间接硬建模方法在混合物太赫兹时域光谱解析中的应用研究

发布时间：2018-04-03 00:34

本文选题：参数化模型　切入点：外推性　出处：《光谱学与光谱分析》2017年10期

【摘要】：太赫兹(Terahertz,THz)波通常是指位于微波和近红外之间的电磁波。由于很多化学和生物分子的振动和转动模式正好都位于THz波段,因此可以利用物质的这些"指纹谱"特性开展定性和定量分析研究。目前用于THz光谱的定量分析主要有主成分回归(PCR)以及偏最小二乘回归(PLSR)等方法,这些算法在建模时往往需要大量的样本进行监督学习,模型精度对训练样本依赖性较高,同时模型的外推性不易保证,在样本量不足或者外推性要求较高的场合,这些算法的使用会受到一定限制。针对这些问题,该研究提出一种利用光谱的参数化模型——间接硬建模方法(IHM),进行混合物太赫兹光谱解析和量化分析技术方案的研究。首先使用S-G平滑滤波方法滤除光谱中的噪声影响;同时考虑到太赫兹光谱特性,消除了人工基线对光谱解析产生的影响;随后,开展了IHM建模与分析研究,重点讨论了在两个训练样本数目情况下模型预测准确度的问题;为了验证该算法的可行性,制备了利福平、乳糖一水合物、微晶纤维素以及硬脂酸镁的四元混合物进行实验与建模分析;使用回归相关系数R和均方根误差RMSE对定量模型进行评价。将IHM方法和PLSR方法进行了比较,理论分析和实验结果表明,相对于传统方法,IHM方法建模所需的训练样本数量可减少至2个,与此同时量化分析准确度获得了提高,同时外推性也有所提升。
[Abstract]:Terahertzt THz wave is usually the electromagnetic wave between microwave and near infrared.Since the vibration and rotation modes of many chemical and biological molecules are located in the THz band, the qualitative and quantitative studies can be carried out by using these "fingerprint spectrum" characteristics of matter.At present, the quantitative analysis of THz spectrum mainly includes principal component regression (PCA) and partial least squares regression (PLSR). These algorithms often require a large number of samples to monitor and learn when modeling, and the model accuracy is highly dependent on training samples.At the same time, the extrapolation of the model is not easy to guarantee, and the use of these algorithms will be limited in the case of insufficient sample size or high extrapolation requirements.To solve these problems, an indirect hard modeling method based on the parameterized model of spectrum is proposed to study the technical scheme of terahertz spectral analysis and quantitative analysis of mixtures.First, S-G smoothing filtering method is used to filter the noise in the spectrum. Considering the terahertz spectrum characteristics, the influence of artificial baseline on spectral analysis is eliminated. Then, IHM modeling and analysis are carried out.In order to verify the feasibility of the algorithm, rifampicin and lactose hydrate were prepared.The quaternary mixture of microcrystalline cellulose and magnesium stearate was tested and analyzed, and the regression correlation coefficient R and root mean square error (RMSE) were used to evaluate the quantitative model.Compared with the IHM method and the PLSR method, the theoretical analysis and experimental results show that compared with the traditional method, the number of training samples needed for modeling can be reduced to 2, and the accuracy of quantitative analysis is improved.At the same time, extrapolation is also improved.
【作者单位】：浙江大学控制科学与工程学院工业控制技术国家重点实验室;
【基金】：国家自然科学基金项目(61473255,61307127) 国家教育部博士点专项基金项目(20110101110063) 浙江省自然科学基金项目(Q14F050010)资助
【分类号】：O433.4

【相似文献】