【作者】 张亮；

【导师】 师俊玲；

【作者基本信息】 西北农林科技大学 ， 食品科学， 2010， 硕士

【摘要】 展青霉素是一种神经毒性真菌毒素,广泛存在于苹果浓缩汁中。扩展青霉是苹果中主要的展青霉素产生菌。开发快速简便的扩展青霉和展青霉素检测技术有助于及时控制产品中展青霉素的水平。近红外光谱检测法具有对样品无损、速度快、重现性好等优点,适用于在线检测。为了开发扩展青霉和展青霉素的近红外检测技术,论文研究了近红外光谱分析技术在纯培养、单菌种污染苹果、多菌种污染苹果及不同污染深度时对扩展青霉的定性定量检测特性,以及其对水溶液、鲜榨苹果汁和浓缩苹果汁中展青霉素的定性定量检测特性:确定了近红外光谱法检测扩展青霉和展青霉素的条件与方法。获得以下主要研究结果:（1）平皿纯培养的条件下,利用全谱区信息就能有效地将扩展青霉与其它苹果致腐菌区分开来,对数据进行矢量归一化法+一阶导数法处理以后,区分效果更好。在孢子悬浮液和苹果污染体系,需要对数据进行矢量归一化法+一阶导数法处理,选取5883～5342 cm^-1,4579～4324-cm^-1。波段信息,才能有效地将扩展青霉与其它霉菌区分开来,而且对表层污染（3mm）的区分效果优于深层污染（6mm）。在孢子悬浮液中,近红外对扩展青霉的检测限为1.5×10¹个/ml;在污染苹果的情况下,对扩展青霉的检测限为1.5×10³个/ml;与其它霉菌污染苹果时,检测限为1.5×10³个/ml。（2）近红外光谱用于展青霉素检测时,需要对光谱进行适当的预处理,并选择最适主成分数及5883～5342 cm^-1,4579～4324 cm^-1。波段范围内信息才能实现对不同体系中展青霉素进行定性定量分析。不同体系中结果如下:水溶液中,用矢量归一化法预处理光谱,主成分数为3,所得模型的决定系数（R²）为94.03%,交互检验方根差（RMSECV）为0.152μg/L;内部交叉验证的预测集决定系数（R²）为93.1%,预测标准偏差（RMSEP）为0₃lμg/L。检测限为9.29μg/I.,浓缩苹果汁中,用最大最小归一化法预处理光谱,主成分数6,所得模型的R²=93.53%, RMSECV=0.42μg/L;内部交叉验证的预测集的R²=92.65%, RMSEP=0.532μg/L,平均样品回收率为99.97%;检测限为9.76μg/L鲜榨苹果汁中,用矢量一化法预处理光谱,主成分数7,所得模型的R²=92.01%, RMSECV=0.7μg/L;内部交叉验证的预测集的R²=90.73%, RMSEP=0.264μg/L;检测限为9.54μg/L上述结果表明,近红外光谱法可以很好地应用于扩展青霉和展青霉素的定性定量检测。更多还原

【Abstract】 Patulin is a mycotoxin with neurotoxicity, widely found in apple juice concentrate. Penicillium expansum was reported as the major patulin-producing fungus in apple. Development of quick and easy method for testing Penicillium expansum and patulin is important to effectively control of the two harmful factors. Near infrared spectroscopy method has great potential in online detection due to its no damage to sample, quick and good reproducibility. In order to explore good method in test of Penicillium expansum and patulin, the thesis investigated the ability of near-infrared spectroscopy in detecting Penicillium expansum under different conditions including pure culture, solely contamination and multiple contamination of different mould in apple. The capability was evaluated in quality and quantity. The application of near infrared-spectroscopy in detection of patulin was also evaluated in different conditions including water solution, fresh apple juice and apple juice concentrates. The cluster analysis and predicting model establishment were used in quality test evaluation and quantity test evaluation, respectively. The main results were obtained are listed as following:（1）In the case of pure culture in solid medium, Penicillium expansum can be effectively distingushed from other fungi by using the information in whole wavelength region without any data pretreatment method, while the data pretreatment of first derivative method and vector normalization gave better results. In cases of spore suspension and contamination in apple, data pretreatment of first derivative method and vector normalization was needed to yield most effective distinguish when information was extracted from the wavelength regions of 5883⁵342 cm^-1 and 4579⁴324 cm^-1. The result was better to the contamination shallow under apple surface （3 mm） than that deeper （6mm）. The testing limit of Penicillium expansum showed as 1.5x101 per ml for spore suspension and single mold contamination and 1.5x103 per ml for simultaneous contamination of different molds in apple.（2） In order to give good distinguish and quantitive test of patulin in different conditions, it is necessary to use suitable data pretreatment, main factor number and extraction of information from the spectroscopy in the wavelength regions of 5883⁵342 cm^-1. Results obtained for different conditions are listed in the follows.For water solution system, the best data pretreatment method is vector normalization at main factor number of 3. The best predicting model was obtained with coefficient of determination（ R² ）of 94.03% and cross-validation root deviation （ RMSECV ）of 0.152 ug/L. The internal cross-validation of the predicting model yielded coefficient of determination （R²） of 93.1%, standard deviation of prediction （RMSEP） of 0.31μg/L. The detection limit was calculated as 9.29μg/L.For the apple juice concentrates, the best data pretreatment method was maximum and minimum normalized pretreatment with main factor number of 6. The predicting model was obtained as R²=93.53% and RMSECV=0.42μg/L. The internal cross-validation of the predicting model showed R²=92.65% and RMSEP=0.532μg/L with the detection limit of 9.76μg/L.For the fresh apple juice, the best data pretreatment showed as vector normalization with main factor number of 7. The best predicting model yield R²=92.01%, RMSECV=0.7 ug/L and internal cross-validation of R²=90.73% and RMSEP=0.264μg/L with the detection limit of 9.54μg/L.更多还原