【作者】 白英奎； 孟宪江； 许晓杰； 欧玉奎； 常守业；

【Author】 Bai Yingkui~1 Meng Xianjiang~1 Xu Xiaojie~2 Ou Yukui~3 Chang Shouye~4 ~1(College of Communication Engineering, Jilin University ,Changchun 130025,China) ~2(College of Physics, Jilin University,Changchun 130025,China) ~3(School Of Electronical Engineering, Beijing Jiao Tong University, Beijing 100044,China) ~4(Flood Control and Draught Fighting Head Office of Jilin Province,Changchun 130025,China)

【机构】 吉林大学通信工程学院； 吉林大学物理学院； 北京交通大学电气工程学院； 长春防汛抗旱指挥部；

【摘要】 提出了一种聚类分析和偏最小二乘法(PLS)结合的新的近红外(NIR)多组分分析法。此方法可以“由粗及精”地预测组分浓度。首先利用聚类分析判别测试样本大致的浓度范围,然后利用此浓度范围附近的训练样本建立PLS校正模型,预测样本的组分浓度。和传统的PLS比较,改善了模型的适应性,显著地提高了预测精度。实验及数据处理结果证明了此方法的有效性。更多还原

【Abstract】 It presents a new NIR multi-component analysis method with clstering analysis and partial Least Square Regression(PLS). This method can predict concentration of the componets from roughness to precision. First, rough concentration range of prediction sapmles is judged. Then, PLS correlation model around this concentration range is establised, and the concentration of prediction component is predicted. The experiment and the result of data process show this method improves the model’s applicability, evidently enhances prediction precision Comparing to traditional PLS.更多还原