Estimation in paper machine control

The problem of online estimation of basis weight and moisture content in paper machines is discussed, and algorithms for separating cross machine and machine direction (MD) variations using scanned data are proposed. Because of its inherent nonlinearity, the moisture scheme uses a bootstrap algorithm, assuming known MD dynamics. For basis weight, the model linearity can be used to develop an extended Kalman filter to estimate the more complicated MD dynamics. Both algorithms have been tested on industrial data. Results from the basis-weight algorithm when applied to industrial scanned and stationary data collected from an operating paper machine show that a second order autoregressive moving average (ARMA) model gives the best fit to the data in terms of sum of squares of prediction errors and in terms of the whiteness of the residual. Furthermore, there is a very good agreement between the MD models estimated with the scanned data and the single point data.<>