Identification and control of Wiener-type nonlinear systems with applications topH processes

A recent trend in the literature of nonlinear system identification deals with cascade or block-structured approaches where the nonlinear system is divided into a series of dynamic linear and static nonlinear elements. In this thesis, a new and improved method for the identification of Wiener systems (i.e. a dynamic linear subsystem followed by a static non-linearity) is proposed with particular interest in the practical application of this method to pH processes. This interest in pH systems is motivated by the fact that they are frequently encountered in the chemical process industries and, due to their nonlinear and time-varying characteristics, represent a significant challenge in terms of identification and control.;The proposed nonlinear identification method follows a straightforward and effective approach to identify the static nonlinearity in the form of its inverse, which is advantageous with respect to control system design. The frequency sampling filter (FSF) model is chosen to represent the linear dynamic element of the Wiener system because of its advantages over the more popular finite impulse response model (FIR) and other transfer function models. The inverse of the process nonlinearity is identified as a polynomial or spline function representation. The proposed approach works with a variety of input excitation signals, including random binary and periodic. Very little prior information is required and orthogonal estimation methods can be utilized for automatic model order determination.;The proposed identification method is experimentally evaluated on a pH pilot plant with excellent results. A key benefit of this method when applied to pH systems is that it provides a new, simple and fast way of obtaining the titration curve, i.e. the pH static nonlinearity. A recursive estimation approach that can track on-line the changes in the time-varying static nonlinearity is also developed and experimentally evaluated. Based on the recursive estimation method, a new on-line titration device, placed in parallel with the pH process, is proposed. This on-line titrator is used to update the inverse of the time-varying titration curve in order to provide the correct nonlinearity compensation for linearizing feedforward-feedback control.