Please login first
Data Science and Numerical Simulations: Shaping the Future of Metal Process Modeling
* 1 , 2
1  Senior Scientist, LKR Light Metals Technologies Ranshofen, Austrian Institute of Technology, Vienna 1210, Austria
2  Research Engineer, LKR Light Metals Technologies Ranshofen, Austrian Institute of Technology, Vienna 1210, Austria
Academic Editor: Chuang Deng

Abstract:

Abstract

With the emergence of data science techniques, such as reduced-order schemes, real-time modeling, machine learning (ML), and smart control schemes, material process modeling and simulations are undergoing a revolutionary phase. Although traditional analytical methods and advanced numerical simulations still provide estimations of multi-physical and multi-scale material processes, generating real-time predictions remains challenging for these techniques. Additionally, data quality and availability issues have slowed the development of data models, while long computational times have hindered the use of advanced numerical simulations for process control. This paper presents the outcomes of research on the simultaneous use of data models and detailed numerical simulations, highlighting their unique roles in control and data generation for future process modeling. While many data processing and handling schemes exist within the data science field, only a few are suitable for material process applications due to their transient and multi-physical natures. As more physics and phases are considered in numerical simulations, the computational time and resources required become enormous, even for today’s parallel and clustered computers. Recently, integrating certain data techniques within numerical simulation frameworks has drastically reduced computational time (e.g., recurrence computational fluid dynamics). This research scrutinizes the efficient use of these simulation techniques for creating fast databases and utilizing these databases for real-time predictions in transient material processes (e.g., casting processes). Consequently, both data models and numerical simulations, along with experimental validations, play crucial roles in generating accurate and reliable metal process modeling. The objective is to integrate these techniques into digital twin and shadow frameworks, driven by industrial digitalization, to enhance greener and more efficient manufacturing. Finally, predefined simulation scenarios were used to produce reliable data models for accurate real-time predictions in metal casting process optimization and control.

Keywords: data models; machine learning; metal processes; digitalization; digital twin;
Comments on this paper
Currently there are no comments available.



 
 
Top