3D viewing of model sensitivity properties in MIMO controllers

An intuitive 3D graphical mapping is proposed for a problem of model sensitivity analysis to variations of model parameters. This problem appears in a class of linear optimization applications used in the steel and aluminum industry, though similar problems are at the core of many research and practical applications. Understanding the effects of parameter variations in linear optimization problems allows control experts to select multiple input-multiple output (MIMO) control policies that better match existing plant conditions. Introducing mathematically-oriented tools of this type is a significant departure from the existing industrial process visualization style: existing human-machine interfaces are based either on the implementation or on the physical structure of the control system - mainly for the purpose of an easier fault detection and isolation. By contrast, the proposed method capitalizes on the mathematical model used for the MIMO control and it is thus a truer predictor of the effects operator changes will have on the system.