Open Access Open Access  Restricted Access Subscription or Fee Access

Efficient Detection Methods on a Composite Plate with Interior Embedded Fiber Optic Sensors via Impact Test



Next-generation marine vessels are adopting composite materials for substructural elements, and due to a lack of knowledge about these materials’ failure modes, the need for monitoring them is substantial. In this study, a structural health monitoring system is implemented on composite plates with Fiber Bragg Grating strain sensors, internally embedded during material manufacturing, to monitor the effects of impact loads under ocean wave conditions. Taking advantage of the properties of FBG strain sensors, the composites plates were specifically designed to use this material to more easily examine the progression of impactinduced damage and delamination. The placement of the fiber optic sensors was done strategically in an asymmetric way to examine locations with the highest strain levels for multiple modes of vibrations, as determined by a finite element model of the composite plate. The failure scenarios were experimentally conducted on numerous plate specimens via dynamic impact simulation by dropping a hammer from various heights until fiber breakage (highly-localized damage). In each loading case, after each prescribed level of damage, the composite plate was subject to a vibration test. The main objective of the study is to examine the sensitivity of various feature extraction techniques from the FBG sensor data to both damage detection and damage localization. The first class of techniques compared involves basic signal statistics (mean and moments, peak statistics, higher-order moments). The second class involves modal property extraction, and the final class involves autoregressive models, which can be used both in a predictive and model-model comparative way. In all cases, comparisons will be made in a statistically-rigorous sense by employing receiver operating characteristics, since all tests were conducted in a supervised learning mode.

Full Text: