Damage Recognition for Structural Health Monitoring

In the field of structural health monitoring, researchers focus on the design of systems and techniques capable of detecting damage in structures. However, it is difficult to develop robust detection schemes that are invariant to environmental and operational conditions. In this report, we investigate several signal processing and machine learning techniques for developing such robust systems. From experimental data of a pressurized pipe, we extract 212 different data features and implement three different classification algorithms for detecting and localizing damage: adaptive boosting, support vector machines, and a combination of the two. The third algorithm shows the best overall performance in terms of accuracy, ranging from 81% to 100% in detection tests and 70% to 100% in localization tests. Through feature selection, we also demonstrate the effectiveness of features related to the Mellin transform and curve length.