We use Ventana Vensim for System Dynamics modelling and simulation. It is cross platform (Linux, Mac, Windows), and you can easily exchange files between these systems. Besides that, Vensim is pretty much standard - I believe that >80% of the SD community actually uses it. The GUI is useful, but a little bit old-fashion, in particular when compared to AnyLogic, for instance. The simulation runs stable, we never had problems with that. Optimization capability is rather limited (just local search on single parameters). And a big plus for teaching and personal use: there is a "Personal Learning Edition" that is free of charge and has most of the features of the commercial version.

Thanks for your answer, Professor Armin Fügenschuh.

From your experience with models involving partial differential equations, linear mixed-integer programming and continuous optimal control, what can you say about  Vensim interface and its resources for these purposes? How easy, user-friendly and efficient is this software?

Vensim is only useful for System Dynamics models, that is, models described by ordinary differential equations and algebraic state equations. It is not possible to model partial differential equations, or mixed-integer linear programs. Also, Vensim's capabilities for optimal control is rather limited (only local optimization of constant control is possible). If you can live with these restrictions, then you find in Vensim a user-friendly and efficient software, together with a lively community supporting it (the annual SD conference starts tomorrow in Delft, see conference.systemdynamics.org for more information).

Thank you very much! The conference seems to be very interesting, indeed.

Vensim is mostly convenient for numerically solving differential equations that are not easy to write down explicitly. For discrete event simulation of logistical systems it is not appropriate. For example, you cannot simulate an M/M/1 queue with vensim.

Olá Wederley.

Que mundo pequeno! Você conhece o software XMDS2? Pelo que vi da sua pergunta, pode ser que lhe seja útil. Tenho um artigo publicado que o utiliza...

Grande abraço,

Renato. 

Vensim can be used to integrate partial differential equations, finite element methods are essentially subscripts and matrices, easily implemented in vensim by Vector Select functions (see manual). Optimal control problems, even in uncertain systems, are trivial in vensim (it has Kalman Filter and Fred Schweppe statistics). However, Vensim does not support conceptual mistakes, like finding “optimum” f(t) because time cannot cause anything, structure cause behavior, - the optimum is a structure not a number-, f(t) is suboptimal, so you have to build the structure g(t) –levels, rates and parameters- where  f(t) come from, then you find optimum parameters of g(t) by a mouse click, results from g(t) >= results from f(t), g(t) is a structure, so it tells you how to implement the optimal control.

If you have further questions do not hesitate to contact me, dfranco45 gmail

Vensim and any System Dynamics software do not use any analytical expression of solutions to differential equations, ordinary or partial, the reason is that existence theorem of solutions of ordinary and PDE, proves, by Picard Iterates, that all solutions are numerical, so the so called analytical solutions are numerical solutions before the advent of computers, learnt by heart

Vensim PLE is a free version very limited, so it does not support VECTOR SELECT function

Ordinary and Partial differential equations can be solved by Vensim, the firts one are obvious, the second one can be integrated either by separation of variables, or by the finite element methods which are trivial by using Vector Select function, as described in Vensim Manual.

However, it does not make sense to use partial or ordinary differencial equations if you have vensim, since it allows a better formulation of solutions to PDE or ODE.

thank you M. Armin Fügenschuh

Vensim has now Ventity an organized way to implement Agent Base Simulation, where the relationships among agents is structured. Stochastic modeling is absolutely trivial, all kinds of random numbers can be generated to implement montecarlo approach. Even advanced uncertain systems which require Kalman Filter or Fourier Transform, can be handled by a mouse click.