It has been shown that fluorescence spectroscopy of sugar in aqueous solution carries important quality and process information related to beet sugar factories, which is accessible by multivariate analyses. A method for measuring crystalline sugar directly on-line in the process should be advantageous. In this paper we compare the solution measurement technique with two methods of fluorescence measurement on solid sugar. Surprisingly, it was possible to measure fluorescence through the sugar crystals by using the same transmission techniques with 90° detection as with the sugar solutions. This method was compared with a 45° front-surface reflection method. Sugar samples from six different sugar factories were examined. The spectral responses were reasonable, but they were influenced by the heterogeneous sample composition and the sample geometry. It was possible with the two methods to separate sugar samples according to factory with the use of principal component analysis (PCA). Seasonal time trends were found in weekly samples from the same factory. Partial least-squares regression (PLS) was used to predict quality parameters, where color (range: 6–41), ash (range: 0.003–0.018), and α-amino-N (range: 0.28–5.07) could be modeled with errors of 2.3–2.6, 0.0015–0.0016, and 0.40–0.42, respectively. Model errors for similar solution data have been determined to 2.4, 0.0012, and 0.266, respectively.
