MEASURING FAT CONTENT OF GROUND BEEF STREAM USING ON–LINE VISIBLE/NIR SPECTROSCOPY

Thirty blocks of frozen beef from three truckloads each of nominal 50% and 15% fat meat were studied. The 27-kg blocks were ground and formed into a continuous stream on a conveyor using a custom molding head attached to the grinder. A Perten diode-array type NIR/visible spectrometer was used to take reflectance measurements from the stream, which moved at 1.0 m/min. Spectral measurements were made each 3.3 cm, and samples were collected from the stream such that each sample was represented by two successive measurements. Fat content of each sample was determined by ether extraction, and a prediction model was developed relating fat content to spectral measurements. Standard error of prediction (SEP) values for the calibration and validation data sets were 1.00% to 1.68% and 2.15% to 2.28% fat, respectively. Predictions for the fat content of an entire block gave SEPs of 0.70% to 1.05% fat. The results show that the molding head is effective, that the Perten machine is suitable for on-line measurements, and that the accuracy of predictions is high.