Quality parameter assessment of grated Parmigiano–Reggiano cheese by waveguide spectroscopy

Parmigiano–Reggiano (P–R) cheese is still one of the most valuable Protected Designation of Origin (PDO) cheeses of Italy. The rind percentage (from 12% to 50%), the ripening (from 13 to 31 months), the moisture content and the differences between true P–R and competitors were determined by means waveguide spectroscopy. Preliminary tests were carried out in the 2–20 GHz frequency range (with a span of 1 GHz) to investigate which 1 GHz frequency range contains most information on the rind percentage and on the months of ripening. Partial Least Squares (PLS) regression was used to predict rind percentage, months of ripening and moisture in the previously selected frequency ranges (2–3, 5–6 and 17–18 GHz for the rind percentage; 2–3, 5–6 and 16–17 GHz for the months of ripening). Moreover, Soft Independent Modelling of Class Analogy (SIMCA) analysis was used to discriminate the samples according to the rind percentage. Principal component analysis (PCA) was used to discriminate true P–R cheese from competitors. PLS models (test set validation) showed R^2 values up to 0.944 (root mean square error of prediction in test set validation, RMSEp = 3.4%), 0.966 (RMSEp = 1.2 months) and 0.786 (RMSEp = 0.99%) for the prediction of rind percentage, months of ripening and moisture percentage, respectively. As a result only a part of the prediction of rind percentage and ripening can easily be attributed to the moisture. For each considered frequency range, all samples belonging to the classes characterized by 12% and 50% of rind were correctly classified. Competitors were clearly separated from P–R cheese by the PCA analysis of the spectral signals and their moisture was predicted with a R^2 values up to 0.942 (test set validation). In this case, the discrimination power can be mainly attributed to moisture content.