A neural network based control of a simulated industrial lime kiln

This paper describes a neural network approach for control of a highly nonlinear multivariable kiln process within pulp and paper industry. In the lime kiln process the fuel flow rate and the draft flow rate through the kiln are used as manipulated variables. The controlled variables are the burning zone temperature and the exit gas temperature. A standard feedforward neural network is used as the process model. It approximates well and with sufficient accuracy the highly nonlinear dynamics in the lime kiln being used in a nonlinear control strategy. A neural-controller was designed to control adaptively the nonlinear plant. In the study, the quality of the control is analysed.