Open Access
Subscription or Fee Access
Automated Near-Optimal Feature Extraction Using Genetic Programming with Application to Structural Health Monitoring Problems
Abstract
Structural health monitoring systems at their core measure structural response and infer real-time damage and performance information. Conventional data processing involves extraction of low-dimensional features from time series measurements that are then input to a classification or outlier detection algorithm. Desirable features are highly sensitive to changes of interest in the structure while also robust in the presence of noise and varying operational and environmental conditions. Traditional feature design requires experts with domain-specific knowledge resulting in an expensive and time-consuming process. Recently, genetic programming, an evolutionary computation method, was adapted to provide automated, data-driven development of feature extraction processes with minimal user input in a supervised learning approach. This study experimentally validates the genetic programming approach with comparisons to common feature sets. Demonstrated applications include ultrasonic damage detection, condition monitoring for rotating machinery, and vibration-based structural health monitoring.