Topological preservation techniques for nonlinear process monitoring

Document Type

Conference Proceeding

Publication Date

1-1-2014

Abstract

• In this work, SOMs were applied to all aspects of process monitoring • Fault detection was performed by using a Gaussian mixture model of the normal region. • For the fault variable identification, a novel method was introduced using residuals of the deviation from the normal region. • Fault diagnosis was done by creating a map for each fault (MSOM), which makes incorporating new faults into the scheme much simpler. • Applications: • Tennessee Eastman Process (TEP) • Previous studies of the TEP have considered particular feed-step-change faults; here, an analysis of dynamic faults such as random variations, sticky valves and a slow drift in kinetics were included. • Preliminary results in an industrial system Results indicate that MSOM is able to improve upon linear distance preservation techniques and standard SOM-based approaches in process monitoring tasks.

Publication Source (Journal or Book title)

Computing and Systems Technology Division 2014 - Core Programming Area at the 2014 AIChE Annual Meeting

First Page

929

Last Page

939

This document is currently not available here.

Share

COinS