近红外全局隐含温度补偿模型的预测精度分析

发布时间：2018-06-28 06:31

本文选题：近红外光谱 + 温度补偿　；参考：《光谱学与光谱分析》2017年04期

【摘要】：温度波动影响含氢基团之间的作用力,从而影响近红外光谱的吸收强度和波峰位置等,导致近红外测量精度的降低。针对温度变化对近红外光谱建模精度的影响,对全局隐含温度补偿方法进行了研究,并对其预测精度进行了分析,分别从预测方差和置信区间两个方面对此类模型的精度进行了理论探讨和验证。同时通过温度的连续变化实验,即在温度连续变化的过程中,等时间间隔采集各样品的近红外光谱,研究了温度变化对光谱主元的连续模式影响,探讨了温度变化影响模型预测精度的方式和途径。最后对某高分子聚合物的粘度测量问题进行了实验验证和误差分析,得到标准温度下所建未经温度补偿的模型和全局隐含温度补偿模型的建模精度分别为:RMSEC=0.243 0,R_c=0.871 6,RMSEP=0.243 2,R_p=0.869 3;RMSEC=0.258 2,R_c=0.870 6,RMSEP=0.265 2,R_p=0.856 0,而当温度变化时,二者预测最大置信区间分别约为1.8和0.9kPa·s。虽然全局隐含温度补偿模型相比于标准温度模型建模精度略降低,但预测精度提高了一倍左右。理论分析和实验结果均表明,全局温度补偿模型具有较高的预测精度,且对温度的变化有较强的鲁棒性和可靠性。
[Abstract]:The temperature fluctuation affects the interaction force between the hydrogen groups, thus affecting the absorption intensity and peak position of the near infrared spectrum, which results in the decrease of the precision of the near infrared measurement. Aiming at the influence of temperature variation on the modeling accuracy of NIR spectrum, the global implicit temperature compensation method is studied, and its prediction accuracy is analyzed. The accuracy of this model is theoretically discussed and verified from two aspects of prediction variance and confidence interval. At the same time, the influence of temperature change on the continuous mode of spectral principal component is studied by the experiment of continuous temperature change, that is, in the process of continuous temperature change, the near-infrared spectra of each sample are collected at equal time intervals. The ways and means of the influence of temperature change on the prediction accuracy of the model are discussed. Finally, the viscosity measurement of a polymer is verified by experiments and error analysis. The modeling accuracy of the model without temperature compensation and the global implicit temperature compensation model established at the standard temperature is: RMSEC / 0.243 0 / RCX 0.871 6 / RMSEPN 0.243 / 2 / RMSEP 0.243 / 69 / 3 / RMSEC / 0.258 / 2 / RMSECP / 0.265 / 0.265 / 0.56 / 0.56, respectively. When the temperature changes, the predicted maximum confidence intervals of the two models are about 1.8 and 0.9ksPa, respectively. The results show that the maximum confidence ranges of the two models are about 1.8 and 0.9ksPa, respectively, when the temperature varies, the predicted maximum confidence ranges of the two models are about 1.8 and 0.9ksPa, respectively. Although the global implicit temperature compensation model is slightly less accurate than the standard temperature model, the prediction accuracy is about twice as high as that of the standard temperature model. The theoretical analysis and experimental results show that the global temperature compensation model has high prediction accuracy and strong robustness and reliability to the temperature variation.
【作者单位】：江南大学自动化研究所轻工过程先进控制教育部重点实验室;
【基金】：国家自然科学基金项目(61473137)资助
【分类号】：O657.33

【相似文献】