Fuzzy Activated Neural Models for Product Quality Monitoring in Refineries
| Authors: | Xibilia Maria Gabriella, Università di Messina, Italy
Fortuna Luigi, Università di Catania, Italy
Graziani Salvatore, Università di Catania, Italy
Barbalace Nicola, Università di Messina, Italy |
|---|
| Topic: | 6.1 Chemical Process Control |
|---|
| Session: | Process Control Applications |
|---|
| Keywords: | Neural networks, products industry, quality control, nonlinear models, fuzzy models, identification algorithms |
|---|
Abstract
In the paper the problem of estimating the octane number of powerformed gasoline produced in a refinery is addressed. The model is designed in order to replace the existing measurement device during maintenance operation guaranteeing the continuity of product quality monitoring and control. Linear and nonlinear Moving Average models based on MLP neural networks have been designed to take into account the two different working points of the process and different strategies are compared. The models obtained are presently implemented on line in the refinery to be tested over a long period.
