论文部分内容阅读
建模样品水分含量的差异对近红外光谱分析模型的稳健性影响很大。以小麦蛋白质含量为研究对象,定标集的水分含量分布设为3个不同δ值的正态分布和1个均态分布,分别建立预测小麦蛋白质含量的模型。结果表明,当用4个模型预测水分范围在10%～15%,蛋白质含量平均值为14.31%的样品时,3个正态分布预测的结果分别是13.63%、13.83%、14.04%,但均态模型的预测结果是14.24%,明显比正态模型预测效果好。当预测中间水分样品的蛋白含量时,正态模型预测准确性要优于均态模型。应用中,可先用较宽水分含量范围建立的模型对待测样品水分进行粗测,后用与待测样品水分粗测值适配的窄水分含量范围的模型来测定样品的蛋白质含量,可减少由于建模水分的差异所带来的预测误差。
The differences in the moisture content of the modeling samples have a great influence on the robustness of the NIRS model. Taking the wheat protein content as the research object, the water content distribution of the calibration set was set as the normal distribution and the homogenous distribution of three different δ values, and the models for predicting the protein content of wheat were established respectively. The results showed that the prediction results of the three normal distributions were 13.63%, 13.83% and 14.04% respectively when the four models were used to predict the moisture content in the range of 10% -15% and the average protein content was 14.31% The prediction of state model is 14.24%, which is obviously better than that of normal model. When predicting the protein content of the intermediate moisture sample, the prediction accuracy of the normal model is better than that of the homogeneous model. Applications, the first model can be established with a wide range of moisture content of the sample to be measured crude water, and then with the sample to be tested for the water content of crude moisture content of the model to determine the scope of the model to determine the protein content of the sample can be reduced Prediction errors due to differences in modeling moisture.