Fast characterisation of steel cleanness by advanced mathematical analysis of spark and laser source optical emission data
This multi-partner research project, supported by ECSC funding, has evaluated the potential of single pulse spark and laser-based optical emission analysis techniques as rapid methods for characterising steel cleanness. Critical factors associated with sample preparation and the capture of raw data were carried out initially to allow robust statistical methods to be developed to identify inclusions within the pulse discriminated data streams (e.g. MnS, CaSiAl, etc.). Directly measuring the amount of material ablated and relating this to the inclusion emission intensities has provided quantitative measurements of inclusion in a range of steels. The application of advanced mathematical techniques, such as multi-variate techniques for cluster analysis and neural networks, provided a direct interpretation of inclusion populations. Indeed a trained probabilistic neural network successfully graded Ca-modified steels in comparison with the conventional metallographic method of classification based on microscopy. The development of a dedicated website was seen as a central feature of this project. In addition to the transfer of locally generated data, the web server was used to present interpreted results. In addition, a trained probabilistic neural network was deployed directly on the project's web server as an active agent for demonstrating the evaluation of grade classification for Ca-modified steels. The successful application of neural network software to data from circulated samples across the partners' instruments demonstrated the scope of application and the potential use of interpretation software directly from analytical instruments.
Bibliographic Reference: EUR 22079 EN (2006), 116 pp. Euro: 25
Availability: Katalogue Number: KI-NA-22079-EN-S The paper version can be ordered online and the PDF version downloaded at: http://bookshop.europa.eu
ISBN: ISBN: 92-79-02081-1
Record Number: 200719286 / Last updated on: 2007-07-10
Original language: en
Available languages: en