SST法和混合建模法对近红外定量模型维护的比较
Comparison of spectral space transfer method and hybrid modeling method in NIR quantitative model maintenance
-
摘要:
为比较SST模型转移算法和混合建模法在近红外定量模型维护中的差别，对两种模型维护方式的维护过程（参数选择、样品数选择）、维护结果（对烟碱CV值的影响）进行了考察。结果表明：①SST法需要优化参数，混合建模法不需要优化参数。②SST法需要4个维护样品即可使RMSEP趋于稳定，混合建模法需要至少12个维护样品RMSEP才趋于稳定。③经过两种方法维护后，烟碱CV值均更接近于化检值计算的CV值。此外，达到同样的维护效果时，SST法比混合建模法能减少模型维护人员的工作量。 Abstract:In order to compare the difference of spectral space transfer （SST） method and hybrid modeling method in NIR quantitative model maintenance, the maintenance process （including parameter selection and sample number determination） and maintenance result （the influences on the CV value of nicotine） of the two methods were investigated. The results showed that: 1） The parameters of SST method needed to be optimized, while those of hybrid modeling method did not. 2） SST method could achieve satisfactory predictions by only using 4 standardization samples, however hybrid modeling method needed at least 12 standardization samples to achieve the same results. 3） After maintaining the model with either method, the CV values of nicotine calculated through model prediction results were closer to the CV values of nicotine calculated through FIA （Flow Injection Analysis） results. To achieve the same prediction result, the workload of model maintainer adopting SST method was less than that of hybrid modeling method.
下载: