Industrial Kiln Multivariable Control: MNN and RBFNN Approaches

Abstract

Artificial neural networks have been recognized as a valuable framework for nonlinear identification and control. In this paper, we discuss and compare the use of two types of neural network architectures: (1) the MNN (multilayer neural network), and (2) the RBFNN (radial basis function neural network) for modelling a second-order nonlinear chemical process-a lime kiln in the pulp and paper industry. The simulation results showed that the MNN performs better in this practical case. Therefore, it was used in an IMC (internal model control) strategy. The neurocontroller was analysed with regard to its performance and robustness against disturbances. (9 References).