I'm postponing the model overview to discuss some conclusions regarding FCC models. It appears that we may have another answer to the question: "How do you determine which flow model of the regenerator is best?" My original answer was to do a tracer study and determine what flow model fits the data best.
Two models for comparison
We now have two models to compare. The first model (Model 1) is the PFR-CSTR-PFR model with the Dasila et al. paper as an example. The second model (Model 2) is the PFR-CSTR-PFR model I have developed. These are very different models even though the reactor types appear to be identical.
Model 1 assumes plug flow of the gas through the dense bed. That gas reacts with catalyst assumed to be in a CSTR, well mixed phase. This CSTR phase has no gas. The third sub-model is the plug flow of gas and entrained catalyst in the freeboard region.
Model 2 assumes plug flow of gas in a bubble phase with no catalyst. The gas-catalyst dense emulsion phase is assumed to be a CSTR. The third sub-model is like the Model 1 freeboard model.
A major difference
The two models can produce nearly identical oil yields and flue gas compositions. The major difference is the level of coke on the regenerated catalyst. Model 2 coke level is about 20 times higher than the Model 1 value.
The reason for the difference
The overall oxygen level in the dense bed plug flow reactor is much higher than the level in the CSTR. Since both models have to remove the same amount of coke that was added in the riser reactor (same yields), then something has to compensate in the reaction rates for the difference in oxygen levels. The possibilities are temperature and coke level.
The bed temperatures should be nearly identical if the flue gas compositions are identical. The same amount of coke is being oxidized to the same products, therefore producing the same amount of heat. Also, both models assume the same amount of gas and catalyst remove this heat. Both models also assume that the gas and solids have a uniform temperature. Thus, bed temperature for a single case will not separate the models. After some additional fine tuning of the coke reactions since the last post, there is only a 10 C degree difference between the bed temperatures predicted by the two models.
Since temperature isn't compensating for the different oxygen level, then the coke level in the regenerator has to be higher in Model 2
Determining the better model without the tracer study
For these two models, it appears that comparing the coke level on the regenerated catalyst would be a good way to select the better model. In order to do so, the following are required.
You might think that step three is not a good idea because it might make either model look good. However, for the reasons discussed in the previous section, the coke level will still be the discerning variable. If you have just one model, a comparison with the data for the coke level should tell you if you have the right model.
Another way to select the models
You will note that the two models responded differently to increases in air flow. The bed temperature increased with Model 1 and decreased slightly with Model 2. Thus, a set of data at different air rates could be used to select the model. You may also find other variables that have different responses to a change in an operating variable.
There, however, is one reason that we can't use this method with the two models in their current state...they have different decay functions for the reactions in the riser. Model 1 does not respond to changes in coke level, whereas Model 2 does respond. It is very likely that the parametric response to air rate will be different if the decay function in Model 1 were changed to a function of coke instead of a function of residence time. Note that the decay function doesn't affect the comparison of the coke level for a single data set. The reason is that I specified that the two models produce the same yields. Rate constants for the reactor could compensate for the decay function for a single data set, but not for a parametric study.
Which do I think is the better of the two models?
When I started this project, I was not a big fan of Model 1 because there was no backmixing of gas along with the catalyst. However, seeing the video of a CFD model of one regenerator has changed my mind. (I discussed the video here.) In that model, it appeared that the dense bed of catalyst might be more like a fixed bed with plug flow upward of gas. I don't think the catalyst was well mixed, but the possibility of plug flow for the gas does support Model 1.
I have discussed and compared just two models. There are others. It appears that the best idea is to build the models you think most likely fit your reactor and then determine what variables can be used to select the best model using plant data. The alternative approach is to do the tracer study and make the selection using simpler versions of the models.
In the paragraph above I said "most likely fit your reactor." Reactors have different locations for catalyst feed and effluent and for air distribution. These differences will likely result in different flow behavior and thus different sub-models will be needed. If you have some experience with gas/solid flow, you may be able to select good model candidates.
Dasila, Prabha K, Indranil Choudhury, Deoki Saraf, Sawaran Chopra, and Ajay Dalai. “Parametric Sensitivity Studies in a Commercial FCC Unit.” Adv. in Chem. Eng. Sci. 2012, no. January (2012): 136–49.