I have been playing in sumulation a bit more and i am looking into how stuff reacts when it is dropped and hits another item, or i have been looking into using counter weights balance on items, but i get some odd results somtimes like parts going though each other or on contact suddenly flying off etc..
I notice that when i play with the settings in 3d contacts I notice this has a grate impact on these effects.
is there a set of standard settings or a list of settings for different materials that can be entered in the 3d contact properties boxes:
Also it can take loads of time to set up stuff like this, does anyone know if futre developments in simulation will make this quicker / easier?
I would apreciate any input on this.
P.S Hi Jd I already own the simulation 2011 book.
As far as I know, there is not published information about joint properties guidance for Dynamic Simulation.
It is not an easy task to answer. Take for example the classic spring, mass, damper case. The critical damping is equal to two times the square root of the stiffness multiplied by the mass.
We don't know if the user wants a underdamped, overdamped or critically damped system. Also, we have to take into account the mass (easy) and the stiffness. If the stiffness is not defined, there are now two unknowns to a three variable equation.
Friction coefficent can also vary greatly depending on the materials in contact, surface roughness, lubrication, etc.
Please let us know if you have any additional questions, comments or suggestions.
I've logged wishlist item 1455553 for a Joint Properties Guidance. I know it's not much help, but it's the best I could do for now.
Best regards, -Hugh
[Edit: changed 'divided' to 'multiplied' I shouldn't go by memory ]
thank you for your reply, I understand your comments and if I were using a spring, bush or a bearing etc.Then this information will be very specific and it will also be available from the suppliers of the items. It is not in this area I need the information for.
Basically I want to use the simulation environment to test out simple ideas, theories and what if scenarios. Have you seen working model 2d? This is excellent and simple for this type of stuff but it only works in 2d. the main reason for me upgrading to simulation was to do real world experiments in 3d using physics simulations.
in playing around, I sketched up a square block 1 mtr square and a long square bare 200mm square by 2 mtrs long and i set the gravity on to the block which is grounded, I also set up a large flat area around 10mtrs square to represent the ground. I set the 3d contacts and spatial joints between them all, I then tried a couple of experiments:
A) I placed the long bar ontop of the grounded block and then dropped another block from a few mtrs up onto the end of the long bar.
B) I placed the long bar ontop of the grounded block off set to one side and then placed another block on the end of the bar to see how it behaves with a cantilever effect, the intension being to position them to find tipping points etc..
Initially I had some very weird results like the rectangle bar shaking and then flying off into space, or parts meshing into each other, or the block dropping from above bouncing off the grounded block but the it went through the floor plate even though all the contacts and joint were set, (if i suppressed the grounded block then the other block didn't go through the floor??)
by playing with the main settings in 3d contacts I managed to change the results and got what looked like reasonable results, however I need as accurate results as possible, that i can make decisions with.
now I am sure that if I set up the materials for these items there should be settings that are standard for each given material?
As I said I purchased the inventor simulation upgrade because I need a 3d physics environment to play in for some of our designs and concepts and I have been told by our resellers that inventor can do everything I need.
but it seems so long winded to set stuff up like this and the results are questionable.
I have had demos in solid works and from i can see it is really quick and simple to set stuff up like this. Literally by group selection and clicking a couple of buttons and your away.
can you advise on any information on the 3d contact settings for given materials and is there any good developments in the simulation side on the horizon for inventor?
Hi Hugh or anyone else.
is there no information on these settings? also what is on the horizon with development in simulation?
My apologies. I did make a post yesterday, but had to delete it. When I re-read it, I wasn’t satisfied with what I was trying to get across. Let’s see if I can answer your questions.
<<thank you for your reply, I understand your comments and if I were using a spring, bush or a bearing etc.Then this information will be very specific and it will also be available from the suppliers of the items. It is not in this area I need the information for.>>
This is truly how DS joints work. The damping in a DS is not a material property, and the stiffness is not the material stiffness (Young's modulus). Stiffness in DS is a spring stiffness within the joint. Material damping varies greatly on the stresses, assembly techniques, etc....it is not really a define-able and useful material property. Material damping is used in dynamic response of body deformations, excitations, etc.
In DS all bodies are considered rigid (this is one reason why DS doesn’t take material stiffness or damping into account). Take for example a rubber ball hitting a wall. In real-life the ball hitting the wall would deform a little bit and spring back. Since all bodies in DS are considered rigid, the way we can simulate this effect is we allow penetration. We control the penetration by stiffness and damping values, which is literally like attaching a spring and damper between the two bodies. For 3D contact, the joint properties take effect only while the bodies are in contact.
The 3D contact in DS is what we call an imprecise contact. This is opposite to the 2D contact, which is precise. 2D doesn’t allow penetration, so it doesn’t include stiffness or damping. It uses the coefficient of restitution and friction coefficient. Imprecise means that the geometry is faceted, and not necessarily the true shape. The 3D contact was designed to be used for impacts, and it doesn't like static non-movement. If I rest a block on a table with a 3D contact, it will start moving around. I asked Development about this years ago, it is a solver limitation that has something to do with too much over-constraints with multiple contacts points and some penetration occurring, or something like that. I’ve logged it again as 1456022 in hopes Development will re-focus on this issue. I cannot explain the passing through behavior unless a 3D contact was not active. For this simulation, it would be better suited using 2D contacts. 2D contacts are preferred over 3D contacts whenever possible due more than one reason.
<<by playing with the main settings in 3d contacts I managed to change the results and got what looked like reasonable results, however I need as accurate results as possible, that i can make decisions with.
now I am sure that if I set up the materials for these items there should be settings that are standard for each given material?>>
DS doesn’t take material properties into account. It uses the inertial properties calculated by Inventor, and the parallel-axis theorem to include the body’s inertia in the joint. When a standard joint is created, there are two complementary coordinate systems, the parent and child. All the F=ma calculations are done via the joint relationships and body geometry is really just 'going along for the ride' with joint coordinate system for display effect. When you start adding motion, force, or contact joints, then the actual geometry can be taken into account. Damping and stiffness of the materials are not taken into account. Even if we could take them into account, with different materials coming into contact, it is not a simple thing to combine stiffness and damping characteristics of the two materials into one or more properties that the joint can use. This is what Autodesk mechanical (formerly Algor) does for their MES (mechanical even simulation) drop tests, etc. They are basically combining dynamics and FEA in one solver.
<<can you advise on any information on the 3d contact settings for given materials and is there any good developments in the simulation side on the horizon for inventor?>>
There are published documents about friction coefficients for certain materials and when combining dissimilar materials. But Inventor doesn’t have all these materials, and will not be able to determine this empirical data for user-defined materials. E.g.http://www.engineeringtoolbox.com/friction-coeffic
Damping is not a reliable material property. For an interesting read about material damping, see http://www.eng-tips.com/viewthread.cfm?qid=49887. The damping DS uses is viscous damping: http://en.wikipedia.org/wiki/Damping
As always, Autodesk employees are not at liberty to comment whether or not new functionality or issues will be addressed in upcoming releases, service packs, etc. until after they are released. What I can say is that Autodesk is continuing to invest heavily in the Simulation space.
To answer your question, how to find reasonable Stiffness, Damping and Friction for 3D contacts, I have done the following:
1) Lookup the correct friction coefficient via published documentation (if possible)
2) If you get too much penetration, tweak the stiffness value =10x. Keep tweaking until you can simulate reasonable stiffness.
3) Now you have 2 of three variables known in the critical damping equation, so you can solve for critical damping value and use it
4) If you desire an under-damped or over-damped system, tweak the critical damping value accordingly
I’ll try to answer your other questions in the bowling alley / skittles thread about components being in the grounded group initially (opposite to the assembly modeling paradigm)
Best regards, -Hugh
thankyou for both of your post replies, I fully understand your replies and comments and it has opened my eyes on the way DS actually works.
I now feel quite disappointed in DS in it's current state.
when I was moving from Rhino 3d to a parametric package, I took a long time to decide between solidworks and inventor and I chose inventor and to be honest i do really like it in general (although after a few years of using it and learning inventors ways it does have issues that make stuff seem so long winded to set up and preform.)
about a year and half a ago I decided to upgrade to the full DS package, again before i did this i looked at many options and was particularly interested in a package called "working model 2d" which was a couple of thousand pounds to purchase and I nearly did, however it was limited to 2d and it seemed like it had not been developed for around 8 years or so. I re-looked at solidworks and this would of been a major spend but I was convinced by my reseller that inventor will do all that i wanted and more.
Now don't get me wrong, inventor does do a lot of stuff and it is really good in many areas but sometimes it can do it in a round about way that is basically long winded and messing about.
"inventor" is not currently living up to it's name! perhaps it should be called "Engineer" because the way I see it after a few years of using inventor, it is very good at engineering stuff if you go about it in a real world engineering approach i.e:
when modelling a part that is to be machined you start off with a block and then you add or remove bits to create the finished model as you would in real life.
When creating a simulation, It is very good at mechanisms or stuff that is engineered from known principles, as you say with 2d contacts it is known as accurate simulation, so when designing straight forward designs like a shaft running through a bearing or gears meshing it is very good and I have partially read the dynamic simulation book and seen online tutorials and DS is good for a range of stuff.
I think "Inventor" is not so good for inventing! when you don't know the data up front it becomes a whole lot more difficult to create designs. I want to and indeed i do create moving mechanisms, moving designs and functional solutions however I need to do as much as possible in the 3d cad environment to work out the solutions and it needs to be done as quickly and accurately as possible to be profitable as a business when designing solutions.
I am not a mathematician, scientist or a PHD but I am good at solving problems and I use software to assist in my problem solving abilities.
So when I looked at Inventor dynamic simulation I had hoped it would perform for me when inventing stuff of an unknown outcome. for an instance if I were to need to design a device to sort out balls that were hopper feed into the top of a machine and then as they went through the machine there were sorted by size into slots that then conveyed them along to drop into an individual box for the size etc..
with this sort of simulation I know you could set this up in inventor but it would be an absolute nightmare to setup in time and effort and then I would wonder if it would give correct results and then when you run the simulation I would gaurentee it would take forever to run.
in a simulation like this there would be gravity, 3d contacts, friction etc.. the balls at the top would be poured into the hopper they could jam up so the design would need changing to make sure they were free to flow, the balls would need to fit just right through the slots based on their size so if the ball was too big it would carry on until it finds a slot it could fit in, it may be necessary to have an ejection system to remove unwanted balls etc.. etc..
it would be necessary to run the simulation get a result, change the design, run the simulation, change the design etc.... I don't think Inventor dynamic simulation would cope.
I would like to think this area of Inventor will be improved dramatically soon But i don't think so, even in 2d sketching with sketchblocks which were introduced a few years ago I asked for a contact constraint between non conecting parts so they would move upon mementy contact, as this would make it really useful, but it never gets added?
I am just looking into a couple of add ons to Rhino 3d "grasshopper" and "Kangaroo" ok the naming could be better but the products look very interesting and they are being developed by very small business possibly by one person that is programming and yet they are touching on very new techniques in 3d design.
When you look at the power and size of Autodesk I would expect the software to change in leaps and bounds each year.
I realize that Autodesk staff must keep the lid on future developments in the software but equally I would love to know that it is worth hanging on with inventor and it's yearly subscriptions because these areas will be developed.
Again I would say: "inventor" is not currently living up to it's name! perhaps it should be called "Engineer"
I am going to look into Solidworks in more detail as I believe these areas of simulation are much much better.
Trying to model contacts with a spring-damper model seems like it should work. It usually doesn't. Back in the early 1990s, video game developers tried to put physics engines into video games. The first games to do this were awful. (See this postmortem of Trespasser from 1999.) Things flying apart came up during game play. Working Model 2D had the same problems.
There are two basic approaches to solving this problem. One is to solve multiple collisions as a linear complementarity problem in an impulse/constraint system. This is how most games do it. The results are somewhat unrealistic, but fast to compute. With this approach, all collisions take place in zero time, and objects reverse direction instantaneously in a collision, even if simulated in slow motion. The effect is that light objects and heavy objects behave the same way, and is why bounces in video games look unrealistic.
The other approach is to model contact as forces exerted in a spring-damper model. As you've observed, if you use a linear spring rate, you have a tuning problem. If the spring is too weak, not enough force is developed to stop the moving objects before they interpenetrate too much. If the spring is too strong, the numerical integration becomes unstable. Integrating a "stiff" system of nonlinear differential equations is a difficult problem.
The solution to this is to use non-linear springs to simulate contact. The non-linear function must increase faster than linearly with penetration depth. The spring forces become very large if excessive interpenetration occurs, but are small for light contacts.This is physically reasonable for many contacts; if you force two elastic spheres together, the contact area increases roughly as the square of the penetration depth, so the force goes up with the cube of the penetration depth. Exponentials work better numerically, although they're not quite as realistic physically.
This makes the system of equations being integrated almost tractible. The integration still has a huge dynamic range, and it's necessary to have elaborate checking on overflow and stability to keep the computation from failing. Double precision arithmetic is required, which is why this approach isn't used on game consoles. They usually lack a fast 64-bit FPU.
We solved this problem back in 1997, and here's the result: Spiral Stair Fall. This is the demo that started the cliche of ragdoll physics - a body falling downstairs. As a further demo, we did a large, heavy legged mecha. That's simulated with the creature having the density of water, and you can see the slow, graceful fall, bounce, and collapse.
There's been considerable progress in the field since them. If joints are coming apart in simulations, the mathematical approach is probably inadequate.Solutions to that problem exist. They are non-trivial.
John Nagle / Animats
thank you for your reply, I am starting to understand how this is working more now and I can understand how spring-damper is used in the simulation. I understand the potential variants in each model.
I assume that different material types and configurations will actually have set values, i.e a solid steel square bar 100mm x 100mm will always have set figures based on what it is, and a brass round bar 50mm round the same, a metal box, a steel ball, a wooden ball, a rubber ball etc... is this just a case of getting a load of data and setting it up?
I mean any of the above will have the required values once worked out and by changing the material, length, width, diameter, thickness etc.. it should be simple math to adjust the calculation.
hugh quoted in an earlier email
"The critical damping is equal to two times the square root of the stiffness multiplied by the mass."
which is a simple known calculation by autodesk
so with the size and power of autodesk is it not a case of acquiring and inputting a load of data which relates to the parts material settings (most of which is already in the software)
then when running a simulation in 3d space with gravity, there is no real guess work?
I have come across a few physics simulators and a lot of them are programed by very small companies. and they look amazing. I think with modern engineering this needs to be built into the likes of inventor especially as Autodesk has the resources to do it and do it well.
Inventor already has the simulator platform but it desperately needs improving.
the potential to have these tools to hand is huge, there are a lot of inventors and designers in the world that are held back because they are practical as opposed to physic scholars, by putting these tools on our computers and making them easy and quick to use with accurate results, will make better designers, inventors and creators!
I would like to know the direction Autodesk is going in with development as it will affect my software decisions.