【作者】 谈爱玲；

【导师】 毕卫红；

【作者基本信息】 燕山大学 ， 光学工程， 2012， 博士

【摘要】 随着国民经济的迅速发展，水体环境污染日益严重，石油类物质是主要污染源之一。快速、准确的水中石油类污染物检测方法的研究对于及时掌握水质变化、有效控制水污染事故、保护水资源具有十分重要的理论研究价值和现实意义。本文对目前水中石油类污染物检测与分析的各种方法进行相关文献的调研和总结，研究基于光纤近红外消逝波吸收光谱探测与分析技术的水中石油类污染物检测新方法。提出光纤近红外消逝波吸收光谱探测结构，并对该结构进行理论和实验研究；研究适用于水中石油污染物种类定性分析和多组分石油污染物浓度定量分析的化学计量学算法。论文主要内容有：首先，详细介绍近红外光谱基本原理和化学计量学算法；分析汽油、柴油和煤油三种典型的石油类污染物组成基团的近红外光谱特性，论证利用光纤近红外光谱技术检测石油类污染物的可行性。其次，基于光纤消逝波吸收传感原理，对具有疏水亲油薄膜的光纤近红外消逝波吸收光谱探测单元进行研究；对消逝场能量与纤芯半径、剩余包层厚度、疏水亲油薄膜厚度、薄膜折射率和探测单元长度等关键参数之间的关系进行数值仿真研究，为探测单元的优化设计提供理论依据。然后，在讨论疏水亲油材料特点的基础上，利用聚苯乙烯高分子溶液对腐蚀后的光纤进行涂覆，在光纤外形成聚苯乙烯薄膜层，该薄膜既起富集水中油污的作用，又把水分子挡在消逝场的范围之外，避免水分子在近红外段强吸收产生的干扰；详细介绍探测单元的腐蚀和涂覆制作过程；配置汽油、煤油和柴油三类典型石油类污染物的油水混合溶液，并基于傅里叶变换近红外光谱仪搭建探测单元性能测试系统，将光谱仪自带探头测量的纯油、纯水与油水混合溶液的近红外光谱与探测单元直接探测的油水混合溶液的近红外光谱进行比较分析。再次，针对传统近红外光谱分析方法未充分考虑吸光度数据非负的特点导致分析结果缺乏合理解释的问题，研究基于带有稀疏约束的非负矩阵分解特征提取算法结合支持向量机分类的水中单一石油类污染物种类鉴别定性分析方法，深入讨论特征提取算法参数和支持向量机分类器参数对分类正确率的影响，优化近红外光谱定性分析模型。最后，针对多组分混合的复杂石油类污染物中各组分浓度定量分析问题，分别研究建立基于偏最小二乘回归和粒子群优化的偏最小二乘支持向量机回归的汽油、柴油和煤油三组分浓度的定量分析模型，给出了定量分析模型的最优参数并利用三组分的最优模型对验证集进行浓度预测，比较两种回归方法所建模型的预测结果。本文在利用光纤消逝波探测水中石油类污染物吸收光谱和石油类污染物近红外光谱分析两个方面进行了较为深入的理论研究和实验工作，能够为光纤近红外消逝波吸收光谱探测在环境监测领域的实际应用提供有价值的参考。更多还原

【Abstract】 Pollution of the water environment is worsening with the rapid growth of the nationaleconomy. Petroleum pollutant is one of the major sources of pollution in water. Accurate,rapid, and convenient detection method of petroleum pollutants in water has veryimportant theoretical value and practical significance in grasping the changes in waterquality, effectively controling water pollution accidents and the protection of waterresources.This paper summarizes relevant literature of the various testing methods forpetroleum pollutants in water and proposes a new method to detect petroleum pollutants inwater based on fiber-optic evanescent wave absorption spectroscopy in combination withNIR analysis technology. The fiber-optic NIR-evanescent wave absorption spectroscopydetection unit structure is proposed. Theoretical and experimental researches are done forthe detection unit; suitable NIR chemometrics algorithms for qualitative identification ofsingle petroleum pollutants and quantitative analysis of multi-component complexpetroleum contaminants are discussed. The major contents of the paper are as follows:First, the paper introduces the basic principle of the NIR spectroscopy and commonlyused chemical metrology algorithm in detail; The analysis of three typical oil pollutantsNIR spectra of gasoline, diesel and kerosene are conducted to demonstrate the feasibilityof using fiber-optic NIR spectroscopy to detect petroleum pollutants in water.Second, the paper proposes the optic fiber NIR-evanescent wave absorption detectionunit structure with hydrophobic oleophilic film, numerically calculating the relationshipbetween the evanescent wave energy with the main parameters including the radius of thefiber core, the remaining thickness of the cladding layer, hydrophobic film thickness, filmrefractive index and the length of the detection unit, providing theoretical basis for theoptimal design of the detection unit.Third, after discussing the characteristic of the hydrophobic oleophilic materials, theidea of coating the polystyrene polymer solution on the surface of the corrosion singlemode fiber is studied. The hydrophobic oleophilic coating not only avoids the watermolecules’ strong absorption interference in the near-infrared region, but also plays the role of adsorption of oils in water; the corrosion and coating process of detection unit isdescribed in detail and the gasoline, diesel and kerosene oil-water mixed solution is made,and the detection unit performance test system is set up based on the fourier transformnear-infrared spectrometer; then the paper compares the near-infrared spectra of theoil-water mixed solution gained by the detection unit with the NIR spectra of pure oil,pure water and oil-water mixed solution gained by the probe measurements of thespectrometer.Fourth, traditional NIR methods do not take full account of the absorbance datanon-negative characteristics, resulting in the analysis lack of reasonable explanation. Forthis problem, the qualitative discriminate method of single species petroleumcontaminants based on non-negative matrix factorization feature extraction combined withsupport vector machine classification algorithm is studied. Non-negative matrixfactorization algorithm and support vector machine classifier parameters on classificationaccuracy are discussed in depth to optimize NIR qualitative classification model.Last, for the problem of quantitative analysis of the complex multi-componentmixing petroleum pollutants, the quantitative analysis model of gasoline, diesel andkerosene, a three-component mixed pollutant solution is established based on partial leastsquares regression algorithm and partial least squares support vector machine regressionalgorithm respectively. The optimal parameters of quantitative model are given based onparticle swarm optimization. The paper uses the three-component model to predict theconcentration of the validation set, and compares the predicted results with two regressionmethods.This paper conducted in-depth theoretical studies and experimental work of usingfiber-optic NIR evanescent wave absorption spectroscopy to detect petroleum pollutants inwater. The research can provide a valuable reference for the use of fiber evanescent waveabsorption spectroscopy detection and NIR spectroscopy technology in the field ofenvironmental monitoring applications.更多还原