Particle filter-based hybrid damage prognosis considering measurement bias

Hybrid methods combining physical knowledge and data-driven techniques have shown great potential for damage prognosis in structural health monitoring (SHM). Current practices consider the physics-based process and data-driven measurement equations to describe the damage evolution and the mapping between the damage state and the measurement, respectively. However, the bias between the measurements predicted by the data-driven equation and the sensor-based measurements obtained from any SHM system, arising from uncertainties like damage geometries and sensor placement or noise, can lead to inaccurate prognosis results. To improve the prognosis performance in case of bias, this paper adopts a methodology typically applied in sensor fault diagnosis and develops a new hybrid prognostic model with a bias parameter included in the measurement equation and the state vector. Particle filter (PF) serves as the estimation technique to identify the state and parameters relating to the damage as well as the bias parameter, and the remaining useful life (RUL) can be predicted by the physics-based process equation, with PF posterior estimates of the related parameters and state variables as an input. The experimental study of an aluminum lug structure subject to fatigue crack growth and equipped with a Lamb wave monitoring system demonstrates the improved estimation and prediction performances of the new prognostic model.

Automated summary

The study utilized hybrid methods combining physical knowledge and data-driven techniques for structural health monitoring, with improved prediction performances based on the physics-based process equation demonstrated in experimental study.