Automated identification of web shrinkage and alignment in paper and film machines

This paper considers identification of the cross-directional (CD) response of the paper machine from the input-output data. In particular, the paper is devoted to identification of the mapping between the CD actuators and the scanned profiles of the paper properties taking into account effects of paper shrinkage and misalignment. We describe a novel approach to identification of alignment and shrinkage using advanced optimization technology. The approach allows us to achieve reliable identification of alignment and shrinkage within 10-20 scans of the profile data, despite the process noise. Instead of averaging only in machine direction (MD), as it is commonly done for classic bump tests to determine bump centers, we compute model fit error for the entire array of data. Such design of the algorithms effectively averages out the noise in both MD and CD which results in much better noise rejection, improved accuracy, and much shorter required test interval than classical bump test methods.