近红外光谱技术在卷烟配方设计与相似度评估中的应用综述

Review of the Application of Near-Infrared Spectroscopy Technology in Cigarette Formulation Design and Similarity Assessment

摘要: 本文系统综述了近红外光谱技术在卷烟配方设计与相似度测量中的关键方法与应用进展。在配方设计方面，重点阐述了光谱数据的预处理、特征提取、基于遗传算法等非线性优化方法的配方回归模型，以及叶组配方的解析与数字化设计平台构建。在相似度评价方面，探讨了多种算法(如余弦相似度、Jaccard系数、汉明距离)的融合应用，以及基于Fisher投影的质量判别模型。研究表明，这些技术能够有效提升配方预测的准确性、优化效率和产品质量一致性，并为卷烟叶组配方的配方构建、快速维护与稳定性控制提供了系统的技术框架与方法支持。

Abstract: This paper systematically reviews key methods and application advances in near-infrared spectroscopy technology for cigarette blend formulation and similarity measurement. Regarding formulation design, it elaborates on spectral data preprocessing, feature extraction, formulation regression models based on nonlinear optimization methods such as genetic algorithms, as well as the analysis of leaf group formulations and the construction of a digital design platform. For similarity evaluation, it explores the integrated application of various algorithms (e.g., cosine similarity, Jaccard index, Hamming distance) and quality discriminant models based on Fisher projection. Research indicates that these technologies can effectively enhance the accuracy of formulation prediction, optimization efficiency, and product quality consistency. Furthermore, they provide a systematic technical framework and methodological support for formula construction, rapid maintenance, and stability control of cigarette leaf group formulations.