太赫兹时域光谱技术(THz-TDS)是近十多年发展起来的一种新的远红外光谱技术, 在气体研究方面有了一定进展, 尤其是对极性气体, 而对非极性气体研究较少。 本文以干馏气、 天然气以及各种沼气的主要成分CH4, C2H6和C3H8气体(非极性气体)为研究对象, 首先对CH4, C2H6和C3H8三种纯气进行测量, 利用THz-TDS技术得到其太赫兹频域谱和相位谱, 然后将其以不同比例、 不同种类混合成二元气体, 进一步研究混合气体的频域谱和相位谱。 实验结果表明CH4, C2H6对太赫兹波的吸收很小而C3H8对太赫兹波有一定的吸收, 这与C3H8极性增强的物理特性相符合。 为了实现对烷烃混合气体的压强和各成分浓度的定量分析, 本文利用BP人工神经网络法对上述二元混合体系的太赫兹频域谱进行分析, 对训练集和预测集分别计算了混合气体的压强和各成分浓度的预测值与实际值的相关系数, 训练集和预测集的相关系数取值分别为0994～0999和0981～0993。 研究表明, 利用太赫兹时域光谱技术结合BP人工神经网络数学方法可以实现对烷烃混合气体的压强和各成分浓度的定量分析, 使THz-TDS技术在气体领域研究范围更加广阔。

Terahertz time-domain spectroscopy (THz-TDS), a newly developed far-infrared technique in recent years, has been widely applied for the detection of polar gas and non-polar gas. The principal components in distillation, natural and marsh gas include methane (CH4), ethane (C2H6) and propane (C3H8). The three kinds of alkane gas are separately measured for the frequency domain and phase spectra, and then mixed as a series of two-component systems which are finally measured with THz-TDS. The frequency dependent amplitude and phase spectra prove that C3H8 has stronger absorption in THz range compared with CH4 and C2H6. Such trend is agreement with the polarity intensities of alkanes. In addition, back propagation artificial neural network (BPANN) is employed to quantitatively analyze the contents and total pressure with the frequency-amplitude spectra as the input. The correlation coefficients R between actual and predicted values are calculated in both training and prediction sets, where r range from 0994 to 0999 and from 0981 to 0993, respectively. Therefore, the combination of THz-TDS and BPANN can realize the quantification of multiple properties in alkane mixtures, which would promote the application of THz-TDS technique for gas detection.