【作者】 周亦斌；

【导师】 王俊；

【作者基本信息】 浙江大学 ， 农业机械化工程， 2005， 硕士

【摘要】 农产品品质的无损快速检测和评价在近二三十年来一直是科研工作者感兴趣的课题之一,农产品的无损检测是不破坏测定对象,在其原有状态下获得测定对象的内部品质、成分等与品质相关的物理化学性质的检测方法,并可对其内部品质(包括糖度、酸度、硬度、内部病变等)进行评价的一种新方法。迄今为止,国内外对无损检测的方法进行了不断的研究与探索。结果表明,水果的坚实度与芳香成分变化是用来衡量水果品质的一个重要指标,然而在实际中测定水果坚实度以及测定芳香成分的方法都会引起水果的破损,而且测量速度不快,不适合于在线测量。为了克服这些缺点,许多学者利用水果的光学、声学、电学等物理特性来研究无损检测水果坚实度与芳香成分的方法,但到目前为止,这些方法在商业中都还没有得到普及推广。 饮料的品质评价除了通过一些常规的理化指标检测外,大多数是通过人的感官评定的,感官审评是通过人的感觉器官来评价饮料的特征和品质,是一门应用感官分析技术的学科。人的感觉器官的灵敏度易受外界因素的干扰而改变,从而影响评定的准确性。为了在饮料生产、流通过程中有一个严格、一致的标准,采用仪器测定饮料,用科学计量上的品质指标来评价饮料品质是必要的手段之一。 本研究主要利用电子鼻技术来评价番茄的品质(成熟度、成熟过程和损伤检测)和黄酒品质(陈年数和品牌的区分识别),拟获得电子鼻信号与番茄与黄酒品质之间的相关性。为今后进一步进行农产品与饮料的无损检测技术探索了一条新的途径。 在本课题的实验中,采用了德国的Airsense Analysis GmbH公司生产的电子鼻PEN2和吸附解吸附EDU。该系统具有较好的客观性、可靠性、重现性等方面的优点。同时,它测量快速,操作方便。它是目前国内外研究无损检测手段中较先进的方法。 1 利用电子鼻分析番茄的成熟度、贮藏过程和损伤后的气味变化 不同成熟度番茄的电子鼻检测分析表明,通过PCA和LDA分析,半熟期的番茄与成熟期、完熟期的番茄可以较好地用电子鼻进行区分,成熟期和完熟期的番茄有部分区域发生重叠。按坚实度指标对番茄成熟度进行重新划分后所进行的分析可知,电子鼻可以区别番茄的坚实度差异,采用PCA方法分析时,电子鼻可以100%地区分不同成熟度的番茄。提取55S时刻的电子鼻信号进行PLS分析和预测,在PLS分析中采用完全交叉确认(Full Cross Validation)的方法。通过不同潜在变量下的剩余方差分析图,建立4个潜在变量的校准模型。在所建立的PLS模型中,预测坚实度与实际坚实度的相关系数R=0.950,平均标准误差(RMSEP)为0.129。可见电子鼻信号与坚实度之间有较好的相关性,可以采用电子鼻信号预测番茄的坚实度。 成熟期和完熟期番茄在贮藏过程中的电子鼻检测分析表明,通过PCA和LDA分析,成熟期的番茄在Dayl-Day6、Day7-Day11、Day14-Day17之间可以较好地进行区分。通过PLS分析,建立8个潜在变量的预测成熟期番茄贮藏期的模型,预测成熟期番茄贮藏期的预测值与实际值的相关系数R=0.989,平均标准误差(RMSEP)为0.666。完熟期的番茄在贮藏中电子鼻检测的PCA分析可知完熟期番茄气味变化的趋势,但Day1-Day5、Day6-Day11之间变化没有较明显的分界,Day6-Day11、Day14-Day17之间可以较好地进行区分。采用PCA分析来区分完熟番茄的贮藏时间不理想。从LDA分析可知,Day1-Day5、Day6-Day11、Day14-Day17之间可以较好地进行区分。通过PLS分析,建立建立8个潜在变量的预测完熟期番茄贮藏期的模型,预测值与实际值的相关系数R=0.980,平均标准误差(RMSEP)为0.912。 成熟期番茄损伤后在贮藏过程中的电子鼻检测分析表明,在传感器未优化之前,从LDA分析中可知,在第1、4、7天时,未损伤的4组样品、未损伤和损伤混合的4组样品存在部分重叠;在第10、13天,两者发生了分离,因此在第7天之后可以进行两者的区分。对传感器进行加载分析和对其信号的分析表明,MOS2、MOS6、MOS7、MOSg和MOS9这5个传感器对检测的样品挥发物响应信号变化较大,说明这5个传感器对样品挥发物反应敏感;MOS1、MOS3、MOS4、MOS5和MOS10这5个传感器对检测的样品挥发物反应不敏感。在多次对比分析发现,在5个不敏感的传感器中去除MOS4、MOS5和MOSlO这3个传感器后,在LDA分析中,电子鼻可以在番茄损伤后的第1天即可以对未损伤的样品、未损伤和损伤相混的样品进行检测区分。在以后番茄损伤的检测中,可以采用MOS1、MOS2、MOS3、MOS6、MOS7、MOS8和MOS9这7个传感器更多还原

【Abstract】 Nondestructive inspection of farm produce has been an interesting field in the past of 20 to 30 years. Without damaging the farm produce, the nondestructive method for the evaluation of fruits quality can be used to detect the internal quality and ingredients, and so on, which are related with fruits’ quality, and to evaluate internal quality, including sugar content, acidity, firmness, inner pathological changes, etc. By far, a great number of studies have been done on the nondestructive detection methods in domestic and overseas. And the results of many researches indicate that the firmness and aroma are important indexes for evaluating the fruit quality. However, the measurements for the fruits firmness by current methods usually damage fruits and operate slowly, in other words, it is not appropriate for the online measurement In order to overcome these disadvantages, many researchers have been studying on the nondestructive measurement techniques for the fruits firmness depending on the physical properties of fruits, such as optics, acoustics, electrics, and so on. Up to now, these methods have not been widely used in the commerce.Besides some routine examination such as physical and chemical variables, the quality of drink is evaluated by the experts through senses traditionally. The sensitivity of human’s sense, however, is liable to be changed by external factors, therefore, it is not easy to make an accurate evaluation through human’s sense. Alternatively, instrumental techniques such as gas chromatography (GC) with headspace sampling, and techniques such as GC combined with mass spectrometry (GC/MS) can be used to identify and quantify individual aroma components. These techniques usually have following disadvantages: 1) time consuming and/or labour intensive; 2) require sample preparation, 3) difficult to automate. The electronic nose (E-nose) offers a fast and non-destructive alternative to the measurement of volatile emission of samples. Commercially available E-noses use an array of sensors combined with pattern recognition software.This research work is focused on the quality evaluation of tomato and rice wine with E-nose. Based on the research, the correlation between the inner characteristic and response of sensors has been analyzed, and prediction models of tomato firmness and wine-year have been established. It provides a new method of nondestructive evaluation of food industry for further research.The electronic nose (PEN2) and enrichment unit (a sample pretreatment including adsorption sampling and subsequent thermal desorption can be applied with PEN2) produced by Airsense Analysis Gmbh corporation was used. This system is provided with objectivity, reliability and repetition.In the same time, the electronic nose responses to the fruit and wine aroma, so it is a non-destructive method to evaluate quality of fruit and wine. It has a convenience operation and celerity inspection.1 Study on the maturity, shelf life and damage of tomato with Electronic noseThe electronic nose was used to assess the different mature stages of tomato. When the maturity was distinguished according to color, the score plot of the principal component analysis(PCA) and linear discriminant analysis (LDA) for the E-nose measurements show that immature tomatoes can be distinguished from mature and over mature tomatoes. But the result is not good for mature and over mature tomatoes. When firmness was considered, the PCA analysis was able to classify the 100 % of the total samples in each different mature stage. The result shows that the difference of firmness can bedistinguished by E-nose. In order to assess the potential of electronic nose technique to predict fruit quality parameter, the electronic nose signal and the result derived from well-established traditional technique such as firmness were related using a multivariate technique. The calculations were carried out using ’The Unscrambler V9.1 1986-2004’ (CAMO, PROCESS, AS, OSLO, Norway), a statistical software package for multivariate calibration. Partial least square (PLS) which used Full Cross Validation test was used to build the prediction models. When signal of E-nose at 55S was used to PLS analysis, the calibration model for the firmness with four latent variables was determined through the residual variance analysis by each different component, the correlation for the firmness model was 0.950 and average standard error of prediction (RMSEP) was 0.129. The firmness could be well predicted with the electronic nose signal.The experiment of the changes of two different mature state tomatoes had been monitored during shelf life (Dayl-Day 11, Dayl4, Dayl7). Based on E-nose, a clear distinction among mature tomatoes on Dayl-Day6, Day7-Dayll and Dayl4-Dayl7 is available. And it is also possible to distinguish over mature tomatoes on Dayl-Day5, Day6-Dayll and Dayl4-Dayl7. The calibration models for the shelf life with eight latent variables were determined through the residual variance analysis by each different component, the correlation coefficient (R=0.989 and R= 0.980), average standard error of prediction (RMSEP=0.666 and RMSEP=0.912) for mature and over mature tomato respectively.The electronic nose was also used to assess the feasibility of damage inspection on tomatoes. Before sensors were optimized, the score plot of the principal component analysis (PCA) for the E-nose measurements had showed that undamaged tomato groups (5 undamaged tomatoes per group) could not be well distinguished from damaged groups (4 undamaged tomatoes and 1 damaged tomato per group), the score plot of linear discriminant analysis (LDA) for the E-nose measurements showed that it could be distinguishedon7th day after daunage. Based on loading analysis and the evolution of the signals generated by the sensor array, it can be inferred that the sensor MOS2> MOS6> MOS7> MOS8 and MOS9 have higher values, MOSK MOS3. MOS4^ MOS5 and MOS10 are not sensitive to the tomato volatile components. After sensors were optimized, seven-sensor-array with MOSK MOS2n MOS3^ MOS6. MOS7> MOS8 and MOS9 was used for analysis, it could be clearly distinguished on 1st day after damage based on LDA analysis, and the sum of variance was improved by 2.04%. So a subset of few sensors can be chosen to explain all the variance. This result could be used in further studies to optimize the number of sensors.2 Study on rice wine with Electronic nose and EDUIn order to analyze the potential of electronic nose with EDU to assess the rice wine in different year (same brand), different feature parameters had been extracted from the response curve from training samples measurement files and were applied to PCA analysis. The average during first five seconds signal is most appropriate for pattern identification from PCA analysis, the average during last five seconds and the average during sixty seconds take second place, the seventh second and the maximum responses of all sensors are unsuitable to pattern identification. It can be gotten that feature parameter is important determinant of pattern identification. Pattern file used to identify wine in different year (same brand) was created through the average of first five seconds signal. After the training, the pattern file was investigated in order to evaluate the discrimination capabilities with measurement files from measurement samples. The plots of the cluster analysis (CA) showed that samples were able to classify the 100 % of the total measurement samples. In order to assess the electronic nose to predict the different year old wine, the calculations were carried out using PLS analysis. Signal of the average of first five seconds was used to PLS analysis and to build the prediction models. Through the residual variance analysis by each different component, the calibration更多还原