PlumX Metrics
Embed PlumX Metrics

Design and implementation of a neural network based soft sensor to infer sulfur content in a Brazilian diesel hydrotreating unit

Chemical Engineering Transactions, ISSN: 2283-9216, Vol: 17, Page: 1389-1394
2009
  • 8
    Citations
  • 0
    Usage
  • 3
    Captures
  • 0
    Mentions
  • 0
    Social Media
Metric Options:   Counts1 Year3 Year

Metrics Details

Conference Paper Description

The diesel hydrotreating (HDT) process in refining oil plants is a conversion process responsible for the specification of this product in oil industry. In this work, the objective was to estimate sulfur content in the outlet stream of the unit, using inferences based on heuristic modeling. Neural networks (NN) were used to correlate the sulfur content, measured offline in laboratories, with variables measured on-line (as temperature and flow rates) in the reaction section of the HDT unit. Historical data was loaded from Petrobras (Brazilian Oil Company) Duque de Caxias Refinery (REDUC) in Rio de Janeiro and treated in order to remove outliers and reduce dimensionality. After that, twenty-four different designs of neural networks were trained to find out the best fit to real data. The chosen neural network was implemented in the refinery's data storing and acquisition system. Very good predicitons of sulfur content were obtained indicating the use of this inference for advanced process control. Copyright © 2009, AIDIC Servizi S.r.l.

Bibliographic Details

Luciana Salvatore; Maurício Bezerra De Souza; Mario C.M.M. Campos

Chemical Engineering

Provide Feedback

Have ideas for a new metric? Would you like to see something else here?Let us know