Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4562938 | Food Research International | 2007 | 11 Pages |
During food extrusion the degree of cooking and the quantity of expensive colorants used, both influence colour and the aesthetic appeal of the product. Thus, the use of a fiber optic equipped spectrophotometer for colour monitoring inside the extruder holds great potential for product characterization and effective quality control. However, this approach presents two difficulties: (1) inadequate sensitivity of the spectrophotometer to dark colours; and (2) the need to correlate the colour monitored inside the extruder (in-line) with that of the final product (off-line). In this paper, a method of spectrophotometer calibration, termed the “virtual white reference”, is shown to overcome the sensitivity and reproducibility problems related to dark-colour measurements. In addition, to predict the final-product colour based on in-line measurements, different artificial neural networks were tested. A multilayer feedforward network with three hidden neurons provided an acceptable correlation between the in-line and off-line colour measurements.