光子学报
光子學報
광자학보
ACTA PHOTONICA SINICA
2009年
11期
2917-2926
,共10页
韩秋燕%申晋%孙贤明%刘伟%宋井玲
韓鞦燕%申晉%孫賢明%劉偉%宋井玲
한추연%신진%손현명%류위%송정령
光子相关光谱%Tikhonov正则化方法%Morozov偏差原理%后验选择策略
光子相關光譜%Tikhonov正則化方法%Morozov偏差原理%後驗選擇策略
광자상관광보%Tikhonov정칙화방법%Morozov편차원리%후험선택책략
Photon correlation spectroscopy%Tikhonov regularization method%Morozov discrepancy principle%Posterior choice strategies
采用基于Morozov偏差原理的后验策略来选择最优正则参量,并采用此方法对单峰和多峰分布颗粒系的模拟电场自相关函数进行了反演,结果表明,对于单峰颗粒体系,当电场自相关函数的扰动误差小于0.05时,反演得到的峰值准确,当电场自相关函数的扰动误差大于0.05时,反演得到的峰值偏离所模拟的颗粒粒径.正则参量初始值在0.000 02~2范围内,在反演所得的峰值准确的基础上,正则参量初始值越小,反演得到的分布宽度越窄.收敛误差在0.000 05~50范围内,在保持反演结果稳定的基础上,收敛误差取值越大,反演得到的分布宽度越窄.对于多峰颗粒体系,当颗粒系中的颗粒粒径差别较小时,峰值向平均值偏移,当颗粒系中的颗粒粒径差别较大时,小颗粒粒径分布以噪音的形式出现.
採用基于Morozov偏差原理的後驗策略來選擇最優正則參量,併採用此方法對單峰和多峰分佈顆粒繫的模擬電場自相關函數進行瞭反縯,結果錶明,對于單峰顆粒體繫,噹電場自相關函數的擾動誤差小于0.05時,反縯得到的峰值準確,噹電場自相關函數的擾動誤差大于0.05時,反縯得到的峰值偏離所模擬的顆粒粒徑.正則參量初始值在0.000 02~2範圍內,在反縯所得的峰值準確的基礎上,正則參量初始值越小,反縯得到的分佈寬度越窄.收斂誤差在0.000 05~50範圍內,在保持反縯結果穩定的基礎上,收斂誤差取值越大,反縯得到的分佈寬度越窄.對于多峰顆粒體繫,噹顆粒繫中的顆粒粒徑差彆較小時,峰值嚮平均值偏移,噹顆粒繫中的顆粒粒徑差彆較大時,小顆粒粒徑分佈以譟音的形式齣現.
채용기우Morozov편차원리적후험책략래선택최우정칙삼량,병채용차방법대단봉화다봉분포과립계적모의전장자상관함수진행료반연,결과표명,대우단봉과립체계,당전장자상관함수적우동오차소우0.05시,반연득도적봉치준학,당전장자상관함수적우동오차대우0.05시,반연득도적봉치편리소모의적과립립경.정칙삼량초시치재0.000 02~2범위내,재반연소득적봉치준학적기출상,정칙삼량초시치월소,반연득도적분포관도월착.수렴오차재0.000 05~50범위내,재보지반연결과은정적기출상,수렴오차취치월대,반연득도적분포관도월착.대우다봉과립체계,당과립계중적과립립경차별교소시,봉치향평균치편이,당과립계중적과립립경차별교대시,소과립립경분포이조음적형식출현.
A Posteriori Choice Strategies based on Morozov discrepancy principle is adopted in order to choose the Optimum Regularization Parameter in the Inverse Algorithm of the Photon Correlation Spectroscopy particle sizing techniques. Using the method analyzed the simulation experimental data of single peake and multimodel peak particles system are analyzed. For the single peake particles system,the peake value is correct when the noise of the experimental data varies from 0 to 0. 05. The peake value is incorrect when the noise of the experimental data is greater than 0.05. The distributing width decreases with the decrease of the original value of Regularization Parameter when the original value of Regularization Parameter varies from 0.000 02 to 2. The distributing width decreases with the increase of the convergence error when the convergence error varies from 0. 000 05 to 50. For the multimodel peak particles system, the peak value incline to the average value when the particles diameter discrepancy is small. The smaller particle size distributing appear as the noise when the diameter discrepancy is large.