Oops. OK, [INAUDIBLE].

The different lines are [INAUDIBLE]. Which kind of weld quality you have in your design. And we have typically what [INAUDIBLE] 100 line in these kinds of welds. A continuous, nice weld. And you follow that line. You go to the 200 megapascals, as we have [INAUDIBLE] range. Those lines meet. And you go down and see how many cycles you get. If we have the continuous standard cycle [INAUDIBLE]. Saying we have roughly 250,000 load cycles for this one on this maximum load. And then in this pretty easy way, you can make a fatigue analysis based on the results in Inventor without making anything incorrect.

I turned down [INAUDIBLE] so I see what you see. OK. So if we go back to my PowerPoint series and summarize what we did. We looked at linear-- no, you don't see what I see. I don't know what's happened. We see completely different things. Try it that way instead. Now at least you see the same as I do.

So what we did was we looked at the-- we had the problem. We made a linear static analysis inside Inventor. We got stressed results. Not as accurate as you might want to have it. But by refining later, we could see what happened. And we could make a nonlinear static analysis. I can turn fullscreen instead. We could make a nonlinear static analysis, and by that make a check of the output load. If you have a collapse scenario or something, we can make that one inside in Nastran in-CAD or inside Inventory if it's below the yield.

We did a buckling check. We increased the thickness and could verify that we have eliminated the buckling problem by increasing the thickness. And I didn't show you, but you also get the stress results within a buckling analysis. You can look at that as well. And we could make a nonlinear static analysis check what happens. Will it collapse? And in this case, we've got some kind of load redistribution. We made the fatigue analysis.

And I'm almost ending up now and trying to say that I don't think those arguments are valid. I think that most of them are because it has been a cost. It is inherited that it is difficult. It was difficult. It was a high cost. You needed to be extremely specialized. But I wouldn't say it's so anymore. I think it's more of a fact that most designers benefit from it. Simulation is possible for everyone. Most products benefit from it. It's definitely not expensive. If you don't have the personal site, it might be expensive to hire someone. And it is for every design stage, not just verifying the final one.

So now I'm ready for questions. Unfortunately, I have just three minutes. Please go ahead if anyone has something. Yes?

AUDIENCE: Are you able to change the temperature of your design in Nastran so that [INAUDIBLE]?

SVEN ERIKSSON: You mean if you get lower Young's modulus or lower yield strength or something?

AUDIENCE: Right.

SVEN ERIKSSON: Yeah. You can make different material models for different temperatures. And you can, of course, run with the temperatures as the temperature as boundary condition. Yes?

AUDIENCE: Can you push back your changes [INAUDIBLE]?

SVEN ERIKSSON: No, not the shell modification. That is inside Inventor. Sorry, that is inside Nastran. But it makes it also easier not to swap all the way. And when you make the changes pretty fast to make it and then just go into the Inventor. As you are inside the Inventor when you work, you can easily go into that. But no, you can't make it that way.

Everyone ready to try Nastran? Everyone ready to try to work a little bit harder with different cases? You're not really convincing. But call Eric if you need help from Autodesk. Call me. Call Eric's colleagues. There are many guys in Autodesk support that can help you. We are many companies like mine, persons like me that feel this is fun and wants to work with it. And we give a lot of trainings.