Mixture Proposal Particle Filtering for Guided Wave Based Fatigue Crack Propagation Prognosis

Abstract

For on-line prognosis of fatigue crack propagation with guided wave based structural health morning, uncertainties always exist in the process of fatigue crack propagation and crack monitoring. The particle filtering is demonstrated as a powerful tool to deal with these uncertainties. However, most literatures adopt the standard sequential importance sampling resampling algorithm, which uses the transition probability density function as the importance density and neglects current measurement information. Satisfactory results may be unable to be achieved if particles are drawn from an inferior transition probability density. This paper proposes a kind of mixture proposal distribution as the importance density that linearly combines the measurement probability density with the transition probability density. In addition, the damage index of the on-line monitored guided wave signal is adopted for quantifying the crack length. Each time a new damage index is available, the mixture proposal particle filtering is adopted to evaluate the posterior probability density estimate of the crack length, as well as the crack propagation prognosis. Simulated crack propagation data is used for validation, and the result shows that the mixture proposal particle filtering outperforms the standard particle filtering algorithm.