In an integrated oil company, the transport through pipes is the cheapest and most effective way to move several products from extraction to distribution.
Because most of pipes crosses areas of environmental preservation and zones densely populated. The pipes operation requests safety procedures and constant monitoring.
With in that context, this article proposes a simulation system for monitoring and detection of leaks in a pipeline (pipes that transports several products) that links a refinery to a harbor. The model used to represent the pipelines is based on paradigm of the artificial neural networks (ANN) whose training is made using flow, pressure, density, and temperature values. These values are measured by control equipment to gas, oily water, oil, and petroleum flows. Each flow possess its own dynamic characteristics of drainage that are very well defined, in spite of non-linear behavior.
Typically, the failures detection in non-linear dynamic systems, as a pipeline, can be done by hardware or software redundancy (or analytic redundancy). Basically, strategies of failures detection based on the concept of analytic redundancy can be classified in a quantitative process model or in a qualitative model.
In this work a rigorous quantitative model is used for pipeline simulation. In the monitoring stage, a simplified linear model is used, where the non-linear dynamics is approximated to a linear dynamics with a specified minimum error. Finally, a qualitative model based on ANN is used for detection of leaks.
BLOCK 2 - - FORUM 9 279