2025 Theses Doctoral

Damage detection, damage localization, and fatigue life prediction for large-area FRP composites

Demo, Luke Benjamin

In this work, a novel self-sensing technology is introduced for fiber-reinforced polymer (FRP) composites, offering an accurate and cost-effective solution for damage detection, localization, and fatigue life prediction. This dissertation implements an innovative approach, transforming structural carbon fiber tows into piezoresistive sensors that enable real-time structural health monitoring (SHM) without the need for additional sensor devices. The self-damage detection and memory (SDDM) hybrid composite material leverages the carbon fiber as a sensor network, with glass fiber providing electrical insulation. Damage detection capabilities are first introduced, demonstrated by tensile testing that revealed two distinct loading peaks and a sharp nonlinear increase in resistance at the point of carbon fiber failure, highlighting its capabilities for damage early warning. Progressive impact tests further confirmed the material's ability to permanently record microdamage, showcasing a self-memory function that could inform life-cycle predictions.

Next, a practical sensor layout was developed, utilizing carbon fiber sensor tow branches connected in parallel each with varying resistances. This novel design can monitor large areas while minimizing the number of connections required to the DAQ circuit, significantly reducing manufacturing costs and complexities. Impact tests on carbon and glass fiber-reinforced composites validated the system’s ability to detect and precisely locate damage, with less than three percent error between the measured resistance and predicted damage location. These results highlight the effectiveness of the proposed damage localization framework, offering an efficient SHM solution for large-area composite structures.

Lastly, this dissertation introduces a low-cost, real-time fatigue life prediction system that leverages the piezoresistive cumulative damage behavior of carbon fiber sensor tows. A Bidirectional Long Short-Term Memory (LSTM) neural network was implemented to predict fatigue life based solely on resistance time-history, with no need for explicit stress inputs. Fatigue tests conducted across various stress amplitudes were used to train and evaluate a LSTM model, with results indicating the model’s ability to accurately predict remaining life. Moreover, testing results showed a sharp increase in resistance before failure, demonstrating the carbon fiber sensor tow's damage early warning capabilities for both cyclic and quasistatic monotonic loading. This system presents a promising, cost-effective SHM method that not only ensures structural safety but also extends the service life of FRP composites through accurate fatigue life prediction.