Student Research Project or Thesis: Physics-Informed ML for Payload and Wind Loads on Cable Car Towers

Fraunhofer-Gesellschaft · Stuttgart

ThesisGerman needed: not said

Still open on 11 Nov

You apply on the company's own website.

Key facts

Pay: Pay depends on your qualifications
Language needed: German and English
Subject you study: Automation Engineering, Computer Science, Mechanical Engineering, Mechatronics, Physics, Aerospace Engineering

Modeling: Build a differentiable physics model of a cable car tower in PyTorch, covering cable statics, influence lines and wind loads, with parameters that can be learned directly from measurement data
Design: Develop hybrid architectures combining a physics model and a neural network, and compare them with purely physics-based simulations, such as MATLAB, and purely data-driven approaches
Analysis: Systematically assess different ways of incorporating physics, including physics-based features, synthetic pretraining, integrated physics layers and physics terms in the loss function
Validation: Test models using real field data from a cable car test facility and investigate transfer learning to other towers
Integration: Assess the resource efficiency and feasibility of implementing the best model on a low-power STM32 microcontroller

Current enrollment in a master's degree in Mechanical Engineering, Mechatronics, Computer Science, Physics or a comparable subject in Germany or the EU
Practical experience in machine learning with Python, ideally some knowledge of PyTorch; MATLAB experience is an advantage
Good basic understanding of engineering mechanics, particularly statics, and enthusiasm for combining physics and AI through physics-informed machine learning
Structured and independent working style and good German or English skills

Varied tasks in applied research with direct access to real long-term field data from a cable car test facility
Modern equipment including a powerful GPU workstation and microcontroller boards
Active involvement in an ongoing research project
Close technical supervision by research staff, with the aim of publishing a scientific paper together
Friendly and committed team

Modeling: Build a differentiable physics model of a cable car tower in PyTorch, covering cable statics, influence lines and wind loads, with parameters that can be learned directly from measurement data
Design: Develop hybrid architectures combining a physics model and a neural network, and compare them with purely physics-based simulations, such as MATLAB, and purely data-driven approaches
Analysis: Systematically assess different ways of incorporating physics, including physics-based features, synthetic pretraining, integrated physics layers and physics terms in the loss function
Validation: Test models using real field data from a cable car test facility and investigate transfer learning to other towers
Integration: Assess the resource efficiency and feasibility of implementing the best model on a low-power STM32 microcontroller

Current enrollment in a master's degree in Mechanical Engineering, Mechatronics, Computer Science, Physics or a comparable subject in Germany or the EU
Practical experience in machine learning with Python, ideally some knowledge of PyTorch; MATLAB experience is an advantage
Good basic understanding of engineering mechanics, particularly statics, and enthusiasm for combining physics and AI through physics-informed machine learning
Structured and independent working style and good German or English skills

Varied tasks in applied research with direct access to real long-term field data from a cable car test facility
Modern equipment including a powerful GPU workstation and microcontroller boards
Active involvement in an ongoing research project
Close technical supervision by research staff, with the aim of publishing a scientific paper together
Friendly and committed team