W. Yu, D.I. Wilson, and B.R. Young (New Zealand)
Control performance assessment, minimum variance con trol, valve stiction, spline, Gaussianity and linearity test.
The field of control performance assessment (CPA) has re ceived considerable attention in the engineering research literature for linear systems. However, CPA for nonlin ear systems is far less mature. This paper addresses the problem of assessing the quality of the control performance when the control valve suffers from a particular nonlinear phenomena known as valve stiction. The strategy assumes that the output can be decomposed into two parts: a nonlin ear component introduced by the stiction, and a linear re mainder. A smoothing spline curve is used to fit the nonlin ear part where the determination of the extent of smoothing depends on a Gaussianity and linearity test on the residuals between the output and the spline curve. Once the spline curve is identified, statistics such as the minimum variance performance bound can be estimated for the residuals us ing standard time series identification techniques. The re sults of a simulation example illustrate that the proposed methodology is quite efficient and accurate enough to pro vide the statistics for CPA for linear systems with nonlin earity introduced by valve stiction.
Important Links:
Go Back
