Preliminary work at IPEK has shown that ergonomic parameters, such as the angle of the elbow, can be derived from tool-integrated sensors (including pressure-sensitive film on tool handles) during drilling. Existing models primarily operate on a discrete-time basis and neglect the temporal dynamics of movement and load trajectories. This master’s thesis aims to investigate the extent to which temporal neural networks (e.g., TCN, LSTM, Transformer) can utilize the temporal structure of sensor data to improve the estimation of ergonomic characteristics compared to discrete-time approaches.

Based on an existing dataset from tool-integrated sensors, various temporal model architectures (e.g., TCN, LSTM, Transformer) are to be implemented, trained, and systematically compared with existing point-in-time discrete models. Using ablation and sensitivity analyses, the added value of temporal modeling for deriving ergonomic motion and load characteristics is to be evaluated and scientifically classified. Optionally, a supplementary measurement study may be designed and conducted to expand the dataset.

• You are studying mechanical engineering, mechatronics, or a related field

• Structured, independent, and meticulous approach to work

• Interested in the intersection of data-driven modeling and ergonomics

• Solid knowledge of machine learning and deep learning, ideally with a focus on time series or sequence models

If you're interested, I look forward to hearing from you!