Naslov
Soft sensors for Estimating Crude Distillation Product Properties

Autori
Buhanec, Gregor ; Lukač, Goran

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Skup
2010 Annual Conference on Student's Research Societies of the University of Pannonia

Mjesto i datum
Veszprém, Mađarska, 10.11.2010

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
soft sensor; crude distillation unit; multivariate regression; neural network

Sažetak
In this study soft sensors for estimating crude oil distillation properties have been developed. Based on available continuous measurements of temperatures and flow rates of certain process streams, the soft sensor for estimating initial and end boiling point of heavy naphtha of splitter unit has been developed. Research has been conducted based on experimental data from an industry plant. Several linear and non-linear models have been developed using multiple linear regression and artificial neural networks. Statistical analysis of data has been conducted, and results have been critically judged. Non-linear models, based on multilayer neural networks, are proved to be suitable for prediction purpose. Results show that soft sensors can be used for estimation of product properties as an alternative to laboratory analysis. In this way, it is possible to continuously estimate properties of different refinery products, and also implement methods of inferential control in order to optimize industry plant operation.

Izvorni jezik
Engleski

Znanstvena područja
Kemijsko inženjerstvo

POVEZANOST RADA