Hybrid resampling scheme for particle filter-based inversion

Abstract

A novel online hybrid resampling (HR) scheme based on the combination of residual and multinomial resampling schemes is proposed. It can be implemented within the framework of the sequential Monte Carlo-based inversion algorithm, also known as particle filter (PF). Based upon the degeneracy of each particle, the choice of best resampling scheme among both candidates is made at each instant iteratively. Consequently, the inversion performance of PF improves by reducing the computational complexity of the resampling scheme. The proposed online HR scheme is incorporated here within the framework of the PF-based inversion scheme once applied on nuclear power plant steam generator non-destructive testing measurement data. The promising results showcase the efficacy of the technique.