Some model-based controllers are nonlinear, such as those based on neural networks. But for empirical models to capture the range of process behaviors, they must be calibrated with extensive process testing throughout all operating conditions. In contrast, first-principles models express process nonlinearity. These models could be considered digital twins, as the same model that is useful for control would support many other analysis and forecasting applications. This article by ISA Fellow Russ Rhinehart explains first-principles models.
It comes from his new book, published by the International Society of Automation, titled “Nonlinear Model-Based Control: Using First-Principles Models in Process Control,” which goes into greater detail by providing illustrations and set-up examples.