Control relevant identification for third-order Volterra systems: a polymerization case study

In this work the problem of identifying third-order Volterra models from input-output process data is addressed. The identification problem involves the rational design of input sequences that exploit the Volterra model structure. The criterion used to measure the model fitness is the minimization of the prediction error variance (PEV). Explicit estimators that utilize plant-friendly input sequences for the identification of bias, linear, nonlinear diagonal, and third-order sub-diagonal kernels are presented. As an application of this technique, an isothermal polymerization reactor case study is considered; it was found that the third-order Volterra model does an efficient job of capturing the nonlinear reactor behavior.