Most of the advanced nonlinear control algorithms require a model of the system to be controlled. Unfortunately, most of the processes in the chemical industry are nonlinear, and fundamental models describing them are lacking. Thus there is a need for the identification and control of nonlinear systems through available inputoutput data. In this article, we briefly introduce the input-output model used (polynomial ARMA models), and analyze its stability and invertibility. This paves the way to the development of a nonlinear-model-predictive controller. Implementation issues such as modeling of disturbance, state and parameter estimation are discussed. The theory presented is illustrated through examples.