红外光谱分析技术已被液体、 固体和气体混合物各组分的定量检测广泛应用。 针对红外光谱定量检测过程中存在的计算过程复杂、 受背景及噪声干扰、 多组分谱线混叠的问题, 一方面, 通过获取谱线二阶导数去除背景及噪声干扰, 并一定程度上区分混叠谱峰。 在获取二阶导数谱线过程中运用Savitzky-Golay滤波器进行平滑滤波处理, 并根据谱线频率特征选取Savitzky-Golay滤波器最佳滤波参数, 解决了滤波参数缺乏标准化选取方法的问题。 另一方面, 根据被测组分的吸收分布情况生成特征吸收窗分别对待测原始吸收谱线以及二阶导数谱线加以处理, 提取出在浓度计算过程中占据重要性更大的特征吸收的区域, 从而排除非吸收区域中背景、 噪声及其他组分对计算结果的干扰。 以一组丙烷、 丙烯和甲苯混合气体为例, 运用二阶导数方法以及特征吸收窗方法进行定量检测, 并与直接利用原始谱线的检测结果进行比较分析。 实验结果表明采用二阶导数谱与特征吸收窗的定量检测方法精度更高, 相对误差基本在5%以内。

In the field of detection, infrared spectroscopy has been widely used to quantitatively detect components in liquid, solid and gas mixtures . Aiming at the problems of complicated calculation process, background and noise interference, and spectrum overlap from multi components in infrared spectroscopy quantitative detection. On the one hand, a method based on second order derivative spectrum for the original absorption line was used to quantitative detection. The background and noise interference were removed in second order derivative spectrum, and the aliasing absorption peaks were distinguished to a certain extent. In the process of obtaining second order derivative line, Savitzky-Golay filter was used for smoothing filtering the spectrum. The optimal filter parameters were selected according to the spectral line frequency characteristics, which makes up the lack of the standardized methods in filtering parameters selecting. On the other hand, the characteristic absorption windows in term of the absorption distribution of corresponding components were applied to both original absorption line and second order derivative spectrum. The characteristic absorption region of more importance in the concentration calculation was extracted, so as to eliminate the interference of background, noise and other components in the non absorption region. The multi components mixtures of propane, propylene and methylbenzene were used as samples for quantitative detection with both methods mentioned above together with the method using original absorption lines. Then a comparison of the results by different method was analyzed. The experiment results show that the quantitative detection methods by using the second order derivative spectrum and the characteristic absorption window can achieve higher accuracy with relative error less than 5%.