Experimental characterisation of laser surface remelting via acoustic emission wavelet decomposition
Bor Mojškerc, Dunja Ravnikar, Roman Šturm
Abstract
This paper presents a novel technique of monitoring laser surface remelting (LSR) via measurements of process generated acoustic emission (AE). The LSR conditions include a variable laser power level and an ambient air or argon 5.0 inert gas atmosphere. The microstructure and microhardness of the remelted surface layer are evaluated. The recorded AE signal datasets are analysed via the wavelet decomposition technique. The resulting energy and energy proportions of approximation and detail coefficients are calculated. The decomposed AE signal features are evaluated in correlation with the LSR process conditions. A supervised machine learning cubic support vector machine classifier type is used for the classification of laser pulses. The experimental results show an overall 98% classification accuracy into the corresponding LSR laser power level and atmosphere conditions, confirming the utility of monitoring the LSR process and the resulting material surface layer characteristics via the AE wavelet decomposition technique.