Efficient Use of Computational Model Data Via "Zonal" Modeling

Computational models have become an integral part of the modern chemical engineer's toolkit. Such models are very powerful in representing and simulating the fundamental physics of a given process. For instance, computational fluid dynamics (CFD) and discrete element method (DEM) models are used extensively in studies relating to flow and mixing behaviour in fluid and granular processes respectively. One of the key challenges of using CFD and DEM models is their inherent computational complexity. Typical simulation run times for these models range from a few hours to a few days (esp. for DEM models). Due to this reason, such models tend to not be available for widespread consumption in the larger process engineering community.

We present a methodology here, variously called "network-of-zones", "compartment-based" and "multi-block" modeling that enables a reduction in the complexity associated with such models. Zonal modeling is essentially a spatial-averaging of simulation data that allows for information to be represented at just the right level of granularity. We will describe this approach and its relative merits as compared to traditional dimensionality-reduction techniques such as PCA. We will present some examples of application of the methodology to practical engineering problems - and also indicate extensions where it can used in conjunction with techniques such as PCA, SVM etc.