Damage Localization Methodology using Pattern Recognition and Machine Learning Approaches

Abstract

Structural health monitoring (SHM) is regarded as a very important field of development in engineering applications. Some benefits of the applications of a SHM system include the improvement of the security of the structure and the reduction of time and cost of its maintenance. Therefore, the service life of the structures can be lengthened. Besides, the implementation of a SHM system provides real time information about the current state of a structure, not only in civil applications but also in military applications, like aircraft, buildings, wind turbines and others. Gathering the maximum information about the current state of the structure to be diagnosed allows a better decision both in the preventive maintenance or the corrective maintenance. The information that can be gathered is the presence or not of damage, the size of damage or even the extent of damage. However, damage localization offers some very interesting possibilities since it allows to accurately determine where the damage is located thus saving time and effort. This work presents a damage detection and localization methodology by means of machine learning approaches which works by analyzing data from a piezoelectric sensor network attached to the structure under test. Validation of the methodology is performed by using an aluminum rectangular pipe with several damage positions