Gaussian process regression model for damage localization in plates based on modal data

Abstract

The applications of plate like structures in different fields of engineering are increasing. In this paper, a new damage detection method investigated based on Gaussian process regression model (GPR). GPR is an efficient learning machines which has been used in different fields of engineering. To identify damage, mode shaped and natural frequencies of damaged structures used to train GPR. Finite element modelling of numerical examples and Gaussian process regression (GPR) model are carried out within the MATLAB environment. To show the effectiveness of presented approach, a two-fixed supported plate and a cantilever plate was studied. In other work, a comparative study has been done using a cantilever plates. The natural frequencies were contaminated with noise in above mentioned numerical examples. Results reveal that the proposed method works well using the only first mode data which may be noisy. In other word, GPR can be trained using limited sample numbers for training.