Monday, November 7, 2011

An interesting presentation at 2011 AIChE National meeting

I’m privileged to be able to regularly attend the AIChE Spring and National (fall) meetings. I attend presentations, meet with customers and prospects, and attend division dinners. Chemstations has been a long-time exhibitor at these meetings, and often sponsors the meeting. I’m vice-chair of the admin committee for DIPPR (www.aiche.org/dippr/), which often meets during the National meeting.

One of my favorite papers at the 2011 National meeting in Minneapolis was a presentation by Dr. Becky Toghiani titled Group Contribution Methods in Undergraduate Chemical Engineering Thermodynamics. Dr Toghiani is a professor at Mississippi State; we’ve known her for years and admired the way she teaches her students the fundamentals. In her presentation, Dr Toghiani discussed one way she has her students use computer programs (CHEMCAD and Cranium*) to learn about the limitations of predicting chemical properties.

Several of the exercises involve predicting well-known chemicals and comparing the results to database values. The students start by comparing predicted pure component properties to experimental data, noticing that if Tc, Pc, and Tb are not well predicted, other properties that are correlated from these values (VP, Cp, HOV) will not be well predicted either. This shows students some of the limits of correlations.

The students then move on to mixture behavior, where they model the separation of binary mixtures using UNIFAC to predict non-ideal VLE. After modeling with predicted chemicals, they look at experimental VLE data to see how well the predicted behavior matches the system. One of the systems is a mixture of o-xylene and p-xylene. Use of a first-order subgroup model (such as Joback) to predict these chemicals will lead to these predicted isomers having the same properties and UNIFAC subgroups, which means they cannot be separated. Use of a second, order model will create slight differences in critical properties and allow for (very difficult!) separation. Naturally, the students are not warned that “cannot be separated” is the wrong answer; this is a process to help them learn to think.

These problems teach Dr Toghiani's students to remember “what is behind the curtain” when they are using modern tools and methods. The students also learn to inspect and think about their results, and ask “Should I try to approach this problem in a different way?” These are the types of lessons we really like; they show the power of a process simulator, and they also teach that a simulator is a tool, not a magical device. It’s important to understand the engineering concepts and the models underneath, even if you don’t have to do the math yourself.

* I should note that I was sitting next to a long-time friend of Chemstations during this presentation: Dr. Kevin Joback, President of Molknow (http://www.molknow.com/) and author of Cranium.