论文部分内容阅读
目的:探讨样本数据重复采集对所构建近红外(NIR)定量校正模型稳健性的影响,初步阐释该影响产生的原因。方法:以银黄液为研究载体,采集样本的近红外光谱,并以高效液相测定值为参考值,采用偏最小二乘算法建立黄芩苷定量校正模型,对潜变量因子累积贡献曲线进行深入探讨,在潜变量空间阐述重复采样对所建立的定量校正模型的影响。结果:在对重复采集光谱平均后,以最优光谱预处理方法建立的定量预测模型达到理想预测结果(RMSECV=1.824)。该模型潜变量因子累计贡献率曲线下的面积,明显大于其他光谱建模方式,即所得的模型更加稳健。结论:多次测量取平均能够显著提高模型的预测性能,使所得的模型更加稳健。
OBJECTIVE: To investigate the effect of repeated sample data acquisition on the robustness of the NIR quantitative calibration model, and to explain the causes of the effect. Methods: Near-infrared spectra of samples were collected with silver-yellow liquor as carrier. The calibration curve of baicalin was established by partial least-squares method and the cumulative contribution curve of latent variables was established Discuss the influence of the resampling on the established quantitative correction model in the latent variable space. RESULTS: The quantitative prediction model established by the optimal spectral pretreatment method achieved the ideal prediction result (RMSECV = 1.824) after the average spectra were collected repeatedly. The area under the cumulative contribution rate curve of the latent variables in this model is significantly larger than other spectral modeling methods, that is, the model obtained is more robust. Conclusion: Averaging multiple measurements can significantly improve the predictive performance of the model and make the resulting model more robust.