Jun Chen, SFB 1313 doctoral researcher at the Institute for Parallel and Distributed Systems (research project D02), will give her milestone presentation on "Parallel numerical coupling methods for interface problems in porous media simulations".
Date: Wednesday, 21 May 2025
Time: 2 pm CEST
Title: "Parallel numerical coupling methods for interface problems in porous media simulations"
Venue: Room 2.013, Universitätsstraße 38, 70569 Stuttgart, Campus Vaihingen. If you are interested to participate online in Jun Chen's milestone presentation, please contact Samaneh Vahid Dastjerdi to receice the Webex link >>> samaneh.vahiddastjerdi@mechbau.uni-stuttgart.de
Abstract
Porous media systems are common in many engineering applications and are often modelled using multi-physics and multi-scale approaches due to their complex internal structure. The goal of this doctoral project is to develop scalable and robust partitioned coupling methods for interface problems in porous media simulations, using the coupling software framework preCICE. With ongoing study, we address two key challenges central to this objective. First, we investigate the instability introduced by quasi-Newton (QN) acceleration methods when applied to problems involving physically bounded variables. To mitigate overshooting effects, we propose and evaluate several limiting strategies. Second, we address the high computational cost of many-to-one multi-scale simulations, where the dominant cost arises from repeated micro-scale evaluations. Building on the adaptivity feature of Micro Manager, we introduce a convergence-aware adaptivity scheme that dynamically relaxes similarity thresholds based on coupling progress. This method significantly reduces the number of micro-scale solves and total simulation time, while maintaining acceptable solution accuracy. In future work, we aim to apply these methods to practical problems and continue improving the robustness and usability of the supporting software components.