论文部分内容阅读
建立了1种基于独立成分分析的多元校正方法,结合近红外光谱分析技术进行多组分同时测定。以甲苯、氯苯和正庚烷构成的混合体系为例进行建模和预测,并将预测结果与主成分回归所得的结果进行了比较。结果主成分回归方法对独立测试集32例样本中甲苯、氯苯和正庚烷浓度的预测平均相对误差分别为4.5030、6.0231及9.6042,而独立成分回归方法对三组分浓度的预测平均相对误差分别为4.3871、4.1465及5.8104,说明本方法具有很强的预测能力,可作为1种实用的多元校正方法。
A multivariate calibration method based on independent component analysis was established. Simultaneous determination of multiple components by near infrared spectroscopy was carried out. A mixed system consisting of toluene, chlorobenzene and n-heptane was used as an example to model and forecast the results. The predicted results were compared with the results obtained from the principal component regression. Results The principal components regression method was used to predict the average relative errors of toluene, chlorobenzene and n-heptane concentrations in 32 samples from independent test set to 4.5030, 6.0231 and 9.6042, respectively. However, the average relative error of three components 4.3871, 4.1465 and 5.8104, indicating that the method has strong predictive ability, which can be used as a practical multivariate calibration method.