Visualizing Model-Based Predictive Controllers

Abstract

One common problem with information displays, particularly in process control, is that relevant data is often scattered across several, separate displays that obscure important relationships and fail to show event information. The current displays used for model-based predictive controllers demonstrate several of the problems that this kind of a design can incur. It is hard to get a good sense of the recent, current and near-future status of the controller (situation awareness), and it is difficult to make informed decision when making changes to the controller (putting data into context). This forces users to navigate around a virtual workspace and attempt to compile the data necessary to make an informed conclusion. We have applied several design principles to show how it is possible to re-represent data into hierarchical data layers that support the cognitive tasks of monitoring, diagnosis, and control. This design forms a coherent, coordinated workspace which helps orient users to problems in the controller, with direct navigation to supporting details.