Robust Inversion Methods for Aerosol Spectroscopy

TL;DR

This paper introduces robust inversion algorithms for aerosol spectroscopy measurements from the FASP device, including Bayesian model selection and multi-component aerosol reconstruction, validated through simulations.

Contribution

It extends existing regularization methods with uniqueness criteria and develops a novel multi-component aerosol inversion algorithm.

Findings

Algorithms outperform classical methods in simulations

Extended model successfully retrieves multi-component aerosol distributions

Numerical results demonstrate robustness and accuracy

Abstract

The Fast Aerosol Spectrometer (FASP) is a device for spectral aerosol measurements. Its purpose is to safely monitor the atmosphere inside a reactor containment. First we describe the FASP and explain its basic physical laws. Then we introduce our reconstruction methods for aerosol particle size distributions designed for the FASP. We extend known existence results for constrained Tikhonov regularization by uniqueness criteria and use those to generate reasonable models for the size distributions. We apply a Bayesian model-selection framework on these pre-generated models. We compare our algorithm with classical inversion methods using simulated measurements. We then extend our reconstruction algorithm for two-component aerosols, so that we can simultaneously retrieve their particle-size distributions and unknown volume fractions of their two components. Finally we present the results of a numerical study for the extended algorithm.