Data mining and analysis for characterizing paper from online multisensor measurements
The objective of this thesis is to develop a multi-resolution tool for screening paper formation variations, aiming to detect abnormalities in various frequency regions ranging from millimeters to several meters. The abnormalities detected in different frequency regions give an indication for the paper maker about specific disturbances in the paper production process.
A paper web, running at a speed of 30 m/s, is illuminated by two red diode lasers and the reflected light are recorded as two time series of high resolution measurements constitutes the input signal to the papermaking process monitoring system. The time series are divided into blocks and each block is analyzed separately. The task is treated as a kernel based novelty detection applied to a multi-resolution time series representation obtained from the frequency bands of the Fourier power spectra of the blocks. The frequency content of each frequency region is characterized by a feature vector, which is transformed using the canonical correlation analysis and then categorized into the inlier or outlier class by the novelty detector. The ratio of outlying data points, significantly exceeding the predetermined value, indicates abnormalities in the paper formation.
The experimental investigations performed have shown that the presented paper formation deficiencies monitoring technique and the system can be used for on-line monitoring of paper deficiencies manifesting themselves in a broad frequency range. A software, implementing the technique, was developed and used for online paper formation monitoring at a Swedish paper mill.
Kernel Based Novelty Detection