论文部分内容阅读
为了对90~#汽油,93~#汽油,0~#柴油和航空煤油等4种成品油进行快速准确地定性鉴别,提出了1种基于目标因子分析算法近红外光谱技术快速鉴别成品油的新方法。分别收集了46个90~#汽油,63个93~#汽油,89个航空煤油,138个柴油样品,利用MPA傅立叶变换近红外光谱仪对样品在(10 000～4 400)cm~(-1)光谱区域进行近红外光谱测量。采用主成分分析结合马氏距离及目标因子分析方法对4种成品油进行模式识别定性分析,选择合适的阈值和目标因子主成分数,结合二维投影和三维投影图,比较了成品油样品在近红外光谱上的差异程度,确立了最优定性判别模型。研究结果表明:采用传统的马氏距离法无法对90~#和93~#汽油进行准确识别,而采用目标因子分析算法能够表征336个样品的类别关系,能更全面可靠地表征样品的特征信息。选择24个90~#汽油,42个93~#汽油,103个0~#柴油和59个航空煤油样品进行验证,其正确识别率分别95.8%,97.6%,99.0%和100%,具有良好的分类和鉴别作用,为成品油的鉴别提供了1种新方法,具有广泛的应用前景。
In order to rapidly and accurately identify four refined oil products such as 90 ~ # gasoline, 93 ~ # gasoline, 0 ~ # diesel oil and aviation kerosene, a new method based on target factor analysis (NIRS) method. A total of 46 samples of 90 ~ # gasoline, 63 93 ~ # gasoline, 89 aviation kerosene and 138 diesel samples were collected and analyzed by MPA Fourier transform near infrared spectroscopy at (10,000 ~ 4,400) cm ~ (-1) Spectral range for near infrared spectroscopy. The principal component analysis combined with Mahalanobis distance and the target factor analysis method were used to qualitatively analyze the four oil samples. The suitable thresholds and the principal components of the target factors were selected. Combining the two-dimensional projection and three-dimensional projection images, Near infrared spectroscopy differences, established the best qualitative discriminant model. The results show that the traditional Mahalanobis distance method can not accurately identify 90 ~ # and 93 ~ # gasoline, and the target factor analysis algorithm can characterize the classification of 336 samples, and can more fully and reliably characterize the sample information . A total of 24 90 ~ # gasoline, 42 93 ~ # gasoline, 103 0 ~ # diesel and 59 aviation kerosene samples were selected for validation. The correct recognition rates were 95.8%, 97.6%, 99.0% and 100% Classification and identification of the role of oil for the identification of a new method provides a wide range of applications.