尝试采用电子舌技术对橙汁感官品质进行快速定量评价。试验以3类20种品牌橙汁为研究对象，以人工感官评价结合模糊数学评价橙汁感官品质，获得各个感官指标得分值；同时采集样本的电子舌传感器数据。利用因子分析法确定橙汁各感官指标的权重，根据权重得出橙汁感官品质的总得分。然后对比采用偏最小二乘法和BP神经网络建立电子舌传感器响应值与感官品质总得分值之间的定量预测模型。结果显示，因子分析法可以有效分析不同类型橙汁的感官指标，得到色泽、香气、酸度、甜度、苦涩味、体态的权重分别为0.15、0.06、0.20、0.24、0.15、0.20。当采用主成分数为3，建立的BP神经网络模型效果最优。模型预测集中预测值与参考值的相关系数为0.93；预测集均方根误差为0.20。研究结果可为橙汁感官品质的智能化评价提供参考。

The sensory quality of three kinds of commercial available juice samples (20 brands) was evaluated by sensory evaluation combined with fuzzy mathematics. The samples were also analyzed by using the electronic tongue to obtain the corresponding electronic tongue data. The weight of each sensory indicator of the orange juice was determined using factor analysis method. The total score of the sensory quality was derived based on the weight of each sensory indicator. Subsequently, partial least squares (PLS) and back propagation neural network(BPNN) methods were contradistinctively used to establish the quantitative prediction model between sensor signals and the total score of sensory quality. The results showed that, sensory indicators of different types of orange juice could be effectively analyzed through factor analysis which accurately reflected the quality of the orange juice by the sensory evaluation results. The weights of color, aroma, acidity, sweetness, bitterness, body were 0.15, 0.06, 0.20, 0.24, 0.15 and 0.20, respectively. When the number of principal component was 3, the performance of BPNN model was better than PLS. The correlation coefficient (Rp) between the value predicted by the BPNN model and the reference value in the prediction set was 0.93 and the root mean square error of prediction (RMSEP) was 0.20. The results could provide a reference for intelligently evaluation of the sensory quality of orange juice.

江苏省高校优势学科建设工程资助项目