|
ANALYSIS OF PLANT-WIDE DISTURBANCES THROUGH DATA-DRIVEN TECHNIQUES AND PROCESS UNDERSTANDING
Nina F. Thornhill+, Chunming Xia*, John Howell*,
John Coxx and Michael Paulonisx
+ Department of Electronic and Electrical Engineering, Torrington Place, University
College London, London WC1E 7JE, UK (corresponding author).
* Department of Mechanical Engineering, University of Glasgow, Glasgow, Scotland.
x Eastman Chemical Company, Kingsport, TN, USA.
Plant-wide disturbances can have an impact on product quality and running costs. Thus there is a motivation for automated detection of a plant-wide disturbance and for diagnosis of the root cause. In this article, data-driven techniques are used to analyze plant-wide disturbances caused, for instance, by limit cycle oscillation in a control loop. The control loops participating in the disturbance are detected and displayed on a process schematic. Other numerical signatures derived from the data trends are utilized for the diagnosis of the root cause. The outcome is a visual display that integrates process understanding and data-driven analysis.
Keywords: Chemical industry; control loop performance; diagnosis; non-linearity; plant-wide disturbance; power spectrum; process control; self validation; surrogate data
Session slot T-Mo-A11: New Directions in Control System Performance Monitoring/Area code 7a : Chemical Process Control
 | 