光谱学与光谱分析, 2019, 39 (10): 3313, 网络出版: 2019-11-05  

三维荧光光谱结合组合算法在环境污染监测中的应用: 油种鉴别与定量分析

Application of Three-Dimensional Fluorescence Spectra Combined with Algorithm Combination Methodology in Environmental Pollution Monitoring: Oil Identification and Quantitative Analysis
作者单位
1 华北理工大学电气工程学院, 河北 唐山 063210
2 燕山大学河北省测试计量技术与仪器重点实验室, 河北 秦皇岛 066004
摘要
针对油类污染物成分复杂, 光谱重叠难以识别的问题, 提出采用三维荧光光谱结合组合算法对油类污染物进行了定性和定量分析。 荧光光谱中存在的瑞利散射对三维荧光光谱检测有较大影响, 提出了缺损数据修复-主成分分析(MDR-PCA)方法对矿物油三维荧光光谱的瑞利散射进行处理, 原理是单个荧光光谱激发发射矩阵符合双线性, 可用主成分分析(PCA)法来解析。 MDR-PCA法首先将荧光数据中的散射干扰数据全部扣除, 之后利用主成分分析(PCA)迭代过程对扣除数据进行重构修复后补全数据。 该方法在消除散射干扰的同时充分利用了荧光物质光谱矩阵中的有效信息。 利用不同浓度的矿物油的激发-发射荧光光谱构建了三维数据。 样品数据来源于柴油、 汽油和煤油三种溶质的四氯化碳溶液。 常用于三维荧光光谱数据分析的三线性分解算法有平行因子分析(PARAFAC)、 交替三线性分解(ATLD)和自加权交替三线性分解算法(SWATLD)等。 PARAFAC基于严格意义上的最小二乘原则, 具有抗噪声强、 模型稳定、 微小预期误差等优点, 可以实现三维数据阵列的最佳拟合, 但该算法收敛速度较慢, 对组分数敏感。 ATLD算法通过提取对角主元和切尾奇异值求解广义逆, 极大提高了收敛速度并降低了对组分数的敏感度, 从而实现三线性分解。 然而, 取对角元时易使ATLD方法对噪声敏感。 SWATLD算法既继承了对组分数不敏感、 收敛速度快等优点, 又降低了噪声水平的影响。 但是在抗共线程度方面, SWATLD算法在抵抗共线性程度方面的能力较ATLD略有降低。 基于此, 论文根据三线性分解算法迭代过程中损失函数的变化, 对迭代过程进行划分, 提出了三线性迭代方法的组合算法(algorithm combination methodology, ACM)—将ATLD, SWATLD与PARAFAC组合在一起, 充分发挥各算法的优点, 实现二阶校正算法的优势互补。 采用ACM算法对两组分及三组分矿物油样品的三维荧光光谱数据进行解析, 并对三种矿物油的回收率进行了计算。 柴油的回收率为97.08%, 汽油的回收率为97.34%, 煤油的回收率为97.25%。 解析光谱和回收率表明, ACM算法能够实现油类污染物的种类识别及浓度测量。
Abstract
In order to solve the problem that the composition of oil pollutants is complex and the spectrum overlap is difficult to identify, qualitative and quantitative analysis of oil pollutants was carried out by three-dimensional fluorescence spectroscopy combined with Algorithm Combination Methodology (ACM). Rayleigh scattering in fluorescence spectra has a great influence on the detection of three-dimensional fluorescence spectrum. In this paper, the missing data recoveryr-principal component analysis (MDR-PCA) method was proposed. The principle is that the single fluorescence spectrum excitation emission matrix conforms to bilinearity and can be analyzed by principal component analysis (PCA). The scattering interference data were first deducted completely, and then the deducted part was repaired by using the remaining effective signal data in the iteration process. This method not only eliminates the scattering interference, but also makes full use of the effective information in the fluorescence spectrum matrix. The three-dimensional data were constructed by using the excitation-emission fluorescence spectra of different concentration mineral oil. The sample data were obtained from carbon tetrachloride solutions of 0# diesel, 95# gasoline and ordinary kerosene solutes. The trilinear decomposition algorithms commonly used for three-dimensional fluorescence spectral data analysis include parallel factor analysis (PARAFAC), alternating trilinear decomposition (ATLD), and self-weighted alternating trilinear decomposition algorithm (SWATLD). PARAFAC is based on the strict principle of least squares and has strong anti-noise ability. Its model is the stablest and the error is expected to be the smallest. It can provide the best fit of 3D data array, but the convergence speed of PARAFAC algorithm is slow and correct. The estimated number of components is more sensitive. The ATLD algorithm is based on the Moore-penrose generalized inverse of singular value decomposition to realize the trilinear model decomposition. By using the inverse diagonal element and the tangent singular value to solve the generalized inverse, the convergence speed of the method is greatly improved, and the sensitivity of the algorithm to the component number is reduced, but the operation of the diagonal element makes the ATLD method more sensitive to noise. SWATLD inherits the advantages of ATLD, which is insensitive to the number of components and fast convergence, and has the characteristics of being insensitive to noise levels. However, the SWATLD algorithm has a slightly lower ability to resist collinearity than ATLD. This paper divides the iteration process according to the change of loss function in the iteration process of trilinear decomposition algorithm, and proposes the algorithm combination method (ACM)-combining ATLD, SWATLD and PARAFAC, giving full play to the advantages of each algorithm, and realizing the complementary advantages of the second-order correction algorithm. The three-dimensional fluorescence spectra of two-component and three-component mineral oil samples were analyzed by ACM algorithm, and the recovery rates of three mineral oil samples were calculated. The recovery rate of diesel was 97.08%, the recovery rate of gasoline was 97.34%. and the recovery rate of kerosene was 97.25%. The analytical spectrum and the recovery rate show that the ACM algorithm can realize species identification and concentration measurement of oil pollutants.

陈至坤, 黄微, 程朋飞, 沈小伟, 王福斌, 王玉田. 三维荧光光谱结合组合算法在环境污染监测中的应用: 油种鉴别与定量分析[J]. 光谱学与光谱分析, 2019, 39(10): 3313. CHEN Zhi-kun, HUANG Wei, CHENG Peng-fei, SHEN Xiao-wei, WANG Fu-bin, WANG Yu-tian. Application of Three-Dimensional Fluorescence Spectra Combined with Algorithm Combination Methodology in Environmental Pollution Monitoring: Oil Identification and Quantitative Analysis[J]. Spectroscopy and Spectral Analysis, 2019, 39(10): 3313.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!