论文部分内容阅读
温度校正是红外光谱定性定量分析中的一个关键问题,通过去除光谱数据中的温度效应,可以改善模型的线性度,从而提高模型的预测精度。通常的温度校正方法不仅需要记录训练集光谱的采集温度,而且需要记录测试集光谱的采集温度,这对很多实际应用中的光谱温度校正造成了困难。提出了一种基于模型的光谱温度预测及校正方法,通过训练集数据对光谱中的温度信息进行建模,利用模型的信息,从而能从测试集光谱数据中估计出采集温度,并进行光谱数据的温度校正,降低了温度校正方法对测试集光谱数据采集温度的依赖性。作为方法的验证,进行了两部分的实验:在第一部分的实验中,通过对十个浓度的水-乙醇二元混合物光谱数据的温度预测以及温度校正的实验,证明了本文方法的有效性;在第二部分的实验中,我们采用了Wulfert的经典温度校正方法 CPDS的实验数据和实验方案,对三元混合物的光谱数据进行温度预测以及温度校正,得到了不亚于CPDS方法的温度校正效果,同时也证明了该方法对三元混合物光谱数据的适用性。通过两部分的实验表明,在缺少测试集测量温度的情况下,提出的温度校正方法仍可对光谱数据进行有效的预测和校正,降低了温度校正方法对测试集数据的依赖性,从而提高了温度校正方法的适用性。
Temperature calibration is a key issue in the qualitative and quantitative analysis of infrared spectroscopy. By removing the temperature effect in the spectral data, the linearity of the model can be improved and the prediction accuracy of the model can be improved. The usual temperature correction method not only needs to record the collection temperature of the training set spectrum, but also needs to record the collection temperature of the test set spectrum, which makes it difficult to correct the spectral temperature in many practical applications. A model-based method of spectral temperature prediction and correction is proposed. The temperature information in the spectrum is modeled by the training set data. Using the model information, the collection temperature can be estimated from the test set spectral data, and the spectral data Of the temperature correction, reducing the temperature correction method of test set spectral data acquisition temperature dependence. As a validation of the method, two experiments were carried out. In the first part of the experiment, the temperature prediction of the spectral data of ten water-ethanol binary mixtures and the experiment of temperature correction were carried out to prove the effectiveness of the proposed method. In the second part of the experiment, we used Wulfert’s classical temperature calibration method CPDS experimental data and experimental program, the temperature prediction of the ternary mixture of spectral data and temperature correction, as the CPDS method of temperature correction effect , But also proved the applicability of this method to the spectral data of ternary mixtures. The experimental results show that the proposed temperature correction method can effectively predict and correct the spectral data in the absence of the measured temperature of the test set and reduce the dependence of the temperature correction method on the test set data, Applicability of temperature correction methods.