成都大学学报（自然科学版）2026，Vol.45Issue(2)：166-173,8.DOI:10.3969/j.issn.1004-5422.2026.02.008
基于K-means聚类算法的白酒分段阈值分析
Segmentation Threshold Analysis of Liquor Based on K-means Clustering Algorithm
摘要
Abstract
The traditional sensory-based liquor fractionation process faces a core issue of unstable base liq-uor quality,which severely restricts the large-scale and standardized development of the liquor industry.This study,by taking the CJ fermentation type from a certain distillery as the research object,collects 392 batches of manual liquor fractionation production data,and combines brewing process characteristics.In-stantaneous concentration,cumulative flow,and comprehensive concentration are selected as core indica-tors,and the K-means clustering algorithm is used to perform clustering analysis on the data at various sta-ges of liquor distillation and fractionation.To address the subjectivity in selecting the k value,the study in-tegrates the Elbow Method,silhouette coefficient,and adjusted Rand index to construct a multidimensional evaluation system,scientifically determining the optimal number of clusters.The study shows that the K-means algorithm can classify the liquor collection data into six clusters,and using the cluster centroids as fractionation segmentation threshold forms a targeted liquor collection strategy.The innovation of this re-search lies in selecting differentiated clustering dimensions according to the characteristics of different liq-uor collection stages,and validating clustering effectiveness through multiple indicators.This compensates for the lack of specificity in traditional clustering methods for liquor fractionation and has significant prac-tical value for technical upgrades in the brewing industry.关键词
自动化摘酒/K-means聚类/分段阈值/基酒质量/多指标验证Key words
automated liquor fractionation/K-means clustering/segmentation threshold/base liquor quali-ty/multiple indicators validation分类
信息技术与安全科学引用本文复制引用
郭俊,周海龙,庹先国,刘勇,程帆业..基于K-means聚类算法的白酒分段阈值分析[J].成都大学学报（自然科学版）,2026,45(2):166-173,8.基金项目
泸州市科技计划重点研发项目(2024YF193) （2024YF193）
