Sparse ultrasonic guided waves sensor networks would increase safety and reduce maintenance costs of current civil and industrial structures by providing continuous information about structural health. Common damage modalities that are of concern include fatigue cracks and corrosion in metals, and delaminations and impact damage in composite materials. Ultrasonic guided wave testing is used in various industries due to its superior ability to rapidly inspect large areas in thin structures. Guided waves travel long distances and interrogate the entire thickness of a structure. These properties make guided waves of interest in structural health monitoring (SHM) since a small number of transducers is required to monitor a large structure.
Due to the complexity of guided wave signals, detection strategies in guided wave SHM often are based on detecting changes in the signals based on reference signals collected when the structure is in its pristine state. Damage detection is challenging because there are multitude of other factors that affect the signal such as sensor aging, temperature changes, humidity, and/or varying loading conditions.
In this project, we developed machine learning and statistical signal processing techniques for building reliable and efficient methods for damage classification in structures under varying environmental and operating conditions. In addition, we developed statistical methods for quantifying the probability of detection due to sensor aging and variations in environmental and operating condition.