柿饼干燥中单宁变化规律及BP神经网络预测模型的建立  

Variation of Tannin in Persimmon Drying and Establishment of BP Neural Network Prediction Model

作　　者：陈超锋 王恒沪 黄子珍 滕建文[1] CHEN Chao-feng;WANG Heng-hu;HUANG Zi-zhen;TENG Jian-wen(Guangxi University, Nanning 530004, China)

机构地区：[1]广西大学

出　　处：《中国调味品》2019年第6期50-55,共6页China Condiment

基　　金：广西创新驱动发展专项项目(桂科AA17204038)

摘　　要：研究不同温度、不同初始单宁含量、不同水分含量对柿饼干制过程中可溶性单宁变化规律的影响,并建立BP神经网络预测模型。结果表明:在35~55℃范围内,温度越高,可溶性单宁脱涩时间越短,且每个温度下均出现返涩现象;初始单宁含量越高,脱涩时间越长,但初始单宁含量在低浓度范围内,脱涩时间不受单宁浓度的影响;水分含量影响脱涩速率,水分含量越低,脱涩越困难。通过建立的BP神经模型可知,BP网络结构为4-6-1,BP预测模型的相关系数为0.966,验证集模型的相关系数为0.93,证明BP神经网络可以对干燥过程中的可溶性单宁含量进行预测。The effects of temperature, initial tannin content and moisture content on the variation of soluble tannin in dried persimmon are studied, and a BP neural network prediction model is established. The results show that in the temperature range of 35~55 ℃,the higher the temperature, the shorter the deastringent time of soluble tannin, and there is a back-astringency phenomenon at each temperature,the higher the initial tannin content, the longer the deastringent time, but the low the initial tannin content, the deastringent time is not affected by tannin content, the water content affects the speed of deastringent, and the lower the water content, the more difficult the deastringent. The BP neural network structure is 4-6-1, the correlation coefficient of BP prediction model is 0.966, and the correlation coefficient of validation set model is 0.93 . It proves that BP neural network can predict the content of soluble tannin in drying process.

关 键 词：柿饼 干燥 可溶性单宁 BP神经网络分析 

分 类 号：TS255.3[轻工技术与工程—农产品加工及贮藏工程]