Robust Specifications for Practical Control Design. Insights from the Reaction Curve

In the context of linear time-invariant, model-based control, this paper exploits a straight-forward method for estimating the uncertainty arising from the use of an approximate nominal model to represent an actual plant. The approach relies on comparing nominal and actual time-domain reaction curves to quantify this uncertainty. Designed to support field control engineers and plant operators, the method enables the setting of realistic closed-loop specifications that reflect the true system behavior, while circumventing the complexities typically associated with conventional robust control techniques. The paper details the estimation procedure and demonstrates its effectiveness through simulated and experimental validation. First, a numerical simulation on a nonlinear three-tank system illustrates how different measurement noise profiles affect the derived robust performance and stability bounds. Subsequently, the method is validated using an experimental student-trainer laboratory setup. The results confirm that the derived specification bounds are practical and relevant for ensuring robust stability and performance in real-world applications.