Controlling sugar quality on the basis of fluorescence fingerprints using robust calibration

The aim of our study was to highlight the benefits of robust calibration in the context of process control. Two properties were monitored — the color and ash content of sugar samples. It was shown for the data being studied that robust models, constructed using the partial robust M-regression technique, have a better fit to the majority of the data and prediction properties than the classic partial least squares and N-way partial least squares models. In particular, the constructed calibration models were characterized by a root mean square errors improved by 1.60% and 1.82% and a root mean square errors of prediction (for independent test samples) improved by 2.39% and 1.11% compared to classic partial least squares models constructed for color and ash content, respectively.

► Fluorescence spectra were used for the purpose of process control.
► Classic calibration models were constructed and compared with the robust models.
► Robust models outperform classic ones when process data contained outlying samples.