基于自回归模型的超细颗粒动态光散射模拟

为了获取超细颗粒动态散射光模拟信号, 在分析超细颗粒动态散射光信号特性的基础上, 通过建立动态光散射随机过程的自回归(AR)模型, 利用Levison-Durbin递推算法确定模型参数, 并给出了单峰、双峰分布颗粒信号模拟的模型阶数确定方法, 从而提出了一种基于AR模型的态光散射信号模拟方法。分别对50 nm, 300 nm, 1000 nm, 50 nm与1000 nm, 100 nm与500 nm, 300 nm与1000 nm的单峰、双峰分布颗粒在模型阶数分别为1, 1, 1, 57, 28, 40时进行了模拟, 得到的模拟信号的光强自相关函数与理论值吻合, 用累积法对单峰分布颗粒反演和双指数法对双峰分布颗粒反演, 相对误差分别小于0.58%和3.7%, 因此, 单峰分布颗粒信号模拟需一阶模型, 双峰分布颗粒信号模拟粒径不同所需阶数不同。

Simulation of Superfine Particles Dynamic Light Scattering Based on Auto-Regressive Module

In order to simulate the dynamic light scattering signal of superfine particles, an auto-regressive (AR) model of the random process is made by analysing the character of dynamic light scattering signal. The module parameter is identified by Levison-Durbin algorithm and the method of selecting module order is given when scattering signal of unimodal and bimodal distribution particles is simulated. Thus a simulative method of dynamic light scattering signal based on AR module is put forward. By this method, the scattered signal of 50 nm, 300 nm, 1000 nm and scattered signal of 50 nm and 1000 nm, 100 nm and 500 nm, 300 nm and 1000 nm were respectively simulated when the module order was 1, 1, 1, 57, 28, 40. Their simulative autocorrelation function of light intensity accords with their theoretical autocorrelation function. The unimodal distribution’s relative error of inverting particles size by cumulants and the bimodal distribution’s relative error of inverting particles size by double exponential were respectively less than 0.58%, 3.7%. The results show that the signal simulation of unimodal distribution particles need one order module, and the signal simulation of bimodal distribution particles with differernt sizes need differernt order module.