Slashdot Mirror
Free Software for Chemical Process Simulation?
chthonicdaemon asks: "I am about to embark on a mission: to create a Free (as in freedom) chemical process simulator. My field is control engineering, and as such, dynamic simulation is what it is all about. There is an open source steady-state simulator called sim42, but I have not found one that can do dynamics. There is also a lot of work available in the form of implementation standards, so I am confident that it can be done. Is there anything like this out there that I have not found, and what open technology could I leverage to get this done faster (perhaps something like SML)."
Ok, not my field, but what about NAMD and VMD, a molecular dynamics program suite created by some people on the floor below me?
The Sim in my user name refers to process simulation, this is what I do. Mostly steady state but I do do some dynamic work. I design oil refineries.
If the submitter is following the thread still, I'd be interested to know more of what exactly they'd be interested in using this for in control systems. Are they interested in tuning controllers, training simulations, RTO? Different needs require different tools.
Most of the important points have been made. Thermo package is going to be a big concern. A lot of money is invested in these things. The algotithms are often published, but the coefficients and implementations are proprietary. Without good reliable thermo, your results won't be of much interest to anyone.
For dynamics, another concern is flow driven versus pressure driven. Each has it's advantages and more importantly serious compromises.
Unit operations aren't actually that bad to model mathematically. Having a good flash algorithm will go a long way. Converging distillation with any kind of speed will take some cleverness, but is conceptually pretty simple. It is just scaling the VLE over many components.
I've written simple stuff in Matlab pretty easily. Binary distillations, as an example.
The biggest hurdle you will face is crediblity. Aspen, Simulation Sciences, and the like have been around a long time and have a pretty good track record. Engineers are comfortable with them enough to base design decisions on them. Even at that, any time a new version is released we are probably 6 months away from adoption due to our validation cycle.
The company I work for is really near the top of the game as far as modelling refineries especially. And even with all of our experience and expertise, we are really wary of dynamic simulation. The concepts are really easy, and it doesn't seem like that big of a jump from steady state. But we take an awful long time to develop a dynamic simulation to the point where we are comfortable using it for design work. So much of it has to be hand-crafted as opposed to the simple drop-and-connect of many of the steady state packages.
Long development cycles are big dollars, and most clients aren't prepared to spend them unless the carrot is big enough. Relief system modelling is where we see it the most often.
I hope I am not being too discouraging here. Engineers by nature a fairly cautious lot. If something I design screws up, there are dead people and a smoking hole in the ground.
Oh, and if the question of sequential modular versus equation oriented doesn't mean anything to you, I would walk away.
Laugh while you can, monkey boy!