含噪动态光散射测量数据反演中正则化算法与Chahine算法的比较

Comparison between the Regularization Algorithm and the Chahine Algorithm in Inversions of Scattering Measurement Data of the Noisy Dynamic Light

采用两种常用的粒度反演方法——正则化和Chahine算法,对90nm与250nm单峰分布、50nm与200nm双峰分布、100nm与300nm双峰分布的模拟动态光散射数据,以及105nm、300nm标准颗粒的实测动态光散射数据进行了反演分析.结果表明:噪声水平的高低是影响粒度分布反演准确性的关键因素之一,反演结果的准确性随噪声水平的增加而降低,噪声水平超过某一阈值后,将无法得到有意义的反演结果;不同反演方法具有不同的抗噪能力,在低噪声水平下反演结果无显著差别,随着噪声水平的增加,反演结果表现出很大差异;正则化方法通过正则参数的选择可以有效抑制噪声影响,表现出强于Chahine算法的抗噪能力;与Chahine算法相比,正则化方法不需要假定初始分布,因此,在噪声较大的实验或生产过程中进行颗粒分布测量时,宜采用正则化方法.

In this paper, two kinds of commonly used particle size inversion methods, regularization algorithm and Chahine algorithm, were used to inverse the simulated dynamic light scattering data of the unimodal distribution of 90nm and 250nm, the bimodal distribution of 50nm and 200 nm, and the bimodal distribution of 100 nm and 300 nm of the particles in submicron region, and measured the dynamic light scattering data of 105nm and 300nm particles, for comparing the noise effects of the two algorithms. The inversion results show that, the noise level is one of the key factor can restrict the accurate inversion for particle size measurement. The accuracy of inversion results decreases with the increase of the noise level, and when the noise level increase to a certain threshold value, the meaningful inversion results will not be obtained. Different inversion methods have different anti-noise ability, but there is no significant difference of the inversion results when noise level is very low. With the increase of the noise level, the retrieval results show a substantial difference, regularization method can effectively restrain the noise influence by the appropriate choice of the regularization parameter, which shows a better anti noise capability than Chahine algorithm. Compared with the Chahine algorithm, the regularization method, despite the need for regularization parameter setting, it does not need to assume the initial distribution. Therefore, it is recommended to use in noisv environment.