Parameter estimation from big data using a sequential monte carlo sampler



Green, PL and Maskell, S ORCID: 0000-0003-1917-2913
(2016) Parameter estimation from big data using a sequential monte carlo sampler. In: ISMA2016.

[img] Text
ISMA_2016.pdf - Author Accepted Manuscript

Download (139kB)

Abstract

It is now well established that, through the use of sequential Monte Carlo methods, it is possible to track the time-varying state of mechanical systems 'online', using a continous stream of measurements. The best known of these algorithms is the particle filter-a numerical algorithm that can be applied to a large variety of nonlinear problems and which, in recent years, has been used to aid the condition monitoring of many mechanical systems. In this paper, a Sequential Monte Carlo method is used to estimate the parameters of a model from a continuous stream of measurements, with the aim of establishing how one's parameter estimates converge as more data is analysed. Crucially, for reasons described in this paper, this is a situation where a particle filter is unsuitable. The issue is instead resolved using a variant of a Sequential Monte Carlo sampler. It is shown how the algorithm can be used to identify the parameters of a model from large data sets and, within the context of structural dynamics, it is compared with the performance of a similar, Markov chain Monte Carlo method.

Item Type: Conference or Workshop Item (Unspecified)
Depositing User: Symplectic Admin
Date Deposited: 07 Oct 2016 13:33
Last Modified: 19 Jan 2023 07:29
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3003664