Active Sensing and Damage Classification for Wave Energy Converter Structural Composites

Abstract

Ocean resources have the potential to provide a large source of renewable energy for communities around the globe. Technologies such as wave energy converters must be designed to operate remotely in harsh environmental conditions. These structures are exposed to widely varying structural loads, and there is interest in developing monitoring systems that can identify the presence of damage, estimate its severity, and provide maintenance or control recommendations that could protect the system from failure. The research presented in this paper focuses on using the electromechanical impedance response of piezoelectric transducers to monitor the health of composite materials similar to those used in the fabrication of several wave energy converters. Techniques have been developed to detect and classify discrete damage events such as holes and slots within composite plates, as well as fatigue damage that evolves due to manufacturing flaws such as delamination and laminate waves. Using data collected over a frequency range of 100 Hz to 100 kHz, a series of genetic algorithms and statistical modeling techniques were used to classify damage type and severity. Plate studies with discrete damage (holes, notches) provided a large dataset of 113 observations comprised of seven distinct classes, one baseline and six damage severities. Random forest techniques were used to classify this population, with accuracies of 93.4% obtained. Fatigue studies of rectangular composite beams containing manufacturing defects (delamination, laminate waves), produced a measurement population of 14 instances comprised of six distinct classes. Framing this problem as the time evolution of damage due to fatigue loads allowed the use of hidden Markov models to differentiate the type of manufacturing flaw present, with results indicating 85.7% accuracy given this limited dataset.