Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
AUTOMATIC STRUCTURAL DAMAGE ISOLATION USING SUPPORT VECTOR MACHINE CLASSIFICATION
Submission Author:
Renan B. M. Santos , SP
Co-Authors:
Renan B. M. Santos, Pablo Souza, Eurípedes Nóbrega
Presenter: Renan B. M. Santos
doi://10.26678/ABCM.COBEM2019.COB2019-2278
Abstract
Structural Health Monitoring uses several signal processing techniques to detect and classify damages in engineering flexible structures. These techniques are usually based on vibration analysis, originated from real structures by adequately positioned sensors. This results in a huge number of signals, making this task virtually impossible for human analysis, consequently demanding efficient data science methods. This paper proposes a machine learning architecture for damage isolation in composite structures, constituted by an unsupervised feature extraction using an autoencoder neural network, and a supervised learning classification based on a Support Vector Machine (SVM) algorithm. Aiming a continuous monitoring system, a Lamb wave method is adopted to generate the signals and preprocessing the data to be classified. Periodic inspection is performed by means of an arrangement of piezoelectric transducers forming a circular array of eight sensors with a central actuator, dividing the monitored area into eight regions. Discrete wavelet and Hilbert transforms are applied to the acquired signals, in order to minimize noise and dispersion effects as well as to improve peak amplitude and location estimation. Damage indexes, which result from the autoencoder model, are used as attributes for an SVM classifier. An experimental dataset is used to train both, autoencoder and the SVM algorithm, in order to predict the target structural integrity by comparing a new input data obtained during the inspection phase with a set of healthy signatures and, if there is any damage, it proceeds to effectively find its localization.
Keywords
Structural Health Monitoring (SHM), Support vector machine, machine learning, Artificial Intelligence, Support Vector Classifier
