Obtaining an accurate model has always been a challenging objective in implementation of Model Predictive Control, especially for nonlinear processes. Part of this work here proposed a model building methodology for a complex block-oriented process, namely a Hammerstein-Wiener system in order to meet such a demand. It is a general system of the more simple structures which are known as Hammerstein and Wiener. This methodology uses sequential step test training data determined from an optimal experimental design and simultaneously estimates all the model coefficients under nonlinear least squares objective function. It is evaluated using four process examples and is compared with a recently proposed method in three of them. Even with less frequent sampling, the proposed method is demonstrated to have advantages in simplicity, the ability to model non-invertible systems, the ability to model multiple input and non-minimum phase processes, and accuracy.

This class of modeling method is also being applied to model normal operation plant data. The common problem seen in this type of dataset including high multi-collinearities of the inputs and low signal to noise ratios for the outputs inhibit modelers to acquire cause and effect relationship. The second part of the work here is to introduced this modeling approach that is capable of developing accurate cause and effect models. It is a special application of the Wiener block-oriented system and the unique and powerful attributes of this approach over existing techniques are demonstrated in a mathematically simulated processes and real processes.