New optimizer features

Sunday, February 21. 2010

I've managed to include some very useful features into the Optimizer. One of them is the ability to define constraints other than the independent variables. This is achieved by using barrier functions. The other is the inclusion of more numerical methods to solve the optimization problem through the DotNumerics library. You'll also have the option to define how the numerical derivatives will be calculated (2 or 4-point) and the barrier function multiplier, which defines how close to the constraints the optimizer will go.

I've also redesigned the interface in order to keep all information on the screen while solving the problem:

Posted by Daniel Medeiros at 09:22 | Comments (5)

Comments

Display comments as (Linear | Threaded)

Hats off, Daniel, for some great work on what appears to be a phenomenal project!

This is exactly what the process engineering industry needs. A friend and I began talks recently about developing a software package for chemical engineering that actually makes sense and is easy to use.

In our research we came across your project and were incredibly excited to find it. We have some pretty good ideas about where to go from here. I can't wait to download the source code and poke around. We would love to share some of our ideas with you and hopefully contribute some code.

#1 Adam on 2010-03-15 14:24 (Reply)

I have evaluated DWSIM and I wish to thank Daniel for this project, herebelow some comments:
a) I understand that the project is complex but graphics altough nice are very difficult to manage, other tools (notable the software for fluid flow simulation) give a much better feeling, at first I would recommend to redefine this section, that would make DWSIM a great tool
b) thermodynamics do not work but really I don't care for, in most cases I need vapor liquid equilibria and PR / UNIFAC work fine, for more complex tasks there are the $$$ tools, much better to devote time for improving distillation columns, reactors etc. etc.
c) a final comment, I fear the risk is that project grows too fast, better to devote more time for debugging existing code,
Marco

#2 Marco on 2010-03-16 07:05 (Reply)

Marco,

thanks for your comments. I didn't get your comment about the GUI, please explain it a little better...

Regarding the thermodynamics, there are some parts that works and others don't, mainly because of the simple flash algorithm which is used in DWSIM (and not that the EOS or activity coefficient implementation might be broken...). Now that the basic unit operations set is there, I'll focus on improving property packages and feature validation.

#2.1 Daniel on 2010-03-17 21:42 (Reply)

Daniel,
a) I mean the capability to easily define topologies of medium size plants (say 50/200 equipments/pipes) or small networks (for example flare collectors etc.), the actual graphics are very nice but (in my personal opinion) difficult to use, WinSim or ChemCAD seem much more immediate, with a proper graphical inerface DWSIM would become a really great tool
b) for me thermodynamics in DWSIM are much more that required, in 99% of working time one needs only a basic VLE flash, a equation of state (PR or SRK would be fine) and a model for activity, more important are emphasis on stability and speed of solvers, a simple plug in mechanism would permit professors to add esotic models, where required (note that there are several new variants every year so it's really difficult to code even a limited number of models much better to provide a simple interface for external tools), note that there are many good tools as PD-PLUS (which I used once) with very limited thermodynamics but reliable solvers, this is a very important aspect, in general I think it's much better to get a column reliably converged with ideal model than attempting to apply complex models and get unrealist results, for those cases it's better to use specifis tools with specific databanks
c) with a stable and simple to use graphical engine one could think to extend DWSIM with good global solvers (something able to solve mass + energy balances in a small network), that would make DWSIM really interesting for MANY applications
best regards,
Marco

#3 Marco on 2010-03-18 05:46 (Reply)

I see, Marco. Thank you very much for your comments and suggestions. Be sure I'll consider them on future development of DWSIM.

Best regards,
Daniel

#4 Daniel on 2010-03-18 19:06 (Reply)

Add Comment

Name
Email
Homepage
In reply to
Comment	Enclosing asterisks marks text as bold (word), underscore are made via _word_. Standard emoticons like :-) and ;-) are converted to images.
	Remember Information? Subscribe to this entry

About DWSIM

DWSIM is an open-source, steady-state chemical process simulator written in VB.NET. It has a HYSYS-like interface with many of the features present in commercial simulators:

- Thermodynamic models: Peng-Robinson, Peng-Robinson with Volume Translation and Immiscible Water, Soave-Redlich-Kwong, Lee-Kesler, Lee-Kesler-Plöcker, UNIFAC, Modified UNIFAC (Dortmund), UNIQUAC, NRTL, Chao-Seader, Grayson-Streed and Ideal Gas;

- Unit Operations: Mixer, Splitter, Separator, Pump, Compressor, Expander, Heater, Cooler, Valve, Pipe Segment, Shortcut Column, Heat Exchanger, Reactors, Component Separator, Orifice Plate, Distillation and Absorption Columns;

- Utilities: Phase Envelope, Hydrate Calculations, Pure Component Properties, Critical Point, PSV Sizing, Vessel Sizing, Spreadsheet and Petroleum Cold Flow Properties;

- Tools: Hypothetical Components Generator, Bulk C7+ and Distillation Curves Petroleum Characterization, Reactions Manager and
Component Creator.

- Process Analysis: Multivariate Constrained Optimization and Sensitivity Analysis utility.