Master Thesis Student

Collected and processed machine sensor data (.JSON, BLM) and shared datasets in Parquet format for efficiency. Integrated process and quality data into unified datasets for predictive modeling. Conducted EDA to identify significant features impacting workpiece quality (parallelity, perpendicularity, flatness). Trained non-linear regression models and progressed to deep neural networks for enhanced prediction accuracy. Applied federated learning using the Flower library to preserve data privacy, benchmarking against centralized and individual learning setups. Evaluated multiple aggregation algorithms to improve federated performance.