No abstract

Volterra series are perhaps the best understood nonlinear system representations in signal processing. They can be used to model a wide class of nonlinear systems. However, since these models are non-parsimonious in parameters, the symmetric kernel parameters are used. This model is used to evaluate identification of a pH-neutralization process. The aim is to use this model in nonlinear model predictive control framework. For this purpose various orders of the Laguerre filters and also Volterra kernels are tested and the results are compared in terms of the validation of these models. The results show that to have a good trade off between simplicity of the model and its corresponding fitness, the selected nonlinear Volterra model has the memory of 3 while the number of its kennel is 4. The VAF of this model is 99.63% which is completely acceptable for nonlinear model predictive control applications.

Model Predictive Control (MPC) is one of the most successful controllers in process industries. Process industries need a predictive controller that is low cost, easy to setup and maintains an adaptive behavior which accounts for plant changes, nonlinearities and undermodeling. To this aim, it is necessary to obtain a suitable adaptive modeling that can be easily used in nonlinear MPC framework. Experiments show performance advantages of Volterra series in terms of convergence, interpretability, and system sizes that can be handled. They can be used to model a wide class of nonlinear systems. However, since these models are in general nonparsimonious in parameters, in this paper the symmetric kernel parameters and Laguerre filtering are used to generate regression vector. The performance of the proposed method is evaluated by simulation results obtained for identification experiments of a pHneutralization process.