本平台为互联网非涉密平台,严禁处理、传输国家秘密或工作秘密

基于稳定一致波长的烟叶总植物碱近红外光谱定量模型传递

Quantitative model transfer between different instruments for total alkaloids in tobacco by near-infrared spectroscopy with consistent and stable wavelengths

  • 摘要: 为提升近红外光谱校准模型的稳健性和可共享性,通过比较仪器间差谱的标准偏差(SDDSI)和精密度测试光谱的标准偏差(SDPDS),筛选出仪器间具有稳定一致光谱信号的波长(SWCSS),考察了SWCSS-PLS模型在4台仪器间的传递效果,并与全光谱模型(WW-PLS)、分段直接标准化算法(PDS)校正后全光谱模型的传递结果进行了对比。结果表明,SWCSS-PLS模型传递到4台从机上后,78个外部验证样本的平均相对误差(MRE)为5.22%~5.60%,满足低于6%的企业内控要求,而WW-PLS模型和PDS校正后的WW-PLS模型的MRE分别为9.01%~11.68%和5.43%~8.69%。SWCSS-PLS模型较全光谱模型及PDS校正模型具有更好的稳健性和适应性。

     

    Abstract: In order to improve the robustness and shareability of different calibration models in near-infrared spectroscopy, the screening wavelengths with consistent and stable signals (SWCSS) between the instruments was applied to select wavelengths by comparing standard deviance of difference spectra between instruments (SDDSI) with standard deviance of precision detection spectra (SDPDS). The transfer results of SWCSS-PLS model between four instruments were compared with those of whole wavelength model (WW-PLS) and those after piecewise direct standardization (PDS). The results showed that, the average mean error (MRE) of SWCSS-PLS model for the samples tested on the four slaves were 5.22%-5.60%, which met the internal control requirement of less than 6%. While the MREs of WW-PLS and PDS-corrected WW-PLS model were 9.01%-11.68% and 5.43%-8.69% respectively. The SWCSS-PLS model was therefore more robust and shareable than the WW-PLS model and PDS-corrected WW-PLS model. The improved consistency and stability of the spectral signals between different instruments can be the basis for the robustness and shareability of SWCSS-PLS model.

     

/

返回文章
返回