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

基于时序聚类的国产雪茄烟叶堆垛发酵温度变化特征分析

Analysis of stack core temperature in the fermentation of cigar tobacco leaves based on time series clustering

  • 摘要: 为解决国产雪茄烟叶堆垛发酵过程调控依赖人工经验、缺乏量化指标的问题,对烟叶堆垛垛芯温度的时序变化规律进行分析。以45个国产雪茄烟叶堆垛为研究对象,监测其垛芯温度序列,并采用动态时间规整算法(DTW)度量温度序列间的相似性,结合K-means++聚类算法进行温度时序聚类分析。通过分析各温度变化模式的动态特征,构建了融合温度变化模式、滑动24 h温差、温度阈值及发酵时长的多因素决策框架。结果表明:①在烟叶堆垛发酵过程中,垛芯温度呈现出5种不同特征的温度变化模式,分别为低温稳定型、平稳升温型、快速升温型、高温稳定型和中温波动型。②对于低温稳定型模式,发酵时长超过120 h且垛芯温度低于30 ℃时建议翻垛;对于平稳升温型模式,发酵时长超过168 h且垛芯温度低于35 ℃时建议翻垛;对于快速升温或高温稳定型模式,垛芯温度超过45 ℃或滑动24 h温差≤0 ℃时,触发翻垛预警;对于中温波动型模式,在全周期累计发酵时长超过30 d、当前阶段发酵时长超168 h及滑动24 h温差绝对值≤0时,可判定发酵结束并出库。③采用模型决策框架方法辅助翻垛能够提前预判发酵终点,减少各发酵阶段末期的恒温或降温时间,提高发酵效率。该研究为烟叶堆垛翻垛与出库操作提供了精准量化的决策依据,并为雪茄烟叶发酵过程的数字化调控与标准化生产提供技术支撑。

     

    Abstract: To address the reliance on empirical judgment and the lack of quantitative indicators in controlling the fermentation of Chinese cigar tobacco piles, this study analyzed the temporal variation patterns of pile core temperatures. Focusing on 45 piles of domestic cigar filler tobacco, core temperature sequences were monitored. Dynamic Time Warping (DTW) was employed to measure sequence similarity, followed by time-series clustering using the K-means++ algorithm. By elucidating the dynamic characteristics of each identified pattern, a multi-factor decision framework was constructed, integrating temperature patterns, 24-hour sliding temperature differences, temperature thresholds, and fermentation duration. The results indicate that: 1) Five typical temperature variation patterns were identified during fermentation: low-temperature stability, steady warming, rapid warming, high-temperature stability, and medium-temperature fluctuation. 2) Quantitative decision rules were established for each pattern: for the low-temperature stability pattern, pile turning is recommended when fermentation exceeds 120 h and temperature remains below 30 °C; for the steady warming pattern, turning is suggested when fermentation exceeds 168 h and temperature is below 35 °C; for rapid warming or high-temperature stability patterns, a turning alert is triggered when temperature exceeds 45 °C or the 24-hour sliding temperature difference is ≤ 0 °C; for the medium-temperature fluctuation pattern, fermentation is deemed complete and ready for outbound transfer when the cumulative duration exceeds 30 days, the current stage duration exceeds 168 h, and the absolute value of the 24-hour sliding temperature difference is ≤ 0 . 3) Implementing this framework facilitates early prediction of the fermentation endpoint, reduces the duration of constant-temperature or cooling phases at the end of each stage, and significantly improves fermentation efficiency. This study provides precise, quantitative criteria for pile turning and outbound operations, offering robust technical support for the digital regulation and standardized production of cigar tobacco leaf fermentation.

     

/

返回文章
返回