Projects / Programmes
Development of advanced meshless methods for modelling of material processing on high-performance computers
| Code |
Science |
Field |
Subfield |
| 2.13.01 |
Engineering sciences and technologies |
Process engineering |
Multi-phase systems |
| Code |
Science |
Field |
| 2.03 |
Engineering and Technology |
Mechanical engineering |
meshless methods direct-chill casting of aluminium high-performance computing general purpose GPU
Data for the last 5 years (citations for the last 10 years) on
June 28, 2024;
A3 for period
2018-2022
Data for ARIS tenders (
04.04.2019 – Programme tender,
archive
)
| Database |
Linked records |
Citations |
Pure citations |
Average pure citations |
| WoS |
29 |
276 |
176 |
6.07 |
| Scopus |
35 |
333 |
211 |
6.03 |
Researchers (1)
| no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
| 1. |
36364 |
PhD Boštjan Mavrič |
Process engineering |
Head |
2020 - 2024 |
0 |
Organisations (1)
Abstract
The meshless methods have in recent years found their use in several fields of science and engineering, due to their geometrical flexibility, simplicity of use and high accuracy. In particular a subgroup of local strong-form meshless methods has found numerous applications in the field of modelling of material processing. The applications of the meshless methods in this field have been spearheaded by the group the applicant is part of. Because of high complexity of the processes that the models attempt to describe the computational times on high-end workstations have increased beyond reasonable times useful in practice. The goal of this project is to develop new algorithms and techniques, specialized for local strong-form meshless methods that will allow the models to be run on high-performance computing (HPC) hardware. Recent investments in such hardware in Slovenia, both in local capabilities at Faculty of Mechanical Engineering, University of Ljubljana, and in large scale cluster at IZUM, have increased the interest of metal processing industry in Slovenia in application of the new resources to their processes. The main goal of the project is to lay a groundwork and precompetitively demonstrate its capabilities so that such applications will be possible in the future. Additional complexity is introduced in such efforts due to the fact that the HPC is an exceptionally inhomogeneous platform. Depending on the amount of funding available at time of construction, best available hardware and trends in the field a node of an HPC cluster can contain one or several processors with varying number of computational cores and associated floating processing units. In recent years, an important part of the computational power comes from dedicated accelerator hardware, either in form of graphical processing units (GPUs) or dedicated many-core accelerator cards. Unfortunately, specialized libraries are often used to take advantage of such hardware. The project will focus on the development of a framework that will facilitate the use of such heterogeneous hardware by the models. In first stage the message passing (MPI) paradigm will be used to parallelize the prototypes of the solvers used in the modelling of material processing for use on multiple computational nodes in distributed memory configuration. The implementation of the framework will first be profiled on the local hardware and then iteratively optimized for the HPC cluster at Faculty of Mechanical Engineering and finally for the HPC RIVR hardware at IZUM. In parallel with the profiling of the distributed memory code, the development of a framework that can use accelerator hardware will start. After the prototypes of the models using the framework will be developed, they will again be subject to the same profiling and optimization procedure as in the case of MPI framework. Finally, the performance of the developed frameworks will be critically evaluated and compared to the existing code in terms of computational speed and accuracy. Such frameworks will significantly advance the state-of-the-art in the field, since similar applications have until now only been attempted on less complex numerical models in geophysics that used lower number of unknowns as compared to the models encountered in the field of materials processing. It will further strengthen the applicant and the research group as a leader in this emerging field of mehless computational methods. The underlying frameworks that will be developed and the knowledge that will be obtained will also provide a competitive advantage in search of funding to expand the models to daily use at potential industrial partners in Slovenia and abroad. The models developed from such collaborations will allow the Slovenian metal processing companies to exploit the HPC infrastructure present in Slovenia.
