欢迎访问《兵工学报》官方网站,今天是

兵工学报 ›› 2020, Vol. 41 ›› Issue (5): 1034-1040.doi: 10.3969/j.issn.1000-1093.2020.05.023

• 研究简报 • 上一篇    

基于近红外光谱的梯恩梯凝固点温度检测方法研究

高嘉明1, 佘冲冲1, 陈军2, 李敏2, 侯云辉2, 陈丽珍1, 王建龙1   

  1. (1.中北大学 化学工程与技术学院, 山西 太原 030051; 2.湖北东方化工有限公司, 湖北 襄阳 441404)
  • 收稿日期:2019-05-27 修回日期:2019-05-27 上线日期:2020-07-17
  • 作者简介:高嘉明(1995—),男,硕士研究生。E-mail:18434367844@163.com
  • 基金资助:
    火炸药青年基金项目(QKCZ201705)

The Detection of Solidifying Point Temperature of TNT Using Fourier Transform Near-infrared Spectroscopy

GAO Jiaming1, SHE Chongchong1, CHEN Jun2, LI Min2, HOU Yunhui2, CHEN Lizhen1, WANG Jianlong1   

  1. (1.School of Chemical Engineering and Technology, North University of China, Taiyuan 030051, Shanxi, China; 2.Hubei Dongfang Chemical industry Co., Ltd., Xiangyang 441404, Hubei, China)
  • Received:2019-05-27 Revised:2019-05-27 Online:2020-07-17

摘要: 为实时检测梯恩梯(TNT)生产过程中的凝固点温度,结合化学计量学和光谱学,采用偏最小二乘法建立近红外光谱分析技术,实时检测TNT凝固点温度的定量模型。通过比较1阶导数+标准正态变量变换(SNV)、多元散射校正(MSC)、1阶导数、SNV、1阶导数+MSC、2阶导数6种原始光谱数据优化预处理方法,发现选择光谱建模区间为9 403.8~ 4 597.7 cm-1、1阶导数(17点平滑)+SNV预处理方法建立的模型最佳,模型相关系数R2=0.991,交互验证标准偏差为0.178. 主成分分析及模型验证结果表明:用最优模型预测不同机台的硝化物凝固点温度,近红外光谱分析预测值与人工测定值偏差最大为0.950 4%; 该模型有较好的稳定性、预测性,能够识别不同类型的样本,在短时间内用近红外光谱分析法可测定凝固点温度。

关键词: 梯恩梯, 凝固点温度, 近红外光谱, 偏最小二乘法, 定量分析, 主成分分析

Abstract: In order to detect the solidifying point temperature in TNT production process in real time, the partial least squares method is used to establish a quantitative model for real-time detection of TNT solidifying point by near-infrared spectroscopy combined with chemometrics and spectroscopy. A model established by the preprocessing method with the modeling interval of 9 403.8-4 597.7 cm-1 and the first-order derivative (17-point smoothing)+SNV is the best compared with six preprocessing methods optimized for the original spectral data. The six methods are standard normal variable transformation (SNV), multiple scattering correction (MSC), first derivative, first derivative + standard normal variable transformation (SNV), first derivative + multiple scattering correction (MSC), and second derivative. The correlation coefficient of the model is R2=0.991, and the root-mean-squares error of cross-validation is 0.178. The results of principal component analysis and model validation show that the maximum deviation between the predicted value of near infrared analysis and the measured value is 0.950 4% when the optimal model is used to predict the solidifying point temperature of nitrates on different samplers. The model has good stability and predictability, and can identify different types of samplers. The solidifying point can be determined by near infrared spectroscopy in a short time. Key

Key words: TNT, solidifyingpointtemperature, near-infraredspectroscopy, partialleastsquares, quantitativeanalysis, principalcomponentanalysis

中图分类号: