基于近红外光谱智能反演的三羟甲基丙烷合成过程在线监测

Quality control of trimethylolpropane synthesis by intelligent retrieval of near-infrared spectra

本文发展了近红外光谱技术的智能反演方法,建立三羟甲基丙烷(TMP)生产过程中产品及其副产物的定量分析模型,实现其产物及其副产物含量的快速检测,同时定量监测TMP的工艺合成过程,构建TMP合成过程的在线监控系统.首先采用气相色谱法测定TMP合成过程中产品及其副产物含量,然后采集大量样品的近红外光谱图,最后调用TQ Analyst软件的偏最小二乘回归法(PLS)建立产品TMP、副产物双三羟甲基丙烷、2-乙基-2-己烯醛、2-乙基丙烯醛的4种智能预测模型.结果表明,所建立的TMP、双三羟甲基丙烷、2-乙基-2-己烯醛、2-乙基丙烯醛的4种定量分析模型的校正均方根误差(RMSEC)分别为0.744、0.300、1.500、1.280,相关系数(R)分别为0.9868、0.9840、0.9945、0.9113;预测均方根误差(RMSEP)分别为1.290、0.588、1.440、1.180,R分别为0.9797、0.9209、0.9960、0.9323.最终证实了所建立的TMP合成过程中产品及其副产物的近红外智能反演模型具有较好的预测精度与预测稳定性,可用于三羟甲基丙烷的合成过程的在线监测.

We developed an intelligent inversion method of near-infrared spectroscopy to establish a quantitative analysis model of the product and its by-products in the production process of trimethylolpropane(TMP),to realize the rapid detection of its product and its by-product content,to quantitatively monitor the process synthesis of TMP,and to build an online monitoring system of the TMP synthesis process.Firstly,gas chromatography was used to determine the content of the product and its by-products during the synthesis of TMP,and then the NIR spectra of a large number of samples were collected,and finally,the partial least squares regression(PLS)method of TQ Analyst software was invoked to establish four intelligent prediction models for the product TMP,the by-products bis-trimethylolpropane,2-ethyl-2-hexenal and 2-ethylacrolein.The results showed that the corrected root mean square errors(RMSE(C)of the four quantitative analytical models developed for TMP,bis-trimethylolpropane,2-ethyl-2-hexenal and 2-ethylacrylaldehyde were 0.744,0.300,1.50 and 1.28,respectively,and the correlation coefficients(R)were 0.9868,0.9840,0.9945 and 0.9113,respectively.The root mean square errors of prediction(RMSEP)were 1.29,0.588,1.44,and 1.18,respectively,and the R were 0.9797,0.9209,0.9960,and 0.9323.It was finally confirmed that the developed NIR intelligent inversion model of the products and their by-products in the synthesis process of TMP has good prediction accuracy and prediction stability,and can be used for online monitoring of the synthesis process of TMP.