Data-Driven Structural Health Monitoring Approach Using Guided Lamb Wave Responses

In this paper, a data-driven structural health monitoring (SHM) approach is proposed to conduct in situ evaluation of the structural health state, i.e., damage location and extent, in which guided Lamb wave responses at selected locations are employed. The proposed approach is composed of an offline and an online phase. The objectives of the offline phase are to carry out data dimensionality reduction and to establish the mapping relationship between sensor data and damage status. First, a comprehensive database is established via high-fidelity finite element method (FEM) simulations (ABAQUS software) to determine guided Lamb wave responses (e.g., displacement and acceleration) under various prescribed structural damage conditions. Then, the proper orthogonal decomposition (POD) method is applied to extract key features from these responses under each simulated case. Finally, a neural network-based surrogate model is developed to relate the damage status with modal coefficients of the POD. The goal of the online phase is to quantify the damage location and extent using limited sensor measurements. The gappy proper orthogonal decomposition (GPOD) is employed to reconstruct the full field information based on limited sensor data. Subsequently, the associated damage extent and location are derived by applying the surrogate model developed in the offline phase. The proposed data-driven SHM approach is comprehensively validated using simulation data harvested from both beam and plate examples. The maximum error between evaluated and actual damage values is within 10%. Parametric studies are conducted as well to investigate the effects on damage detection using different sensor placement and sensor types. In summary, the proposed approach could lead to an efficient damage detection technique for aerospace structures.