In a LinkedIn discussion about non-CFD models (e.g. CSTR-PFR, PFR-PFR) of fluidized beds, the question was raised, "How can we determine the nearest approximation." Since the different submodels all relate to fluid flow, I suggested that a tracer study would be an ideal way of selecting the best combination. The other responses all gave opinions about the "best" model, but there were no other suggestions for methods to determine the best model.
As was mentioned in the LinkedIn discussion, fluidized beds have been studied for a long time. However, even if the best model is known, each reactor is different. Thus, the parameters relating to subreactor volumes and exchange rates will need to be determined for a given reactor. That's why the tracer study is still needed.
Why not use reaction data?
Reaction data may be used to determine the flow parameters provided that the reaction parameters are known independently. If the reaction parameters are estimated simultaneously with the flow parameters, you will probably end up with flow dependent kinetic parameters.
If you do have independent kinetics, the flow parameter estimation problem may still be difficult without a dynamic tracer study. A lot more data are required to do the estimation using reaction data.
Another advantage of tracer studies
The tracer method and the steady-state reaction method both result in parameter estimates for the flow models. However, the tracer study also produces the tracer data curve and the model tracer curve. A comparison of the tracer data curve with the model curve can be used to determine a model deficiency and possibly suggest a better model. The reaction method probably won't be able to suggest model improvements.
It's not RTD
Most of the reaction engineering books have a chapter on residence time distribution (RTD) studies and mixed models like I've been discussing. In that literature, the emphasis is on using either impulse or step input tracers. The exit tracer concentration curve is then converted to various normalized age distribution curves, I, E, C, or the intensity function. These curves are then used to determine such things as deadspace volume and bypass flow rates for simple models. I won't be discussing those methods.
Instead, I will be assuming different models and then using the tracer concentration curve directly to optimize the parameters. One advantage of this method is that the tracer input doesn't have to be a perfect impulse or a step increase, provided the actual tracer input function is recorded or known with reasonable accuracy.
Another advantage of the direct modeling approach is that the tracer model becomes the first step toward the full reaction model.