Performance Comparison of Improved Particle Filters for On-line Fatigue Crack Prognosis

Abstract

Fatigue crack growth prognosis is a fundamental task for ensuring structural integrity. Recently, attentions have been gradually paid to methods that combine on-line measurements of structural health monitoring (SHM) with the particle filter based prognostics. However, most studies used the basic particle filter algorithm, which has the intrinsic particle impoverishment problem. There are different kinds of improvements made to the basic particle filter in the field of particle filtering. Few studies have been carried out to discuss their applicability to on-line fatigue crack growth prognosis combining with the SHM technique. Therefore, this paper compares four different improved particle filter algorithms for the on-line application of fatigue crack growth prognosis by integrating the guide wave based SHM. Studies are carried out on the basis of fatigue tests performed on the attachment lug, which is a key structural element in engineering structures. Two cases are involved that the SHM measurement has relatively large or small errors, which are influenced by whether the measurement equation is accurately trained with historical data. The prognostic accuracy of these improved particle filters is discussed. Moreover, effects of the particle number on the performance and computational cost are analyzed.