乳制品成分快速检测实验

Experimental study on rapid detection of dairy ingredients

针对于乳制品的各项成分,通过建立液态乳制品中的非脂乳固体、脂肪、乳糖、蛋白质、密度模型进行分析研究,并且运算法则使用的是以高斯-牛顿法则为基础的偏最小二乘法(PLS)进行非线性回归计算,研究了液体乳制品中各成分的快速检测。模型预测值与真值相关系数分别为0.9912,0.9921,0.9923,0.9935,0.9929,RMSEP分别为0.095 g/100 mL,0.088 g/100 mL,0.113 g/100 mL,0.100 g/100 mL,0.626 kg/m3。通过掺水牛奶实验,验证模型通用性。结果表明:仪器稳定速度较快,在41 min后可达到稳定,仪器极差为4.1356,衰减标准差为0.5057,变异系数为0.057%,仪器测量结果可达到牛奶测量要求。

This paper uses partial least squares Gauss Newton algorithms based on regression algorithm （PLS）, a non fat milk solids, fat, lactose, protein, density model of liquid dairy products, the rapid detection of components in liquid dairy products. The correlation coefficients between the predicted values and the true values were 0.9912, 0.9921, 0.9923, 0.9935, 0.9929, RMSEP, 0.095 g/100 mL,0.088 g/100 mL, 0.113 g/100 mL, 0.100 g/100 mL, 0.626 kg/m3, respectively. The diluted milk experiment, verify the model universal test shows that the instrument is stable and fast, in 41 minutes can reach the stable range of 4.1356 instruments, the attenuation of the standard deviation is 0.5057, coefficient of variation is 0.057%, the measurement can meet the measurement requirements of milk.