动态光散射颗粒分布软测量

发布时间：2018-04-03 14:14

本文选题：动态光散射　切入点：颗粒分布测量　出处：《光学精密工程》2016年11期

【摘要】：考虑传统动态光散射颗粒粒度分布测量用的反演算法复杂、精度不够、抗噪能力差,本文基于大数据思想,提出了一种动态光散射颗粒分布软测量方法。该方法通过调节颗粒粒度分布形状参数获得大量自相关函数及其对应颗粒分布的数据;使用这些数据对子学习机进行训练。最后,针对训练数据维数较高的特点对传统Bagging算法进行改进,并利用改进的Bagging集成算法集成子学习机以提高软测量模型的精度及泛化能力。通过模拟单峰数据和对300nm标准粒径进行软测量开展了验证实验。结果表明,该方法能够较好地测量出不同动态光散射颗粒分布的峰值及分布宽度,模拟单峰数据测量峰值精度可达1nm,300nm和503nm,标准粒径测量精度分别可达3nm和4nm,优于一般的反演算法。该软测量方法为动态光散射颗粒分布测量开辟了新的途径。
[Abstract]:Considering that the traditional inversion algorithm for particle size distribution measurement of dynamic light scattering is complex, the accuracy is not enough, and the anti-noise ability is poor, a soft sensing method for dynamic light scattering particle distribution is proposed based on big data's idea.In this method, a large number of autocorrelation functions and their corresponding particle distribution data are obtained by adjusting the shape parameters of particle size distribution, and these data are used to train the sub-learning machine.Finally, aiming at the high dimension of training data, the traditional Bagging algorithm is improved, and the improved Bagging integration algorithm is used to integrate the sublearning machine to improve the accuracy and generalization ability of the soft sensor model.The verification experiments were carried out by simulating single peak data and soft-sensing 300nm standard particle size.The results show that this method can measure the peak value and the distribution width of different dynamic light scattering particles.The peak accuracy of simulated single-peak data can reach 300nm and 503nm, and the precision of standard particle size measurement can reach 3nm and 4nmrespectively, which is superior to the general inversion algorithm.This soft sensing method opens up a new way for measuring the distribution of dynamic light scattering particles.
【作者单位】：天津工业大学电气工程与自动化学院;天津工业大学电工电能新技术天津重点实验室;
【基金】：国家自然科学基金资助项目(No.61403277;No.71602143)
【分类号】：TP18;O436.2

【相似文献】