Soft Sensor based on Fuzzy Model Identification
| Authors: | Nagai Elaine Y., Federal Center of Technological Education of Parana / CEFET-PR, Brazil
de Arruda Lucia Valeria Ramos, Federal Center of Technological Education of Parana / CEFET-PR, Brazil |
|---|
| Topic: | 3.2 Cognition and Control ( AI, Fuzzy, Neuro, Evolut.Comp.) |
|---|
| Session: | Soft Sensors and Predictive Control |
|---|
| Keywords: | distillation columns, estimators, fuzzy modeling, neural networks, soft sensing |
|---|
Abstract
This paper presents an approach to build a soft sensor based on computational intelligence techniques. The goal is to identify fuzzy models from numerical data. First of all, the fuzzy model input variables are selected from the pull of secondary variables set by applying Kohonen maps. Then, the Lipschitz quotients are used to select the lag structure of the fuzzy model. A fuzzy clustering algorithm is applied to find an initial rule base, and, to conclude the identification process, this initial rule base is simplified by merging the similar membership functions. The validity of the proposed identification method is demonstrated by the development of a soft sensor to infer the top composition of a distillation column.
