基于近红外光谱的<b>SIMCA</b>算法在烟叶模块配方打叶中的应用

基于近红外光谱的SIMCA算法在烟叶模块配方打叶中的应用

摘要: 为快速、无损地对模块配方打叶的初烤烟叶进行预分类，以2011—2014年产于四川凉山和攀枝花地区的172个烟叶样品的近红外光谱为基础，采用SIMCA（Softing Dependent Modeling of Class Analogy）算法建立了香气、烟气、口感、拓展、中上部、下部模块的初烤烟叶预分类模型。经验证，样品的预测归属与实际配方模块基本相符，正确率达到72.2%，对初烤烟叶模块配方打叶的预分类具有较好效果，减少了配方人员工作量。

Abstract: To rapidly and nondestructively presort the cured tobacco leaves of blended modules for threshing, 172 samples of tobacco leaves harvested in 2011—2014 were collected from Liangshan and Panzhihua in Sichuan and their NIR （near-infrared spectroscopy） data were recorded. The presorting models for the flavor, smoke, taste, expansion, middle-upper, and lower modules of flue-cured tobacco leaves were established by adopting SIMCA （Softing Dependent Modeling of Class Analogy） algorithm based on the NIR data. The results of verification showed that the accuracy rate between the predicted attribution and the actual entered blended module of the samples reached 72.2%, which suggested that the models could be used to presort cured tobacco leaves of blended modules for threshing and reduce the blending workload of workers.