Using Optimization to Detect Snowball Effects

In process plants with recycle streams some level control architectures are inoperable. Their use leads to large excursions in manipulated and/or controlled variables and this behavior has been termed the snowball effect. The snowball effect is a steady state phenomenon and it can be analyzed using steady state process models. In this paper a steady state mixed integer nonlinear programming (MINLP) approach is used to analyze for large excursions in process variables. This MINLP approach can be used to detect the likely occurrence of a snowball effect in a plant and to develop control architectures that can avoid this problem. The optimization-based approach is illustrated on a threereactor three-distillation column plant taken from the literature.