If modeling plus 3d visualization is what you need, I would recommend COMSOL Multiphysics, it happens to be what NASA is using right now, it is a new software tool but it supports mixing a lot of aspects to the modeling process that include mechanical, electrical and other physical aspects of your model, also it has extension to support most of the CAD tools available on the market and it has MATLAB link so you can do part of your modeling in the MATLAB environment which allows more customization if you want, optimization is also available.

http://www.comsol.com/

Well, what I am after is integration of models from multiple sources.  COMSOL + MATLAB make a rather limited portfolio of model sources.  Most of the mechanical engineers I need to be working with use 20-sim and possibly SimulationX.  We are not going to make them switch tool, but we can exchange behavioural models using FMI (functional mockup interface).

But let me ask about the CAD tool integration.  Is that plug-and-play, or is it a tedious, manual process to integrate?  

The link below contains of the details of the modules that COMSOL multiphysics use:

http://www.comsol.com/products

There are CAD import modules, and there are live link to common CAD packages; booth for interfacing and booth with FMI in mind.

Also I would like to point out that usually COMSOL alone is enough and the MATLAB live link is just there as a helping tool, in general COMSOL was designed at first as just a PDE solver with no or very little coding, now it includes support for common physics and have solvers that use finite element, boundary element, finite volume and ray tracing approaches for solving different multiphysics problems

Are you seriously suggesting that in order to simulate a complex and composite system you would recode every subsystem model in comsol, rather than building the system from existing building blocks?

The vision of the FMI project 2008-2011 and MAP (Modelica Association poject) is slowly being realized by software developers and by the companies producing the software packages for modeling, but for the time being not every software that is used in the design process is following the FMI standard or able to produce a functional mock-up unit that can be then used in another part of the modeling and simulation process, there is a handful of modeling environments that support FMI including MATLAB and SIMULINK, there is also CAD software packages that support FMI like CATIA, not necessarily COMSOL is what I recommend but you will need a modeling platform that either fully supports FMI or can communicate with the various development tools that were used to develop the existing building blocks of your system created by other engineers in their specific field.

A possible approach for coupling output from multiple computational sources into a single system is currently being pioneered in the AMUSE project at Leiden University, the Netherlands:

http://amusecode.org/

In this case, they use a python wrapper to integrate a variety of codes and, if necessary, allow them to run simultaneously, using each other's data. (For example, an N-body code to calculate a gravitational field and a hydrodynamics code to simulate the behaviour of gas inside the field as it evolves.) The data structure that allows the communication is based on MPI, which allows each code to write its own output without interference and, when necessary, to read the output of another code.

Dear Hans Georg Schaathun,

If you and your people work with 20-sim, etc, then those tools can be used for visualtization. I know that 20-sim has quite some facilities for that, especially from a multi engineering domain / mechatronic context. It now has scripting facilities, making it easier to connect to other tools.

An alternative could be to use Gazebo and ROS maintained by the Open Source Robot Software foundation. They are focussing on robots, which might distract you, although robots are mechatronic systems.

Thank you both.  ROS, Gazebo and AMUSE are useful pointers; I shall investigate further.

I know that 20-sim can visualise, and it works very well when the mech engineers design the entire system and just wants to visualise their own work. It breaks down when mech, elec, and comp engineers make different components for the system. Neither EE og CE find 20-sim a natural tool to use. It is /possible/ to integrate 20-sim with other tools, but only by having the comp/elect engineers take substantial time to learn the tools for a different trade (20-sim).

And this isn't particular for 20-sim. Most tools, and especially commercial tools, prefer t lock-ins over collaboration.

BTW.  I am interested in the design and development of tools and frameworks.  Mechatronic modelling, I leave for others.