被动FT-IR多组分气体非线性定量分析技术研究

Passive Fourier transform infrared spectroscopy has important application value in the field of multi-component gas quantitative analysis, such as industrial exhaust gases, chemical toxic gases, aircraft emissions, combustion gases, environmental monitoring, etc. The complex nonlinear relationship between infrared radiation spectral data and component concentrations makes the existing prediction methods large output errors, and the correction model has poor anti-interference ability and repeatability. This project quantitative description overlapping spectral of gas mixture and builds a component crossover generalized linear model, which can introduce nonlinear factors into the analysis model and reduce the dependency of the calibration data on the mixed component samples. For the concomitant data multicollinearity problem, a statistical analysis method will be used to evaluate the severity of the collinearity and then control initialization parameters for iterative process. The virtual component gas mixture spectral will be created to reduce the decomposition error. Taking into consideration the constantly changing characteristics of gas compositions in the field test, a multi-constrained MCR-ALS step-by-step iterative algorithm guided by the spectral integral energy-time curve will be established to realize complex multi-component gas quantitative analysis. Finally, on-line monitoring experiment with typical multi-component gas by passive FT-IR spectrometer will be built to verify quantitative analysis performance under the standard data sets. And corresponding optimization methods will be proposed at the same time. It is helpful to establish a passive FT-IR gas molecular spectral fingerprint database, which will play a very good role in promoting the passive FT-IR gas remote sensing real-time monitoring technology.

被动傅里叶变换红外光谱技术在多组分气体（如：工业废气、化学毒气、飞机排放、高温燃烧、环境监测等）定量分析领域具有重要的应用价值。红外辐射光谱数据与组分浓度之间复杂的非线性关系使得现有方法预测输出误差大，校正模型抗干扰能力和可重复性差。本项目对多组分气体混叠进行量化描述，构建组分交叉广义线性模型，实现在分析模型中引入非线性因子，降低校正数据对混合组分样本的依赖；针对伴随而来的数据多重共线性问题，通过统计分析方法评估共线性程度，控制迭代参数初始化，用虚拟组分替代降低分解误差；考虑实际情况中气体组分组合不断变化的特点，建立光谱积分能量-时间曲线指导的多约束MCR-ALS分步迭代算法，实现复杂多组分气体定量分析。最后，开展典型多组分气体被动FT-IR在线监测试验，结合标准数据集验证定量分析性能，优化方法模型，建立被动FT-IR气体分子光谱指纹数据库，推动被动FT-IR气体遥感实时监测技术的发展。

多组分气体分析是被动傅里叶变换红外光谱定量遥感的关键技术。光谱数据与气体组分浓度之间复杂的非线性关系是制约气体定性定量算法的核心难题。本项目同时从模型构建和迭代分解方法两方面入手，针对复杂场景中多组分混合气体（团）定量分析的需求，开展标准气样品和待分析气体（团）的被动FT-IR测量方案设计，构建多层气体辐射传输模型，分析路径辐射以解决光谱测量“背景”问题，构建标定试验与自适应定标参数相结合的辐射反演方法；研究气体分子光谱线型展宽机制，结合气体吸收光谱基础线型和展宽效应，从气体分子光谱“走势形状”特点出发，对多组分气体混叠进行量化描述，构建组分交叉广义线性模型，实现在分析模型中引入非线性因子，降低校正数据对混合组分样本的依赖；针对数据伴随的多重共线性问题，通过共线性程度定量分析指导交替迭代初始化参数估计，用虚拟组分光谱替代降低分解误差；考虑实际情况中气体组分组合不断变化的特点，建立光谱积分能量-时间曲线指导的多约束MCR-ALS分步迭代算法，实现复杂多组分气体定量分析。最后，以燃油燃烧火焰为典型多组分气体分析对象，用被动FT-IR测量系统对其实施在线监测，进行气体组分定量分析，结合标准气校正数据集验证气体光谱定量分析性能，研究环境变化与噪声影响分析算法的鲁棒性，优化方法模型，建立被动FT-IR气体分子光谱指纹数据库，最终推广应用到飞机排放监测等领域。本项目研究在国内外重要学术期刊及会议上发表论文11篇，其中8篇被SCI收录，3篇被EI收录，1篇SCI论文2020年7月称为高被引论文；授权美国发明专利1项。成果突破了复杂遥感环境下多组分气体定性定量分析中的瓶颈问题，对于环境监测、气体分析等以气体定性定量为核心的众多热点应用，给出了通用可行的方法模型。