| 摘要: |
| 目的 为提高高危气体检测的安全性与效率，减少实际操作中的安全隐患，研究基于高分辨率透射分子吸收数据库（HITRAN数据库）的高含量H2S混合气体红外光谱定量分析方法，并验证其在工业、环境监测和公共安全领域中的应用可行性。方法 利用傅里叶变换红外光谱技术（FTIR），结合支持向量回归（SVR）和径向基函数（RBF）神经网络模型，对含H2S、CO2和CH4的混合气体数据进行定量分析。通过HITRAN数据库生成高精度理论光谱数据，并采用光谱叠加方法模拟混合气体光谱，同时加入噪声模拟FTIR仪器的响应特性，以更接近实际检测环境。结果 该方法在多组分气体的定量分析中表现出较高的效率和精度，其中基于径向基核函数的支持向量回归（R-SVR）模型效果优于RBF神经网络模型，能够实现更高精度的检测结果。结论 为高含量H2S混合气体检测提供了一种低成本、高效且安全的仿真验证手段，同时为实际应用中的多组分气体检测提供了可靠的技术支持，具有重要的工程实践价值。 |
| 关键词: H2S HITRAN数据库 红外光谱 定量分析 RBF神经网络 支持向量回归 |
| DOI：10.3969/j.issn.1007-3426.2025.03.018 |
| 分类号: |
| 基金项目:中国石油西南油气田公司天然气净化总厂科研项目“天然气净化厂酸气主要组分在线检测技术研究”（2023-01） |
|
| Quantitative analysis of high sulfur-containing gases by infrared spectroscopy based on HITRAN database |
|
YANG Zhenggang1, ZENG Qiao2, XI Ningkai1, GAO Jin1, LI Taifu3
|
|
1.Natural Gas Purification Plant General, PetroChina Southwest Oil & Gasfield Company, Chongqing, China;2.College of Safety Science and Engineering, Chongqing University of Science and Technology, Chongqing, China;3.College of Innovation and Entrepreneurship, Chongqing University of Science and Technology, Chongqing, China
|
| Abstract: |
| Objective To enhance the safety and efficiency of hazardous gas detection while minimizing operational risks, this study investigates a quantitative analysis method for high hydrogen sulfide(H2S)-containing gas mixtures based on the high-resolution transmission molecular absorption database (HITRAN database), and further validates the feasibility of its application in the fields of industrial, environmental monitoring, and public safety. Method Fourier transform infrared (FTIR) spectroscopy was employed in conjunction with support vector regression (SVR) and radial basis function (RBF) neural network models to perform quantitative analysis on gas mixtures containing H2S, CO2, and CH4. High-precision theoretical spectra data were generated using the HITRAN database, and a spectral superposition method was applied to simulate the infrared spectra of gas mixtures. The noise was added to simulate the response characteristics of FTIR instruments, making the simulated spectra closer to real detection scenarios. Result The proposed method demonstrated high efficiency and precision in the quantitative analysis of multi-component gas mixtures. The radial basis function kernel-based SVR (R-SVR) model outperformed the RBF neural network model, achieving higher detection precision. Conclusion This study provides a low-cost, efficient, and safe simulation-based validation method for detecting high H2S-containing gas mixtures. It offers reliable technical support for multi-component gas mixtures detection in practical applications and holds significant value for engineering practices. |
| Key words: H2S HITRAN database infrared spectroscopy quantitative analysis RBF neural network support vector regression |
