Multichannel Optimal Linear Deconvolution Filters and Strip Thickness Estimation From Gauge Measurements

Abstract

A solution is presented to the H2 optimal deconvolution filtering, smoothing and prediction problems for multivariable, discrete, linear signal processing problems. A weighted H2 cost-function is minimized where the dynamic weighting function can be chosen for robustness improvement. The signal and noise sources can be correlated and signal channel dynamics can be included in the system model. The estimation of the thickness of steel strip given X-ray gauge measurements is then considered. The deconvolution problem arises because the thickness at the roll gap is required for control purposes whereas the measurement occurs some time later when the strip reaches the X-ray gauge.