Aplicação da Teoria dos Conjuntos Fuzzy no Estudo da Impedância Eletromecânica

Abstract

Structural Health Monitoring (SHM) can be performed by electromechanical impedance. The impedance signature is changed in relation to the reference, with the modification of the structure. Two major challenges for SHM are the normalization of the collected impedance data and to determine strategies to assess the level of damage to engineering structures and equipment over time. The objectives of this work are to propose a new data normalization technique and to model the damage level of an aluminum beam, using Fuzzy Rule-Based Systems (FRBSs) that are generated by means of Adaptive Neuro-Fuzzy Inference System (ANFIS). For the first aim, the training is carried out with the input variables temperature and frequency, and the output data are baselines signature impedance values. The temperature effect can generate changes in the impedance signature, leading to incorrect structural diagnoses. Because of that, it is necessary to compensate for the effect of this variable for later prediction of impedance signatures without damage, at temperatures that were not necessarily observed in the data collection. Results obtained in the validation, in which a part of the data was used for training the FRBSs and another part intended for validation, both among the baseline data, indicate good accuracy of the predicted signatures since the highest Correlation Coefficient Deviation (CCD) damage index obtained was 0,003800021. To evaluate the damage level, FRBSs were constructed with the input variables being two damage indices, established by the electromechanical impedance signatures. The FRBSs average hit percentages are 95 % . This result can indicate possible inputs for FBRSs in order to identify the damage levels, when these FRBSs output values are not known. Finally, the methodology proposed in this work is used for damage detection process in an experiment to detect corrosion related damage in metallic structures.