Optimized selection of modeling samples in milk component concentration ultrasonic detection

Contraposing to the three primary components of milk such as fat, protein and lactose, in the determinate range, it is the first time to select the modeling samples with the orthogonal principle of orthogonal table, and build the prediction model with the alternating validation of PLS(Partial Least Square). Furthermore, contrasting the prediction results between the two fat PLS models deriving from the orthogonal model and the conventional model respectively. The results show that the prediction error difference between the two fat PLS models is less than 0.02 g/100 g. For these two models, the discrepancy between the predicted concentration and the reference concentration is about 0.1 g/100 g. However, the sample size of the conventional model is about six times than that of the orthogonal model. Furthermore, the prediction error of the third model which derives from the samples being extracted from the conventional model and being equal to the orthogonal model samples is larger than that of the former two models, and its correlation coefficient is smaller than that of the former two models. In addition, the difference between the predicted concentration from the third model and reference concentration is about 0.4 g/100 g.