Backup roll eccentricity estimation and control

A novel recursive identification scheme is described for the estimation of the time-varying frequency of roll eccentricity signal. The method is based on the identification of frequencies, amplitudes and phases of multiple sinusoidal signals. The algorithm uses only the load measurement and does not need any extra sensors or a physical model of the stand. The scheme is recursive and computationally effective. It can therefore be implemented on-line on mills with hydraulic actuators to compensate the roll eccentricity. The virtual sensor is evaluated using the measurement obtained from the British Steel Llanwern Hot Strip Finishing Mill.