摘要
固态发酵过程的窖池温度、入窖水分、入窖酸度和入窖淀粉等因素和出酒率有很大的相关性,针对发酵周期较长,温度数据维数偏高导致其他属性作用降低的问题,引用相对高度的定义,借鉴其中的思想,提出了一种曲线特征提取算法。整个过程采用递归方法,每次提取均取曲线上离曲线两端连接线段最远的点,利用这些点将曲线分成多个曲线段,从左到右遍历这些曲线段,得到曲线段对应的相对高度,并从左到右依次存放在对应的二叉树根节点下面的子节点上,用相对高度描述温度曲线,每个温度曲线对应唯一的特征。实验用支持向量机分别对高斯拟合法预处理后的温度数据、提取出的温度曲线特征、曲线特征加入窖水分、入窖酸度和入窖淀粉进行分类,分别记为M1、M2、M3,实验结果表明M3的分类性能优于M1和M2,M3比M1平均提高了14.2225%,M3比M2平均提高了6.4027%,为下一轮入窖的工艺优化提供了依据,使发酵过程处于更理想状态,提高白酒的出酒率。
Solid-state fermentation process of pits temperature,into the pit water,into the pit acidity and cellar starch and other factors and the rate of wine have a great correlation,for a longer fermentation period,the temperature data dimension lead to lower the role of other properties The definition of relative height is referenced,the thought of it is borrowed and a curve feature extraction algorithm is proposed.The whole process uses a recursive method,each extraction are taken from the curve at both ends of the connecting line farthest point,the use of these points will curve is divided into multiple curve segments,from left to right traversing these curve segments,the curve segment corresponding to the relative height is obtained,and from left to right they are stored in the child node below the corresponding root node of the binary tree,a relative height describes the temperature curve,each temperature curve corresponds to a unique feature.Experimental support vector machine are respectively Gaussian fitting method after the pretreatment temperature data,the extracted temperature curve characteristics,the curve features added pit water,cellar acidity and cellar starch classification,which are recorded as M1,M2,M3.The results of experiment show that the classification performance of M3 is better than that of M1 is M2,M3 was 14.2225%higher than M1 and M3 is 6.4027%higher than M2,which provides the basis for the process optimization of the next round of cellar and the fermentation process is in a more ideal state in order to the liquor output rate.
作者
陈树
张继中
CHEN Shu;ZHANG Jizhong(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122)
出处
《计算机与数字工程》
2019年第8期1895-1899,共5页
Computer & Digital Engineering
关键词
曲线特征提取
支持向量机
二叉树
固态发酵
分类
curve feature extraction
support vector machine
binary tree
solid state fermentation
classification