基于电子鼻技术的玉米气味品质检测研究

Different aroma quality detection of corn based on electronic nose

为探索玉米品质的快速检测方法,利用由10个气敏传感器组成阵列的电子鼻系统对6个品质不同的玉米挥发性气味进行了检测分析,并将10个传感器对不同品质玉米的响应进行了方差分析。结果表明,10个传感器对品质不同的玉米响应差异显著,多重比较显示存在3个冗余传感器。去掉3个冗余传感器后对电子鼻检测信号进行主成分分析,结果显示6个品质不同的玉米能被很好的区分。采用BP神经网络建立传感器信号和玉米菌落总数之间的预测模型。通过测试集对BP网络模型进行验证得到菌落总数的预测值和测试值的相关系数为0.93,预测平均相对误差为2.44%、最大相对误差为15.82%。

An electronic nose comprising ten metal oxide semiconductor was used to measure corn fla- vor in corns of six different qualities. A multifactor variance analysis was applied to analyzed the signif- icant difference existing among the gas sensors response values of corns of 6 different qualities. The re- sult of variance analysis shows that the difference among ten gas sensors and six different quality corns are remarkable. The result of multiple comparisons showed that there are 3 redundant sensors. The principal component analysis was applied to the signal of remaining seven sensors, and the six different quality corns were discriminated well. The prediction models were established between signal of elec- tronic nose and the aerobic bacterial count of corn by BP network. The BP network with test data had O. 93 correlation coefficient between predicted and measured values, with an average relative error of 2.44% , and the max relative error of 15.82%.