SOFT ANALYSERS FOR A SULFUR RECOVERY UNIT

Abstract In this paper soft sensors for a sulfur recovery unit in the ERG PETROLI petrochemical plant located in Priolo, Italy, are designed to parallel the online analyser which is often taken off for servicing. Three strategy have been compared, namely neural networks, neuro-fuzzy networks and nonlinear LSQ techniques. The best performance has been obtained with a neural NMA model and the soft sensor is now installed on the plant for online verification.