基于太赫兹时域光谱分析红层泥岩风化样本

红层泥岩边坡的风化是一个普遍存在而又未完全解决好的问题。因风化而造成的不稳定红层泥岩边坡，在岩土体重力、震动及其他因素作用下，常常发生危害性的变形与破坏，导致崩塌滑坡、落石和泥石流等地质灾害。风化深度是红层泥岩边坡治理的重要依据之一。风化深度是指岩石破裂后在破裂面内侧产生的微裂隙或微裂隙带在风化作用下向岩石内部延伸的垂直距离。选择典型的红层泥岩边坡地带进行钻孔取样。太赫兹时域光谱技术是利用太赫兹脉冲表征物质性质的一种新兴光谱技术。实验使用太赫兹时域光谱系统测量所取得样本的太赫兹透射光谱。对所有样品的透射光谱进行分析，观察到各风化样品间的太赫兹透射光谱差别微小，且没有明显的特征吸收峰。为了获得取样地点的风化深度，基于太赫兹光谱建立一个智能，高效的回归预测支持向量机(SVM)模型来预测采样地点的风化深度。预测结果与实际深度相比较，所得相对误差小于7.09％。结果表明，基于太赫兹光谱的支持向量机(SVM)模型可以有效的区分样本以及预测深度，可以为红层泥岩边坡治理提供重要的参考。

Study on the Samples of Weathered Red-Bed Mudstone Based on THz-TDS

The weathering universally takes place in red‐bed mudstone slope ,which is not completely solved till now .Under the influence of geotechnical weight force ,vibration ,and other factors ,unsteady red‐bed mudstone slope caused due to weathering often lead to harmful deformation and failure ,resulting in collapse of geological disasters such as landslides ,falling rocks and mudslides .Weathering depth is the vertical distance after the rock burst produced inside the fracture surface micro cracks or micro‐cracks in the weathering zone extends to the interior of the rock .Weathering depth can provide an important basis for red‐bed mudstone slope treatment .The rock core was obtained by using the core drilling machine in the typical red‐bed mudstone slope .Terahertz time‐domain spectroscopy is a new spectroscopy approach to characterize material based on terahertz pulse . Terahertz time‐domain spectroscopy is used to measure the terahertz transmission spectroscopy of sample obtained .The THz transmission spectra of samples are being analyzed .Only minor differences are found among the transmission spectra of different samples ,and there are not characteristic absorption peaks been observed .To obtain weathering depth of sampling site ,an intel‐ligent and efficient SVM model of the transmission spectra of samples is employed to distinguish and predict the depth .Com‐pared with the actual depth ,the relative error is less than 7.09% .The results showe that THz spectroscopy based on support vector machine (SVM) model can effectively distinguish samples and the predict depths ,which can provide an important refer‐ence for the red‐bed mudstone slope treatment .