Anomaly Detection in Parallel Time Series in production lines for Philips Personal Health Products

Sanguinazzi, Silvia (2025) Anomaly Detection in Parallel Time Series in production lines for Philips Personal Health Products. Bachelor thesis, Data Science and Society (DSS).

PDF BA-4849744-S-Sanguinazzi.pdf Restricted to Repository staff only Download (5MB)

Abstract

This study investigates anomaly detection in the quality check step of an electric toothbrush production line by analyzing Duty Cycle (DC) measurements collected across six different stations. An initial similarity analysis revealed consistent behavior with deviations occurring mostly at Stations 1 and 5, often linked to maintenance or operational changes. An unsupervised ensemble of classical machine learning models was developed with a majority voting mechanism. To address the temporal irregularities, a rolling window strategy was implemented on each individual model. Additionally, a deep learning model was employed. The results demonstrated a concordance between detected anomalies and documented operational changes.

Item Type:	Thesis (Bachelor)
Name supervisor:	Zwitter, A.J. and Doherty, H. L.
Date Deposited:	05 Aug 2025 09:50
Last Modified:	05 Aug 2025 10:01
URI:	https://campus-fryslan.studenttheses.ub.rug.nl/id/eprint/750

Actions (login required)

View Item