In the experiments i) whole potatoes, ii) whole peeled potatoes, iii) raw potato chips, iv) blanched chips, v) dried chips, and vi) fried chips (10 batches fried at 10 different time intervals, 1, 2, … 10 minutes/batch) were measured. The water content varied from 74%-81% for i) to v), whereas the water and fat content of the fried chips varied from 44%-65% and 5%-12%, respectively.
 
Three different NIR spectroscopic measurement methods were used and compared in this study:

• On-line NIR reflectance (MM710, Infrared Engineering)
• A benchtop lab-system (Zeiss), equipped with a contact transflectance fiber optic probe
• The transflection scanner originally developed by SINTEF and Nofima for measuring water content of split cod (This scanner has since been commercialised by Qvision, a subsidiary of Titech, and is now used in many food applications).

Drying and weighing was used as a reference method for determining the water content, whereas chemical analysis was used for referencing the fat content. Regression models were used to estimate the performance with an RMSEP (root mean square error of prediction) for water and fat content of 1.7% and 0.9%, respectively. The best results were obtained for blanched and dried chips, with an RMSEP of 0.6%.

It should be noted that this was a preliminary study and the sample set was limited. We expect that both the experimental design and the models can be considerably improved. In spite of this, the results convincingly show that

1. Water and fat content of fried potato chips can be measured on-line using NIR methods, and
2. the transflection scanner provided best results in most cases. 

The latter is good news, as the latest QVision scanner can easily be installed in an online industrial measurement situation.