Monitoramento ótico de impactos mecânicos supervisionado por inteligência artificial

Abstract

In this work is presented an intelligent optical sensor capable to localize the position of mechanical impacts on a flat surface structure. The transducers were previously characterized relatively to mechanical strain. In this application, four optical fiber Bragg gratings were installed on a polymethyl methacrylate square plate, subjected to a set of mechanical impacts under controlled conditions. The responses provided by the transducers were used to model and test Artificial Neural Networks. The system showed the capacity to determine the disturbance positions, even at impact points different from the ones used in the training stage. The higher mean square error obtained in the test stage was 1,1 cm relatively to the impact position, wherein the points were located at a distance of 5 cm from each other on a test surface with 60 cm length , 60 cm width and 6 mm height.