Diagnostics and prognostics of multi-mode failure scenarios in miter gates using multiple data sources and a dynamic Bayesian network

Abstract

Current health monitoring approaches for large structures mostly rely on a combination of distributed sensor networks and in-situ inspection. This paper presents a novel online diagnostics and prognostics framework for structures subject to multiple failure modes and demonstrates the proposed method with a high-fidelity finite element model using multiple data sources (i.e., strain gauges and images). The approach aims at an accurate simulation of the interaction between different failure features, and subsequently at the effective estimation and prediction of the damage states based on the generated structural physics. A dynamic Bayesian network is used which incorporates different data sources to evaluate the structures under different kinds of deterioration mechanisms. In diagnosis, the dynamic Bayesian network is used to approximate the damage-related parameters and estimate the timedependent variables. In prognosis, the dynamic Bayesian network gives a probabilistic prediction of the remaining useful life of the structure based on the evolution of the failures. It is found that the proposed framework is highly effective in performing online diagnosis and prognosis using combined data sources.