Spectral Graph Theoretic analysis of process systems: an application to distillation columns

The wide-spread implementation of Smart Manufacturing goals has given rise to increased use of highperformance computation. However, this has also created the need to understand complicated, large-scale hierarchical processes with a fresh perspective. To explore this, the paper presents a novel interpretation of process models as a graph of the underlying variables using Graph Theory and Graph Signal Processing. The construction of variable graphs is demonstrated using a distillation column model, and the associated network features are presented. Filtering properties of this variable graph using Graph Signal Processing is also demonstrated followed by a discussion of how the graph filters can be designed. The clustering properties of large-scale variable graphs arising out of a sequence of three distillation tray variable graphs is demonstrated, highlighting how the information is embedded. The article concludes with possible future applications and usefulness in achieving Smart Manufacturing goals.