卷烟滤棒中三醋酸甘油酯含量的现场快速检测研究

On-site Rapid Determination of Content of Glycerol Triacetate in Cigarette Filter

为了提高卷烟滤棒中三醋酸甘油酯含量的检测效率,该文通过使用手持式近红外光谱仪,结合粒子群优化-极限学习机(PSO-ELM)回归算法建立了三醋酸甘油酯含量的定量预测模型,并与偏最小二乘回归(PLSR)和极限学习机回归(ELMR)进行了比较。实验结果表明:相比于PLSR和ELMR模型,所建立的PSOELM预测模型的决定系数R^(2)为0.9212,远高于PLSR预测模型的0.8604和ELMR预测模型的0.8772;同时,使用PSO-ELM模型的预测均方根误差(RMSEP)为0.39212,小于PLSR预测模型的0.49772和ELMR预测模型的0.47018。以上实验结果表明,所建立的近红外光谱定量模型能够应用于卷烟滤棒中三醋酸甘油酯含量的快速准确测量,为实现滤棒中三醋酸甘油酯含量的现场快速检测提供了良好的技术参考。

Glycerol triacetate is used as the plasticizer and hardener in molding process of cigarette filter,its content will not only affect the rigidity,resistance and filtration performance of a cigarette filter,but also have an impact on the perceived quality of cigarette smoke.In order to improve the efficiency for detection of glycerol triacetate content,a novel hand-held near-infrared spectroscopy(NIR)technique combined with particle swarm optimization-extreme learning machine(PSOELM)algorithm was applied to build a prediction model for glycerol triacetate content.Meanwhile,it was also compared with the traditional partial least squares(PLS)model and extreme learning machine(ELM)model.The experimental results showed that compared with PLSR model and ELMR model,the PSO-ELM model was much better with a determination coefficient(r^(2))of 0.9212,which was higher than 0.8604 for PLSR prediction model and 0.8772 for ELMR prediction model.Besides,the root mean square error of prediction(RMSEP)for PSO-ELM model was 0.39212,which was smaller than 0.49772 for PLSR model and 0.47018 for ELMR model.The above results indicated that the established model could be applied to the rapid and accurate determination of the content of glycerol triacetate in cigarette filter,which provided a good reference for on-site rapid determination on the content of glycerol triacetate.