PCA-BPNN在黄山毛峰茶贮藏时间检测中的应用  

Application of PCA-BPNN in Storage Time Detection of Huangshan Maofeng Tea

作　　者：杨春兰[1] 薛大为[1] 

机构地区：[1]蚌埠学院电子与电气工程系,安徽蚌埠233030

出　　处：《广东石油化工学院学报》2016年第3期34-37,共4页Journal of Guangdong University of Petrochemical Technology

基　　金：安徽省高等学校省级自然科学研究项目(KJ2013Z195);安徽省高等学校优秀青年人才基金项目(2012SQRL218)

摘　　要：为了探索黄山毛峰茶贮藏时间的有效检测方法,利用电子鼻技术对4个不同贮藏时间下的3个等级黄山毛峰干茶叶进行了检测。根据电子鼻传感器阵列响应曲线变化特点,选取了特征变量,并通过主成分分析法提取出4个主特征变量。以主特征变量作为BP神经网络的输入,以茶叶贮藏时间作为输出,建立了黄山毛峰茶贮藏时间的PCA-BPNN预测模型。实验结果表明:PCA-BPNN预测模型的最大预测误差为38.5 d,预测误差超过10 d的样本最大比率为4.44%;与BPNN预测模型相比,PCA-BPNN预测模型性能更优。In order to seek effective detection method for storage time of Huangshan Maofeng tea,3 grades of Huangshan Maofeng tea under 4 different storage time are detected by electronic nose. The feature variables are selected according to the response curves of sensor array. Then,4 principal feature variables are extracted by PCA. PCA-BPNN prediction model for Huangshan Maofeng tea storage time is built by using principal feature variables as input variables. The experimental results show that the maximum prediction error of PCA-BPNN prediction model is 38. 5 days,the maximum proportion of the samples with the prediction errors exceeding 10 days is 4. 44%,and the performance of PCA-BPNN prediction model is better than that of BPNN prediction model.

关 键 词：PCA BPNN 黄山毛峰茶 电子鼻 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]