Predicting Crack Growth in Multiple Degradation Experiment with Dynamic Bayesian Network

Abstract

Ships and marine structures are subjected to fatigue damage from repeated fluctuating loads. Predicting crack growth is important for increasing structural safety and service life, however few approaches currently deal with multiple cracks in a complex structural system. To mimic many of the properties of larger and more complex marine structures in a laboratory-level experiment, a hexagon tension specimen with four propagating fatigue cracks on each corner is designed and tested. This specimen replicates the properties of a marine structure with redundant load paths, crack interaction and component to system level integration. The applied loading cycles and corresponding crack lengths are recorded as major time-varying data of degradation state which is used to evaluate the performance of a dynamic Bayesian network for predicting crack growth. The dynamic Bayesian network models the time-varying process with sequential slices. The dependence among components are controlled by hyperparameters and are integrated into complex system behavior to reflect the structure from components level to system level. The ability of network to predict the experiment is assessed.