Refractory residues classification strategy using emission spectroscopy of laser-induced plasmas in tandem with a decision tree-based algorithm

J. Moros, L.M. Cabalin, J.J. Laserna, Analytica Chimica Acta, in press.

Abstract

The recycling of refractory scraps began to be forged just over a decade ago. Until then, virtually all refractory scraps were disposed off in landfill sites without any application. Over these past few years, a growing interest and a gain steady momentum of the circular economy, the emergent framing around waste and resource management that promotes the notions of their productive cycling, has been the driving force towards the “zero waste” culture across the spectrum of refractory users and producers. In this way, the circular economy, operated following strategies such as, but not limited to, reusing, recycling, and remanufacturing, has played the pillar role in the different essential value chains of the refractory industry to the entering the new era of secondary raw material supply. In any case, prior to starting any sustainable process, it is really necessary to know the wastes and to classify them. In this context, the present research focused on a refractory residue-classification strategy based on combined laser-induced breakdown spectroscopy (LIBS) and a decision tree algorithm for a qualitative analytical performance. This tandem approach allowed the categorization of a rich set of residues in up to 10 different refractory groups. By choosing original LIBS emission intensities and intensity ratios involving the most relevant constituent elements (Al, Mg, C ‒through its related-species CN‒, Si and Zr) of various refractory wastes, a decision tree with multiple nodes that decided how to classify inputs was designed and trained. Categorization performed from LIBS emission spectra of “blind” refractory residues showed
that LIBS data combined with this supervised machine learning algorithm provided good refractory scraps-classification performance, with a classification accuracy of up to 75%. However, some more than justified decisions of the algorithm on allegedly misclassified residues showed that scores for the decision tree could found to be far superior to those obtained. The results achieved support the strategy

 

Related Posts