Damage Detection in Fiber-Reinforced Composite Beams by Using a Bayesian Fusion Method

This paper presents a method for the detection of damage present in composite beam-type structures. The method, which successfully detected damage in steel beams, is applied to a glass fiber-reinforced beam in order to verify its suitability for composite structures as well. The damage indices were obtained using the gapped-smoothing method (GSM), which does not require a baseline model in order to detect damage. Despite the advantage of avoiding the need for a reference model altogether, unavoidable measurement errors make GSM rather ineffective. The proposed method uses the damage indices that GSM provides for synthesizing a set of likelihood functions that is processed under a Bayesian approach in order to reduce the effect of the noise and other uncertainty sources. The quality of the damage detection was examined by investigating an optimal sampling size analytically, and it was demonstrated through numerical simulation. This paper details the theory of the noise suppression method based on Bayesian data fusion, includes an analysis of the optimal sampling size, and presents the experimental results for two glass fiber-reinforced composite beams with a narrow and wide delamination, respectively. A noise-addition process was applied to the simulated data considering two different noise distributions. The composite beam was modeled in ANSYS, and harmonic analysis was used to obtain the frequency response functions at different beam locations. The results were obtained by adding 5, 10, and 15% noise in the simulated data, and they were then validated from the experimental results.