Fast characterisation of steel cleanness by advanced mathematical analysis of spark and laser source optical emission data

In order to develop advanced mathematical algorithms for evaluating spark discharge and laser optical emission spectroscopic (OES ) data that give comparable results among different manufacturers and different spectroscopic equipment a common European platform has been implemented. Measuring data of round robin measuring campaigns have been converted to a unified file format and have been stored in a data base at the Fraunhofer-lnstitut für Lasertechnik (ILT) that can be accessed through a web server by all project partners. All the measuring data stored in the web data base have been processed with advanced mathematical algorithms that have been developed by the partners and the results are stored in the web data base too. Different evaluation approaches have been developed among these is a Probabilistic Neural Network (PNN) based classifier that achieved good results in classifying different cleanness grades. This PNN is accessible online at the web server.