Combining Multiple Neural Networks to Predict Bronze Alloy Elemental Composition

The problem of predicting the composition of alloys measured by means of Laser-Induced Breakdown Spectroscopy (LIBS) analysis is frequently tackled in the literature. In this paper we propose the use of an ensemble of neu-ral networks to model the functional relationship between LIBS spectra and the corresponding composition of bronze alloys, expressed in terms of concentra-tions of the constituting elements. The networks are trained independently and their inputs are determined by different feature selection processes. Their out-puts are then combined by applying an averaging function. The results achieved allow to correctly predicting the composition of unknown bronze alloy samples.