Particle Size Inversion of Dynamic Light Scattering Based on Improved Morozov's Deviation Principle
Dynamic Light Scattering(DLS)technology is an effective method to measure Particle Size Distribution(PSD),and this technology is widely used in chemistry,medical treatment,materials and other fields.DLS technology obtains the Autocorrelation Function(ACF)of the scattered light intensity signal by the autocorrelation operation of scattered light,and obtains the PSD by inverting the ACF of the light intensity.Inverting the ACF needs to solve the first kind of Fredholm integral equation which belongs to ill-conditioned problem,and this kind of equation can be solved by Tikhonov regularization.Tikhonov regularization controls the accuracy and stability of the solution by adjusting the regularization parameter,which is usually selected by L-curve criterion.L-curve criterion introduces the stability analysis through the vector modulus of the solution,but it can not get ideal inversion results under the condition of wide particle size distribution.Morozov's Discrepancy Principle(MDP),another method to select regularization parameter,can select parameter according to the noise level of electric field ACF.But the raw data noise level is usually unknown,which causes it difficult to apply MDP in actual measurement.Compared with the inversion of narrow particle size distribution,the inversion of wide particle size distribution needs more accurate regularization parameters to ensure the accuracy of the inversion results.To improve the accuracy in the widely distributed particle system,we got the noise component of the electric field ACF through wavelet packet decomposition,and measured the amplitude of the noise component to estimate the noise level.The noise level was brought into MDP to establish the fitness function,the initial population was generated within the empirical range of the regular parameters,and then the fitness function and the initial population were brought into the Genetic Algorithm(GA)to find the regular parameters.The method proposed above is named as MDP-GA.Thus,the problem of MDP in actual measurement is solved.Compare the MDP-GA method with the L-curve criterion under simulated and measured conditions.In comparison,different distributions under different noise levels are selected,and take peak error,distribution error and repeatability error as performance indicators.The inversion results show that there is no significant difference between MDP-GA and L-curve criterion under the condition of narrow particle distribution,at the same time,the inversion results of MDP-GA method are better than those of L-curve criterion under the condition of wide distribution.Under the condition of wide particle size distribution,the false peak and peak offset that may occured in L-curve criterion are avoided by the MDP-GA method.
NETL
NSTL
万方数据
