Sixth International Conference on Advances in Mechanical and Robotics Engineering - AMRE 2017
Author(s) : GEORGE GEORGOULAS , PETROS KARVELIS , VASILIS TZITZILONIS , VASSILIOS KAPPATOS
In this paper a method for the estimation of the optimum sensor positions for acoustic emission localization on ship hull structures is presented. The optimum sensor positions are treated as a classification (localization) problem based on a deep learning paradigm. In order to avoid complex and timeconsuming implementations, the proposed approach uses a simple feature extraction module, which significantly reduces the extremely high dimensionality of the raw signals/data. The optimum sensor position is defined by the maximum localization rate. In simulation experiments, where a stiffened plate model was partially sunk into the water, the localization rate of acoustic emission events in a noise-free environment is greater than 99.5 %, using only a single sensor.