Restricted Complexity Approximation of Nonlinear Processes Using a Control-Relevant Approach

Abstract A two-step nonlinear system identification method using restricted complexity models (RCM) is proposed. In the first step, a parsimonious yet full order Volterra model is identified using the orthogonal least squares method. In the second step, using a control relevant approach, the full order model is further reduced to a restricted complexity model which is more amenable to control design and analysis. The minimization problem in the model reduction step is posed such that it can be solved using general optimization routines. A corresponding two-step model validation procedure is implemented to ensure the closed-loop performance of the resulting model. Effectiveness of the proposed method is illustrated by a polymerization reactor example.