Improvement of the performances of a commercial hand-held laser-induced breakdown spectroscopy instrument for steel analysis using multiple artificial neural networks.

In this article, we present a study on the optimization of the analytical performance of a commercial hand-held laser-induced breakdown spectroscopy instrument for steel analysis. We show how the performances of the instrument can be substantially improved using a non-linear calibration approach based on a set of Artificial Neural Networks (ANNs), one optimized for the determination of the major elements of the alloy, and the others specialized for the analysis of minor components. Tests of the instrument on steel samples used for instrument internal calibration demonstrate a comparable accuracy with the results of the ANNs, while the latter are considerably more accurate when unknown samples, not used for calibration/training, are tested.