Non-negative Tikhonov Regularization Inversion Combining Trust-region with Interior Point Newton for Photon Correlation Spectroscopy

摘要

In order to improve the accuracy of inversion Particle Size Distribution (PSD) in the Photon Correlation Spectroscopy (PCS) technology, considering non-negative characteristic of PSD, based on Tikhonov regularization method, two non-negative constraint methods of trust-region (Trust) and Interior Point Newton (IPN) are compared in this study. Combining characteristics of two methods, an inversion method of Trust-IPN-Tikhonov is proposed. This method inherits the advantages of the Trust-Tikhonov and IPN-Tikhonov. The inversion results of simulation data and experimental data demonstrate that Trust-IPN-Tikhonov has smaller peak error, relative error and narrower distribution width than IPN-Tikhonov, compared with the Trust-Tikhonov, Trust-IPN-Tikhonov has not only smaller peak error and relative error but also better smoothness. All in all, Trust-IPN-Tikhonov has higher accuracy, better smoothness and is more consistent with the true distribution.