Off-line parameter estimation for linear control valve with hysteresis

The problem of estimating parameters of linear control valve with hysteresis is considered. The hysteretic behavior of control valve, which is modelled as a switched linear model, is identified by a subspace decomposition technique. Via spectral clustering on subspace, the noisy data of control valve are classified into separate sets, each corresponding to up-trip mode and down-trip mode respectively. Then least squares technique is used to estimate the parameters of control valve. It is shown that the proposed technique exactly estimates the parameters and switching epochs in absence of noise and exhibits dominant advantage in noisy scenario.