Chapter 1: Machine Learning Algorithms for Damage Detection

This chapter poses the damage detection for civil, mechanical, and aerospace structures in the context of a pattern recognition paradigm for structural health monitoring (SHM), where machine learning algorithms are essential to learn the structural behavior from experience, following the same principle of the human brain. These algorithms are especially relevant in cases where the damage-sensitive features extracted from the structural responses are affected by changes caused by operational and environmental variability and changes caused by damage. State-of-the-art machine learning algorithms are presented based on Mahalanobis squared distance (MSD), Gaussian mixture models (GMMs), principal component analysis (PCA), kernel principal component analysis (KPCA), and autoassociative neural network (ANN); the bioinspired algorithms are highlighted as promising algorithms to overcome some of the limitations of the more traditional ones. All the algorithms present different working principles and seek to generalize the normal structural condition in order to detect deviations, from the baseline condition, associated with damage. The applicability of the chosen algorithms for damage detection will be tested on standard data sets from the Z-24 Bridge in Switzerland. These data sets are unique as they combine 1-year monitoring from the baseline condition, when influenced by extreme operational and environmental variability, with realistic damage scenarios.