OSCILLATION DETECTION IN MULTIVARIABLE SYSTEMS USING PRINCIPAL COMPONENT ANALYSIS

Oscillations constitute a common form of plant-wide disturbances. The presence of
oscillations in control loops results in increased process variability which is undesirable
from quality point of view. In this paper, a simple visualization tool is developed to detect
oscillations using Principal Component Analysis (PCA). The advantage of PCA in
dimensionality reduction enables clear visualization of the process performance while
taking into account all variables in the loop. Simulation results using an industrial
Fluidized Catalytic Cracking Unit (FCCU) show successful implementation of the PCA
method in detecting oscillations in the loop due to control valve nonlinearity.