Machine Learning Techniques and Population-Based Metaheuristics for Damage Detection and Localization Through Frequency and Modal-Based Structural Health Monitoring: A Review

Vibration-based damage detection techniques, and particularly frequency and modal-based methods, address the problem of localization and quantification of damage in a structure by using observed changes in its dynamic properties. Most of these procedures are based on an optimal criterion in which the stiffness distribution of an element into a structural system is iteratively updated in order to match the computed natural frequencies with the measured ones at a certain level of deterioration. This paper provides a comprehensive review of the optimization methodologies based on random search procedures that have been employed to find practical solutions in the forward and inverse frequency and modal-based Structural Health Monitoring (SHM) problem. Over the last three decades, the application of machine learning approaches and population-based metaheuristics to solve real-engineering optimization problems has grown massively, especially in the SHM field. Therefore, the literature review provided in this paper is organized into three main sections: (1) Machine learning techniques; (2) Population-based metaheuristics; and (3) coupled methodologies. Finally, the purpose of this paper is to provide the reader with a wider understanding and a critical point of view of frequency and modal-based techniques for structural damage detection, and the various computational methods that have been employed to solve the problem.

Automated summary

This paper provides a comprehensive review of optimization methodologies based on random search procedures used in forward and inverse frequency and modal-based Structural Health Monitoring (SHM) problem.