Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Thesis on AI-based Control for Multi-axial Dynamic Testing devices
The goal of this thesis is to suggest an ML-based scheme for efficient control of a Hexapod device, which serves as a multi-axial dynamic testing facility for engineered components.
This Master thesis opportunity is offered by the Chair of Structural Mechanics & Monitoring at ETH Zurich (Prof. Eleni Chatzi) in collaboration with Michael Herrnberger & Patrick Vaudrevange of TWT Gmbh Science & Innovation.
A Hexapod is a kinematic motion device that provides six degrees of freedom with restrictions on the realizable frequency-dependent amplitude. Its control is challenging, as one needs excellent accuracy, possibly focusing on a frequency-range of special importance, and simultaneously with a perfect repeatability of its motion – independent of the given load.
This poses limitations for testing purposes, where the goal is to reliably reproduce an actuation signal for the tested setups.
A Hexapod is a kinematic motion device that provides six degrees of freedom with restrictions on the realizable frequency-dependent amplitude. Its control is challenging, as one needs excellent accuracy, possibly focusing on a frequency-range of special importance, and simultaneously with a perfect repeatability of its motion – independent of the given load. This poses limitations for testing purposes, where the goal is to reliably reproduce an actuation signal for the tested setups.
The goal of this thesis is to suggest an ML-based scheme for efficient control of such a testing facility.
Specific Tasks:
- Considering a hexapod either in a virtual setup or within an actual laboratory setting (as hosted at the IBK Structures Lab, ETH Zurich: https://chatzi.ibk.ethz.ch/downloads/facilities.html), analyze methods for its control.
- Optimization of hexapod’s motion through machine learning techniques.
- Comparison of the results achieved from various methods in terms of accuracy and repeatability.
For Hexapod’s control, the intention is to investigate usefulness of physics-informed neural networks and reinforcement learning methods, among other tools. The comparison of the results with the repeatability of the data-based approach, will be performed by a traditional control software.
The goal of this thesis is to suggest an ML-based scheme for efficient control of such a testing facility. Specific Tasks: - Considering a hexapod either in a virtual setup or within an actual laboratory setting (as hosted at the IBK Structures Lab, ETH Zurich: https://chatzi.ibk.ethz.ch/downloads/facilities.html), analyze methods for its control. - Optimization of hexapod’s motion through machine learning techniques. - Comparison of the results achieved from various methods in terms of accuracy and repeatability. For Hexapod’s control, the intention is to investigate usefulness of physics-informed neural networks and reinforcement learning methods, among other tools. The comparison of the results with the repeatability of the data-based approach, will be performed by a traditional control software.
This thesis opportunity is not an internship and is not financed. It is open to students at ETH Zurich or students from other institutions, who wish to conduct their thesis at ETH Zurich.
If you are interested, send an email at echatzi@ethz.ch with the title "TWT - Hexapod MSc thesis", introducing yourself along with your CV and a list of taken courses (or transcripts).
This thesis opportunity is not an internship and is not financed. It is open to students at ETH Zurich or students from other institutions, who wish to conduct their thesis at ETH Zurich.
If you are interested, send an email at echatzi@ethz.ch with the title "TWT - Hexapod MSc thesis", introducing yourself along with your CV and a list of taken courses (or transcripts).
