Dynamic Bayesian Network for Fatigue-Crack-Related Risk Assessment of Aerospace Bolts

Abstract

This paper provides a general process for applying dynamic Bayesian networks (DBNs) for the risk assessment of aerospace bolts, which are susceptible to fatigue cracks. The process includes modeling, discretization, verification, calibration, validation, and prediction. The uncertainties of the inspection capability are emphasized. Subsequently, the fatigue-crack growth of a bolt for a fighter jet is predicted, and the corresponding risk is calculated. A simple equation, i.e., Paris’s law, is used to calculate the crack growth over the stress cycles. Because the bolt is inspected regularly using nondestructive inspection methods, the probability of detection (POD) of the inspection is considered by adding the parameter of the POD curve in the DBN. The model is validated via comparison based on a Monte Carlo simulation. Calibration and validation are conducted successfully based on maintenance data.