【作者】 代秀迎；

【导师】 陈斌；

【作者基本信息】 江苏大学 ， 农产品加工及贮藏工程， 2010， 硕士

【摘要】 研究快速的食用植物油品质检测方法对于市场上食用植物油的质量监督具有重要的现实意义。本研究以食用植物油为检测对象,结合化学计量学方法开展了衰减全反射傅立叶变换红外光谱分析技术(FTIR-ATR)定量定性分析的研究,综合多学科的知识对红外光谱数据中的信息提取和模型建立中遇到的关键问题给出了解决方案,论证了该技术的可行性。旨在为植物油品质的快速无损检测提供新的参考。主要有以下研究内容：1.利用红外光谱分析技术结合偏最小二乘法(PLS)定量分析食用植物油中主要脂肪酸含量。在PLS模型建立过程中,通过马氏距离法和学生残差及杠杆值法对异常样品进行了剔除,并优化了光谱预处理方法。结果得到棕榈酸、硬脂酸、油酸、亚油酸和亚麻酸模型预测集相关系数(R2)分别为0.963、0.798、0.997、0.996和0.991。从研究结果可以看出,利用红外光谱结合PLS算法可以很好的检测食用植物油中主要脂肪酸含量,为食用植物油品质的快速检测提供了依据。2.利用红外光谱分析技术结合SIMCA模式识别方法对3种食用量最大的食用植物油进行定性分析。对光谱预处理后建立SIMCA模型,菜籽油、花生油和芝麻油的预测准确率分别为100%、100%和97.5%。结果表明采用红外光谱结合SIMCA分类法对不同品种食用植物油的分类和预测是可行的。3.利用红外光谱分析技术结合SIMCA模式识别方法和PLS+BP神经网络的方法对不同制取工艺的同类植物油进行识别分析。首先利用SIMCA法分别对3种植物油进行工艺判别,菜籽油、花生油和芝麻油的预测准确率分别为80%、75%和95%；然后用PLS+BP神经网络的方法对预测结果不理想的菜籽油和花生油进行了识别分析,对光谱预处理后用PLS+BP网络建模,通过反复试验选择了最佳的模型参数,菜籽油和花生油的预测准确率提高为100%和90%。结果表明红外光谱法结合适当的化学计量学方法识别同类植物油的制取工艺是可行的。4.特征谱区筛选法优化红外光谱检测硬脂酸的定量分析模型。分别利用区间偏最小二乘和遗传偏最小二乘方法来筛选特征光谱区域,并比较了它们的结果。从试验结果看,它们都能在提高模型精度的同时简化模型,遗传偏最小二乘法得到的结果较佳,该模型预测集的相关系数(R2)为0.8260。更多还原

【Abstract】 Study on the rapid detection of vegetable oil quality is an attractive and prospective subject for quality supervision of vegetable oil in market. In combination with the chemometrical methods,the attenuated total reflectance fourier transform infrared spectroscopy(FTIR-ATR) analysis technology was used for vegetable oil quantitative and qualitative analysis.In order to provide new reference for the rapid and nondestructive examination for vegetable oil quality detection,the study was carried out by several subjects methods to come up with solution on the abstraction of weak spectral signal and establishing model and the feasibility of the tecnology was demonstrated. The main contents and conclusions are as follows:1. IR spectroscopy combined with partial least square (PLS) was applied to building the quantitative model to quantitatively predict the main fatty acid contents in vegetable oil.In building model, outliers were removed by mahalanobis distance and studentized residual versus leverage methods,the effects on the spectral preprocessing methods were also discussed. The correlation coefficient(R2) between the prdicted and the reference results for the test set is used as an evaluation parameter for the models:the palmitic acid,stearic acid,oleic acid,linoleic acid and linolenic acid results R2=0.963,0.798,0.997,0.996 and 0.991,respectively.It can be concluded that the main fatty acid contents in vegetable oil can be analyzed fast by IR spectroscopy coupled with the appropriate chemometrics methods,and this real-time,at-site measurement will significantly improve the efficiency of quality control and assurance.2. IR spectroscopy combined with pattern recognition based on SIMCA was applied to buliding the qualitative model to qualitatively predict 3 kinds of vegetable oil which have the largest consumption.The infrared spectra which after preprocessing were used to bulid forecasting model,using SIMCA method. The predicting accuracy of rapeseed oil,peanut oil and sesame oil were 100%、100% and 97.5%,respectively.The results showed that different species of vegetable oil can be classified by IR spectroscopy coupled with the appropriate chemometrics methods. 3. IR spectroscopy combined with SIMCA and back-propagation(BP) neural network was applied to buliding the qualitative model to recognize the similar vegetable oil of different processing methods. The predicting accuracy of rapeseed oil,peanut oil and sesame oil were 80%,75% and 95% by SIMCA pattern recognition method. To optimize the predictions, the infrared spectra of rapeseed oil and peanut oil which after preprocessing were used to build forecasting model again,using PLS+BP regression analysis. The optimal model parameters were selected through intensive trial-by-trail analysis, the predicting accuracy of rapeseed oil and peanut oil were 100% and 90%.It has certain feasibility for infrared spectroscopy to identify different processing technology of the similar vegetable oil with the appropriate chemometrics methods.4. Selection of the efficient wavelength regions in FT-IR spectroscopy was used for determination of stearic acid content in vegetable oil.In order to improve the precision and robustness of the stearic acid model, interval partial least-squares(iPLS) and genetic algorithm partial least-squares(GA-PLS) were applied to selecting the efficient spectral regions.Both models were able to produce better prediction models in relation to the full-spectrum model,and the models were simple and easier.Experimental results showed that the performance of GA-PLS model was better than iPLS model,and the optimal model was achieved with correlation coefficient R2=0.8260 in prediction set.更多还原