Feature Extraction and Sensor Fusion for Ultrasonic Structural Health Monitoring Under Changing Environmental Conditions

Strongly reverberating diffuse-like ultrasonic waves can interrogate large areas of complex structures that do not support more easily interpreted guided waves. However, sensitivity to environmental changes such as temperature and surface wetting can degrade the performance of a structural health monitoring system using these types of waves. Surface wetting is investigated here with a simplified experiment where controlled amounts of water are applied to the surface of a specimen in conjunction with incrementally introduced artificial damage. A feature-based approach is taken whereby differential features between a signal and a baseline are defined that are sensitive to damage but less sensitive to surface wetting, and multiple features obtained from a spatially distributed sensor array are combined via a voting strategy. In addition, the features considered are insensitive to moderate temperature changes, which are unavoidable even in the laboratory. Experimental results show a probability of detection greater than 90% when detecting damage in the presence of modest surface wetting while maintaining a false alarm rate under 5%.