光谱学与光谱分析, 2020, 40 (4): 1051, 网络出版: 2020-12-11   

基于背景水分扣除的水泥生料成分的近红外光谱建模

Near Infrared Spectroscopic Modeling Method for Cement Raw Meal Components by Eliminating Background Moisture
作者单位
1 中国科学技术大学环境科学与光电技术学院, 安徽 合肥 230026
2 中国科学院安徽光学精密机械研究所环境光学与技术重点实验室, 安徽 合肥 230031
摘要
傅里叶变换红外光谱技术(FTIR)在水泥生料成分的在线分析上具有巨大的潜力。 但因现场环境复杂, 空气湿度不稳定, 会对生料样品中Fe2O3, SiO2, CaO和Al2O3四种关键成分的在线FTIR定量分析形成一定干扰。 使用生料在线FTIR分析仪对不同湿度条件下的水泥生料样品进行了近红外光谱采集, 分析了不同湿度对近红外光谱定量分析的影响, 并提出一种消除背景水分吸收的方法。 具体研究内容为: (1) 通过对两种不同湿度条件下的各50个样品的光谱分析得到: 高湿度的样品光谱与低湿度的样品光谱比较, 形状类似, 但吸光度整体降低, 基线倾斜。 表明背景水分影响了样品的近红外光谱。 (2) 分别建立高湿度、 低湿度条件下的样品的定量分析模型, 预测另一湿度条件下的预测集中8个样品的四种成分含量。 得到: ①高湿度模型预测样品中4种成分含量与标准值之间的相关系数(r)为83.74%~92.74%, 均方根误差(RMSE)为0.12~0.83; ②低湿度模型预测的r为67.32%~82.41%, RMSE为0.12~0.84。 表明背景水分影响了水泥生料成分的FTIR定量分析。 (3) 为了消除背景水分造成的影响, 从实测光谱中消除背景水分的特征吸收后, 分别建立了高湿度、 低湿度条件下的样品的FTIR定量分析模型, 并对预测集样品的四种成分含量进行预测。 得到: ①高湿度条件下, 消除背景水分后的模型较未消除前的模型预测的准确度提高, 预测的r为90.73%~97.76%, RMSE为0.12~0.82; ②低湿度条件下, 消除背景水分后的模型较未消除前的模型预测的准确度提高, 预测的r为94.07%~98.69%, RMSE为0.12~0.82; ③高湿度、 低湿度条件下, 消除背景水分后的2个模型预测的r均达到90%以上。 表明了该方法可有效消除背景水分对水泥生料成分定量分析模型预测的影响, 为实现基于FTIR的水泥生料成分的在线分析提供了理论基础和技术支持。
Abstract
Fourier transform infrared (FTIR) spectroscopy has great potential for on-line analysis of cement raw meal components. As the air humidity on site is not stable due to the complex environment, it will cause interference to the on-line FTIR quantitative analysis of the four key components of Fe2O3, SiO2, CaO, Al2O3 in the raw material samples. In this paper, the on-line FTIR analyzer for raw meals was used to collect near-infrared spectra of raw meal cement samples under different humidity conditions. The influences of different humidity conditions on near-infrared quantitative analysis were analyzed, and a method of eliminating the background moisture interference was proposed. The specific researches were as follows: (1) Spectra of each 50 samples at two different humidity levels were analyzed. The results were that sample spectra at high humidity level compared to that at low humidity level were similar in shape, while the absorbance intensities were deceased overall and baselines were inclined. These demonstrated that background moisture affected the near-infrared spectra of the samples. (2) Two FTIR quantitative analysis models for samples under high humidity and low humidity conditions were established respectively, and the four component contents of 8 samples in prediction set under another humidity condition were predicted. The results were that the values of the correlation coefficient (r) between the content values of the four components predicted by model under high humidity condition and the standard values in the prediction set were 83.74%~92.74%, and the values of the root mean square error (RMSE) were 0.12~0.83. The values of R obtained by model under low humidity condition were 67.32%~82.41%, and the values of RMSE were 0.12~0.84. These indicated that background moisture had affected the FTIR quantitative analysis of raw meal cement components. (3) In order to eliminate the influence of water absorption, the characteristic absorption of background moisture from the measured spectrum were removed refer to the mid-infrared spectroscopy technique. The FTIR quantitative analysis models under high humidity and low humidity conditions were established respectively, and the four components contents of samples in prediction set were predicted by these models. The results were as follows: ① Under high humidity condition, the prediction accuracy of the model with eliminating moisture absorption was improved compared with model without eliminating moisture absorption, the predicted values of r were 90.73%~97.76%, and the values of RMSE were 0.12~0.82, ② Under low humidity condition, the prediction accuracy of model with eliminating moisture absorption was higher than that of model without eliminating moisture absorption, and the predicted values of r were 94.07%~98.69%, the values of RMSE were 0.12~0.82, ③ The values of r obtained by models under high and low humidity conditions were above 90%. The experimental results showed that the method could effectively eliminate the influence of moisture absorption on the quantitative analysis model of raw material cement compositions. It provided the theoretical basis and technical support for the online analysis of raw material cement compositions based on FTIR technology.

胡荣, 刘文清, 徐亮, 金岭, 杨伟锋, 沈先春, 成潇潇, 王钰豪, 胡凯, 刘建国. 基于背景水分扣除的水泥生料成分的近红外光谱建模[J]. 光谱学与光谱分析, 2020, 40(4): 1051. HU Rong, LIU Wen-qing, XU Liang, JIN Ling, YANG Wei-feng, SHEN Xian-chun, CHEGN Xiao-xiao, WANG Yu-hao, HU Kai, LIU Jian-guo. Near Infrared Spectroscopic Modeling Method for Cement Raw Meal Components by Eliminating Background Moisture[J]. Spectroscopy and Spectral Analysis, 2020, 40(4): 1051.

本文已被 2 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!