Resilience assessment approach of mechanical structure combining finite element models and dynamic Bayesian networks

Resilience notionally means the ability to adapt changing conditions and recover rapidly from disruptions, which is vital for mechanical structure. Structure failure is usually caused by sudden changes of the fatigue mechanical property. Mechanical properties of structures should be considered when assessing resilience. The work proposes a general resilience assessment approach for mechanical structure through combining finite element models and dynamic Bayesian networks (DBNs). Resilience assessment process is divided into two parts, namely degradation process and recovery process. Degradation states in different time points can be analyzed by the finite element model, which can further provide data when establishing the DBN model of the degradation process. Recovery process is composed of fault diagnosis, resource allocation and maintenance. Fault diagnosis capability and resource allocation capability are calculated as quantitative coefficients, which can influence the maintenance activity. The maintenance capability is simulated by a DBN model through physical model mapping. The DBN model for the recovery process is finally established by integrating the quantitative coefficients and the maintenance model. Subsea wellhead connector attacked by the internal wave is adopted to demonstrate the application of the proposed assessment approach.