Atmospheric Corrosion Prognostics Using a Particle Filter

Abstract

A prognostic algorithm can guarantee high reliability and availability of machinery of structures at acceptable costs. This paper proposes to use a particle filter for atmospheric corrosion prognostics, which bridges the gap between corrosion modeling and corrosion monitoring. The applied particle filter only takes temperature and monitored mass loss as input and is based on Arrhenius equation. The output of the particle filter is a probability distribution of the remaining useful life that considers uncertainties on the process, the model and future weather conditions. The effectiveness of the approach is demonstrated by a case study composed from monthly exposure tests performed by the National Institute of Materials Science in Japan. It is shown that the particle filter estimates suitable model parameters of the corrosion model to give good remaining useful life estimations, while only requiring a relatively simple corrosion model. In new practical applications challenges remain in parameter selection and initialization of the algorithm. Furthermore, the method should be validated on an actual long-term corrosion process.