An intelligent signal validation system for a cupola furnace. I. Methodology

We present a methodology for developing a signal validation technique that can be introduced to improve the operation of the cupola iron-melting furnace. The operation of the digital controllers used depends on the measurement accuracy of the controlled variables. We develop a signal validation system for one of the process variables of the cupola furnace, namely iron temperature. An artificial neural network (ANN) rule-based filter and trend estimator is developed to estimate the measurement signals and to eliminate spikes and other external disturbances from the measurement signals. Analytical redundancy is provided through the use of inferential sensors developed from the identification of input-output dynamic models for the iron temperature using ANN. Another type of inferential sensor that relies on the identification of nonlinear relations between the iron temperature and another temperature measurement across the furnace body is also developed.