【作者】 张红梅；

【导师】 王俊；

【作者基本信息】 浙江大学 ， 农业机械化工程， 2007， 博士

【摘要】 近20年来，电子鼻研究取得了长足进展。目前国外已经有十几家公司可以生产电子鼻，这些设备大多可应用在食品工业中。检测样品时花费小、灵敏度高、重复性好是它们的主要优点，但它们价格较贵，推广使用比较困难。电子鼻及其应用研究在我国尚处于起步阶段，研究开发便携，低功耗，价格适宜的电子鼻近年来己发展成为一个热门的研究方向。本文结合农产品的香气成分复杂的特点，开发一种能依据农产品的挥发性气味对农产品进行检测的电子鼻。全文围绕电子鼻系统的研制和电子鼻在农产品检测中的应用展开研究。实现了电子鼻系统的研制，并成功地将所开发的电子鼻应用于谷物霉变、谷物虫害和水果质量的识别和定量分析中。主要结论如下：设计并实现了一个电子鼻系统。该系统由取样部分、数据采集部分、信号处理部分组成。并采用LabVIEW开发软件，结合虚拟仪器的概念，开发了电子鼻系统软件。传感器阵列由八个商业化的传感器组成，只要更换一个或几个传感器即可组成新的传感器阵列，可方便的测量其它样品。通过所研制的电子鼻系统对不同霉变程度的谷物，新鲜谷物中掺入不同百分比的霉变谷物进行检测。通过方差分析和主成分分析(PCA)对传感器阵列进行优化，去掉冗余的传感器，然后进行不同的模式识别。模式识别结果显示该系统对谷物霉变菌落总数和在新鲜谷物中掺入霉变谷物数量的检测都具有很高的分析精度。探索了水果坚实度、糖度和pH值的电子鼻无损检测方法。采用逐步回归、二次多项式逐步回归和BP网络建立大白桃坚实度、糖度和pH与电子鼻信号之间的关系模型。预测值和测试值的相关系数都很高，相应的误差也很小。对不同采摘期雪青梨采用多元线性回归、主成分回归、偏最小二乘回归的线性建模方法，建立了雪青梨坚实度、糖度和pH值的回归模型，除了pH值的回归模型，其它模型的预测精度总体来说都比较好。不同建模方法所得的数学模型预测结果有一定的差异，其中多元线性回归模型的预测结果是最好的。然后采用二次多项式回归和BP网络建立雪青梨坚实度、糖度和pH预测模型。多元线性回归、二次多项式回归和BP网络对雪青梨坚实度和糖度的预测精度都很好。利用定量分析方法建立水果质量指标与电子鼻信号之间关系的数学模型，并用于水果质量的预测是可行的。电子鼻系统对稻谷、小麦和玉米的生虫情况以及在完好的谷物中掺入不同比例的生虫谷物的样本进行检测。主成分分析(PCA)和线性判别分析(LDA)可以把受虫害侵蚀不同时间的稻谷、小麦和玉米根据受侵害的时间进行很好的分类。分别利用多元线性逐步回归、二次多项式逐步回归和BP网络建立传感器信号和稻谷生虫时间、小麦生虫时间之间的模型，结果显示三种模型对稻谷生虫时间和小麦生虫时间的预测精度都很高，标准预测误差也都很小，所以三种模型都适合利用生虫稻谷的电子鼻信号和生虫小麦的电子鼻信号建立模型对稻谷生虫时间和小麦生虫时间进行预测。对玉米生虫时间的预测结果显示二次多项式逐步回归和BP网络对玉米生虫时间的预测精度都很高，标准预测误差也都很小，所以二次多项式逐步回归和BP网络都适合利用生虫玉米的电子鼻信号建立模型对玉米生虫时间进行预测。更多还原

【Abstract】 In the recent 20 years, the researches of electronic nose have made big progress. There are several companies to be allowed to produce the electronic nose overseas. The device has the advantage of high portability for measurements with lower costs and good reliability. They are particularly useful for the analysis of headspace of liquid or solid samples. Recently, electronic noses have been extensively used to test the food quality. But, the device is difficult to popularize in Chinese which of developing country due to the price is expensive. The research and application of electronic nose at the start stage in our country. So, the research and development of new device with portable, low power and low cost has become a popular research direction in recent years. Development of electronic nose is able to recognize the agricultural product based on the complex fragrance ingredient of agricultural product. The development of electronic nose and its applications to classify of agricultural product were investigated in this paper. An electronic nose was developed and applied to classify and quantitative analysis of different degree of moldy grain, different degrees of insect damage grain and fruit quality by the responses of the sensors to volatiles in the samples. The main conclusion is as follows:The electronic nose system consists of sampling unit, data acquisition unit and signal processing unit. The electronic nose software system was developed by using the concept of virtual instrument in LabVIEW development platform. The sensor array is composed of eight commercial metal oxide sensors, so long as replaces or several sensors then composes the new sensor array then may facilitate detection other new samples.An electronic nose was applied to classify different degree of moldy grain and the percentage of adulteration in grain. A few of redundancy sensors were removed by multivariate analysis of variance and principal component analysis. Finally, responses signals of residual sensor were chose for the different pattern recognition. The results obtained indicated that the electronic nose could predict aerobic bacterial count of moldy grain and the percentage of adulteration in grain with a high accuracy.The nondestructive measurement used to fruit firmness, sugar content and acidity was studied, and the fruit interior qualities of mathematic models were developed. The relationship between sensor signals and the firmness, sugar content and acidity of "dabai" peach were developed using multiple linear regressions with stepwise procedure, quadratic polynomial step regression and BP network. The results exhibits a very good ability in describing the quality indices of the selected three set of peach in training and prediction, as witnessed by the high correlation coefficients and the relatively low average percent error. The multivariate calibration methods, multiple linear regression (MLR), principal component regression (PCR) and partial least-squares regressions (PLS) were applied to predict the quality indices of"xueqing" pear from different picking dates based on the signal of electronic nose. All models for firmness and soluble solids content show a good prediction performance. However the acidity, there was a very poor correlation with the signal of the electronic nose. It was found that MLR led to more precise predictions than the other multivariate calibration methods. The relationship between sensor signals and the quality indices of pear were developed using quadratic polynomial step regression and BP network also. It was found that the forecast result of multiple linear regressions with stepwise procedure, quadratic polynomial step regression and BP for fruit firmness, sugar content and acidity were very precise. The results indicate that it is possible to use this electronic nose technique for measuring fruit quality characteristics.Electronic nose was applied to classify different degrees of insect damage grain and the percentage of adulteration in grain. Principal-component analysis (PCA) and linear-discriminant analysis (LDA) were applied to the generated patterns to discriminate successfully different degrees of insect damage grain. Multiple linear regressions with stepwise procedure, quadratic polynomial step regression and BP network were applied to predict the time of insect damage for grain and the percentage of adulteration in grain. It was found that quadratic polynomial step regression and BP network led to more precise predictions for the time of insect damage than for the percentage of adulteration in grain. Generally, the fresh rice and insect damage rice adulteration in fresh rice, fresh wheat and insect damage wheat adulteration in fresh wheat, flesh maize and insect damage maize adulteration in fresh maize were able to differentiate clearly. The result showed that forecast precision to the time of insect damage from three kind of models were very high, the standard error prediction is also small. Therefore three kind of models all suit use electronic nose signal of the insect damage grain to establishment model for prediction the time of insect damage.更多还原