Mathematical Filtering
Numerical algorithms apply local polynomial regression to equally spaced data points to reduce high-frequency noise in spectral measurements. Application of Savitzky Golay smoothing enhances the clarity of near-infrared spectra collected from moving paperboard sheets. The method uses a moving window to fit a least-squares polynomial to the raw sensor data.
This calculation reduces signal fluctuation without introducing a phase shift or a temporal delay.
Signal Preservation
Traditional moving average filters reduce noise but often flatten the peak heights of characteristic chemical absorption bands. By contrast, Savitzky Golay smoothing retains the higher moments of the spectral peaks, which is essential for accurate quantitative analysis of paperboard properties. The algorithm calculates the first and second derivatives of the spectra simultaneously, revealing subtle chemical features that were previously hidden by electrical noise.
This preservation of peak shape enables the precise detection of wet end additives and sizing chemicals. It allows the control system to make fine adjustments to the chemical flow without overreacting to random sensor noise.
Algorithm Output
Selecting the correct window size and polynomial degree determines the effectiveness of the noise reduction process. Excessive Savitzky Golay smoothing can obliterate actual chemical changes if the window is wider than the narrowest absorption feature. Correct parameter selection balances noise reduction against spectral resolution to ensure stable in-line measurements.