Can Engineering Simulation Software simulate full systems?

[amrbekhit]

Hello all,

I have been looking at the websites of various simulation packages, such as the Solidworks Simulation features, ANSYS, COMSOL and LS-DYNA. Although these packages can simulate many different things, it wasn't clear whether these packages can simulate entire systems. All the examples I found on the websites tended to focus on just one physical aspect, such as flow, mechanical stress etc.

For example, I am designing an underwater robot that moves inside pipes. I would like to be able to model the robot in Solidworks, model the control system in MATLAB and Simulink, create a virtual pipe with fluid flowing through it, place the robot model inside the pipe and study how well the control system moves the robot, how the flow of fluid in the pipe affects the robot, how the various actuators in the robot are working etc. Essentially, I want to build and test as much of the robot as possible holistically in software before building physical prototypes.

Is this possible with software packages out there at the moment? If so, are there any which are more "complete" than others? Or are they pretty much similar in their features (like, for example Solidworks and Inventor)?

--Amr

 

[IRstuff]

No, because most systems are not designed by a single person or discipline. For your specific application, Matlab is the way to go.

What you appear to be asking for would involve designing something, then modelling it, and then designing it over.

What most systems engineers do abstracts the design to the basic fundamentals of the problem, and then levies design requirements onto the specific disciplines.

TTFN

FAQ731-376

 

[GregLocock]

Well, we use ADAMS to model the behaviour of a car, with tires, aero, and the (ESC) control system, so yes, it can be done. ADAMS makes it relatively easy to hook outside software such as matlab into the model. I can include flexible bodies so i can study the stresses at the same time, but that rapidly bogs the model down.

Your specific problem adds a layer of complexity in that you also want to run a CFD analysis at the same time. This is an order of magnitude more calculation than is required for the rest, and will have to be repeated for every time step of your model, which I think is going to take years.

As IRstuff implied, what real engineers do is system engineering, so in your case the control system and motion through the water would be analysed by a different model to the stress model, etc.

None of the mickey mouse programs will do what you want.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[SomptingGuy]

I can't speak for your industry, but in mine (auto), Simulink has become the de-facto standard. If a simulation can't co-simulate with Simulink, it's not worth considering. Likewise, 3D CFD codes must link with 1D codes. Many system models consist of Simulink at the top, with complex physical models solved in their own local environments.

DONT try modelling the entire world in Simulink.

- Steve

 

[FeX32]

Many packages can couple physics, even implement a control algorithm within a custom modified code. Although, the effort in performing what the OP mentions is far greater then actually conducting an experiment imo.


Fe

 

[GregLocock]

Perhaps in more detail this is how I would approach your problem

1) define the required manoeuvres.

2) measure the loads in the actuator shafts caused by those manoeuvres, using CFD.

3) hand those loads over the the structures guys, who will then work out the stresses internal to the machine.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[amrbekhit]

Hello all,

Thanks for your advice. Unfortunately, as this is an individual project, I'm not working with experts in the individual fields. Sounds like trying to simulate the entire thing is not going to be practical. I'll look into splitting it all up.

--Amr

 

[GregLocock]

That makes things easier, really. One approach would be to work out the forces/torques in the control shafts at maximum speed and maximum deflection, and use those for your strength calcs. If this is man rated you want a factor of safety of 4, less if it is a pure robot.

On top of that you need to consider abuse cases like collisions while it is being lifted out of the water, or if it drives into something. They'll be significantly greater than the control forces.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[MechEng2005]

I think it would be wise/necessary to find a way to split it up. The things you describe become quite computationally intensive quickly, even when you don't incorporate the interactions between them (i.e. flow causes body to move). Often, models are made neglecting certain things to help simplify the process. Even if you COULD get a program to simulate all aspects, you are asking for a massive program, at the same time limiting the amount of simplification you can provide for your model. For example, a fillet that has little impact on stress results may be very important for fluid dynamics.

-- MechEng2005