基于电子鼻的“贵妃”芒果糖度酸度无损伤检测技术应用

Non-destructive Test on Predicting Sugar Content and Acidity of Mango by Electronic Nose Technology

以"贵妃"芒果为试材,利用电子鼻检测果实气味响应值,同时测定果实的糖酸度,采用偏最小二乘法(PLS)和BP神经网络建立了基于电子鼻的可溶性固形物、可滴定酸的品质预测模型。两种方法构建的可溶性固形物含量预测模型的建模集相关系数R均大于93%,可滴定酸测模型的建模集相关系数R均大于91%。其中,BP神经网络建模集的相关系数R均略高于PLS,建模均方均根误差(RMSEM)也较低。而预测集相关系数R和预测均方根误差(RMSEP)与PLS的相当或略低,BP神经网络模型对芒果糖酸度预测准确性略好于PLS。结果表明,PLS和BP神经网络模型的预测性能均较好,利用电子鼻技术对芒果品质进行无损伤检测是可行的。

In this study, the odor response value of ＇Guifei＇ mango fruit was detected using an electronic nose （model PEN3）, meanwhile the soluble solids content （SSC） and titratable acidity （TA） were measured by traditional assays. Based on the data of odor response value, SSC and TA obtained by tests, the quality prediction models of SSC and TA by partial least squares （PI.S）and back propagation neural network （BPNN） modeling were established, respectively. The results showed that the correlation coefficient R for SSC prediction model structured by both PLS and BPNN was higher than 93%, while model correlation coefficient R for TA prediction model was higher than 91% by both PLS and BPNN. Comparatively, the correlation coefficients R by BPNN were slightly higher than those by PLS, and the root mean square error of model （RMSEM） by BPNN was louver than that by PLS. In addition, the correlation coefficient R of prediction set and root mean square error of prediction （RMSEP） by BPNN were slightly less than or similar to those by PLC, suggesting that the prediction accuracy by BPNN model for sugar and acidity in mango fruit was slightly better than that by PLS. The present findings indicate that non-destructive detection by electronic nose in combination with BPNN and PLS modeling for predicting SSC and TA of mango is a feasible and promising approach.