Design, Construction, and Testing of an Automated NIR In-line Analysis System for Potatoes. Part I: Off-line NIR Feasibility Study for the Characterization of Potato Composition

An off-line near-infrared reflectance (NIR) feasibility study was conducted to explore the critical steps in the NIR determination of the major potato constituents (dry matter, starch, and protein) in relatively large (10 kg) potato samples. The results were important for the design of an automated industrial analysis system for potatoes with in-line NIR. The 10-kg potato samples were pulped with an industrial rotary saw blade rasp. A critical step in the NIR measurements was the occurrence of phase separation in the potato pulp. Phase separation manifests itself directly after pulping the potatoes and significantly affects the NIR spectrum. Therefore, during the NIR measurements, the potato pulp had to be stirred continuously. The NIR spectra (1,100-2,500 nm) were measured by applying an optical fiber NIR probe (EDAPT-1) connected to the NIR spectrophotometer (Technicon Infralyzer IA 500). NIR models for the concentration of dry matter, starch, and coagulating protein in potatoes have been developed. With the partial least squares regression procedure, promising NIR models were calculated. The NIR models were validated using an independent validation set of potato samples. The root mean square error in prediction of the samples in the validation set was 0.5% (w/w) for dry matter, 0.63 (w/w) for starch concentration, and 0.06% (w/w) for the coagulating protein.