PostDoc Position in Physics-Based Machine Learning for Geophysical Flows
German Aerospace Center (DLR)
Planetary and Solar System Sciences (PS)
The Department of Planetary Physics of the DLR Institute of Planetary Research in Berlin has a long-standing experience in developing models of the dynamics and evolution of the interior of planets and moons, with a major focus on numerical simulations of thermal convection in the mantle of rocky bodies.
During the past few years, our group has been investigating data-driven approaches combining machine learning with traditional numerical methods for the solution of thermal convection problems in the framework of both forward (ref. 1, ref. 2) and inverse modelling (ref. 3). In this project named PLAGeS (Physics-based Learning Algorithms for Geophysical fluid Simulations), which is funded for 3 years by the Federal Ministry of Education and Research, we aim at extending our purely data- driven approach to the rapidly growing field of physics-based machine learning. We will adopt an approach that doesn’t require previously generated simulation data for training and incorporates the underlying physics into the learning algorithms. The final goal of the project is the development of a fast, machine-learning-based thermal convection solver and its application to the modelling of the interior evolution of rocky bodies, with a particular focus on Mercury, Mars, and rocky exoplanets. The project will be carried out at the Department of Planetary Physics of the DLR Institute for Planetary Research in Berlin and in close cooperation with the Model-Driven Machine Learning Group based at the Helmholtz Centre Hereon, which specialises in atmosphere and ocean simulations.
To support our group in this exciting field at the interface between data science and numerical simulations, we are looking for a highly motivated postdoctoral researcher. The ideal candidate will have strong mathematical and programming skills, an established experience in numerical methods for the solution of partial differential equations, and at least a basic expertise in machine learning. Previous experience in planetary physics, thermal convection and computational fluid dynamics is welcome but not necessary. The successful candidate will focus on challenging and impactful research, with minimal administrative duties and will be supported by extensive on-site expertise in planetary science and excellent computational resources.
This is a fixed-term appointment for up to 3 years, with remuneration according to the TVöD 13 level (100%). The intended start date is January 2023 or shortly after.
- Development of a machine-learning-based solver for thermal convection
- Extensive validation and testing against existing analytical and numerical solutions
- Derivation of scaling laws for convective heat transfer in dependence of multiple parameters
(pressure scale, viscosity, thermal expansivity and conductivity, internal heating rate, etc.)
- Application of the developed solver to the thermal evolution of Mars, Mercury and rocky exoplanets
- Presentation of the work at institute’s seminars and international conferences, and publication of the
obtained results in peer-review journals
- PhD degree in physics, mathematics, computer science, engineering or another relevant discipline
- Excellent mathematical and programming skills
- Experience with numerical methods for the solution of partial differential equations
- Basic experience in machine learning
- Ability to work in a structured way independently and as part of a team
- Excellent communication skills in English
Applications must be submitted through the DLR job portal at this address and include a motivation letter summarising research interests, academic transcripts from relevant degrees, CV, and names and contact information of at least two references.