Mr. llias Papadimitriou is a mechanical engineer and has been working as a technical expert powertrain and data analytics at GF Casting Solutions in Schaffhausen since 2010. His background is the development of internal combustion engines, particularly in the area of multibody dynamics, structural analysis, tribology, and NVH. He started his career in 1995 at AVL List in Graz. 2001 joined Ferrari S.p.a in the engine development of F1 and passenger cars in Maranello. In recent years, he has been active in data analytics and the implementation of artificial intelligence methods for quality optimization of industrial engineering processes.
SPEECH TITLE: RISKS AND CHALLENGES OF DEEP LEARNING METHODS IN THE INDUSTRIAL MANUFACTURING PROCESSES
There have been many discussions about using Deep Learning methods in the manufacturing process in the last few years. Some companies deal with these approaches in industry 4.0 digitalization projects. Others try to discover the potential of deep learning methods in specific problems like quality problems or maintenance prediction problems. Regardless of the motivation, the manufacturing industry faces some common issues in the implementation process of deep learning methods. The first issue concerns the fact that deep learning methods have to be applied by engineers. They are rather used to perform modelling and simulation methods based on mathematical models derived from physical laws. This fact leads to difficulties understanding and trusting this data-based training modelling technique. The second problem concerns the data acquisition systems and the data quality itself. There is many data available in the manufacturing industry, but data quality is not always guaranteed. The presentation shows practical deep learning approaches for pilot projects in the high pressure die casting and sand casting process that reduced a scrap of typical product quality issues like porosities and distortion.