• Fahle, Simon

Abstract

Data-driven analytical approaches such as machine learning bear great potential for increasing productivity in industrial applications. The primary requirement for using those approaches is data. The challenge is to not only have any kind of data but data which has been transformed into an analytically useful form. Building upon this initial requirement, this paper presents the current state concerning data analysis and data integration in the industrial branch of hot forming, specifically focussing on radial-axial ring rolling. The state of the art is represented by the results of a data survey which was completed by six of Germany’s representing radial-axial ring rolling companies. The survey’s centre of interest focuses on how data is currently stored and analysed and how it gets depicted into eight different statements. Based on the results of the survey a framework is proposed to integrate data of the whole production process of ring rolling (furnace, punch, ring rolling machine, heat treatment and quality inspection) so that data-driven techniques can be applied to reduce form and process errors. The proposed framework takes into account that a generalized standard is hard to set because of already grown structures and a huge variety of analytical methods. Therefore, the framework focuses on data integration issues commonly found in an industrial setting as opposed to controlled research environments. The paper proposes methodologies on how to utilize the potential of each company's data. As a result, the proposed framework creates awareness for saving the data in a standardized and thoughtful manner as well as building a data-driven culture within the company.

Topics

• impedance spectroscopy
• forming
• machine learning