Novelty detection based on symbolic data analysis applied to structural health monitoring

In this paper, the concept of Symbolic Data Analysis is used in order to classify different structural behaviors. For this purpose, raw information (acceleration measurements) but also processed information (modal data) are used for feature extraction. This paper proposes an original approach in which the SDA is applied to three well know classification techniques: Bayesian Decision Trees, Neural Networks and Support Vector Machines. Results regarding experimental tests performed on a railway bridge in France are presented in order to highlight advantages and drawbacks of the described methodology. The results show that the SDA combined to the cited methods are efficient enough to classify and to discriminate structural modifications, either considering the vibration data or the modal parameters. More robust results are given by modal data than by measurement data.