Orthogonal Signal Correction and Its Application to Soft Sensing

A preprocessing method for multivariate data in chemometrics, called orthogonal signal correction(OSC), is introduced. An improved OSC method is proposed and an application of this method combined with partial least squares(PLS) method to soft sensing of polypropylene melt flow rate(MFR) is demonstrated. It results in much reduction of model complexity with the same prediction ability. Besides, the regression coefficients of the model have clearer physical meaning.