Stochastic modeling of fatigue crack damage for information-based maintenance

The concept of information-based maintenance is that of updating decisions for inspection, repair, and maintenance scheduling based on evolving knowledge of operation history and anticipated usage of the machinery as well as the physics and dynamics of material degradation in critical components. This paper presents a stochastic model of fatigue crack damage in metallic structures for application to information-based maintenance of operating machinery. The information on operation history allows the stochastic model to predict the current state of damage, and the information on anticipated usage of the machinery facilitates forecasting the remaining service life based on the stress level to which the critical components are likely to be subjected. The Karhunen–Loève expansion for nonstationary processes is utilized for formulating the stochastic model which generates the crack length statistics in the setting of a two-parameter lognormal distribution. Hypothesis tests are built upon the (conditional) probability density function of crack damage that does not require the solution of stochastic differential equations in either Wiener integral or Itô integral settings. Consequently, structural damage and remaining life of stressed components can be predicted to make maintenance decisions in real time. The damage model is verified by comparison with experimental data of fatigue crack statistics for 2024-T3 and 7075-T6 aluminium alloys. Examples are presented to demonstrate how this concept can be applied to hypothesis testing and remaining life prediction.