Statistical Performance Assessment of an NDE-Based SHM-DP Methodology for the Remaining Fatigue Life Prediction of Monitored Structural Components and Systems

Abstract

Integrated structural health monitoring and damage prognosis (SHM-DP) methodologies, coupled with sensor-based nondestructive evaluation (NDE) techniques, are becoming increasingly important for the near-real-time condition assessment (i.e., SHM) and future performance predictions (i.e., DP) of aging mechanical systems, civil structures, and infrastructure networks, as well as automotive, naval, and aerospace vehicles. A successful SHM-DP strategy, capable of identifying all critical damage mechanisms while accounting for all relevant sources of uncertainty, can be used as an advanced tool to effectively and optimally manage the life-cycle of the monitored system, recursively forecast its remaining useful life (RUL), and ultimately reduce the overall ownership cost through dynamic reliability-based inspection and maintenance (RBIM) plans, system downtime minimization, catastrophic failure prevention, and potential RUL extension. In this perspective, fatigue damage propagation is one of the most critical and unpredictable deterioration processes for a large variety of structural and mechanical systems that are subjected repeatedly to cyclic and/or random operational loading during their service life. Within this limited scope, the authors developed a comprehensive NDE-based SHM-DP framework for recursively predicting the time-varying system reliability and the remaining fatigue life (RFL) of monitored systems subjected to deterioration by multi-site fatigue damage propagation. This paper provides a brief overview of the proposed framework and then uses a set of experimental fatigue test data to perform a thorough statistical performance assessment of the developed methodology at the local reliability component level (i.e., single damage mechanism and single damage location) including NDE detectability and measurement uncertainty as well as both load and model parameter uncertainty.