Contextual Performance Profiling and Anomaly Detection for Industrial Assets
This project delivers an accessible toolkit for contextual performance profiling and time-series anomaly detection across industrial machinery.
The project successfully addressed the industry 4.0 challenges in monitoring industrial assets and data exploitation through AI-powered anomaly detection and performance profiling. A comprehensive solution was developed that bridges the gap for industries facing difficulties in recruiting specialized data and AI experts. The delivered methodology for contextual performance profiling and anomaly detection has been streamlined and modularized, with validation across multiple industrial machinery datasets. The completed starter kit provides an intuitive computational environment, enhanced with educational resources that have strengthened users' technological competencies. A Python package for time series anomaly detection complements the solution, making advanced analytics accessible to non-AI experts. The starter kit's user interface has been tested with some end-users, demonstrating effective support for training service providers and end users.
Downloads
Starter kit: https://colab.research.google.com/github/EluciDATALab/elucidatalab.starterkits/blob/main/notebooks/SK_1_2_2_Contextual_Anomaly_Detection/elucidata_starterkit_1_2_2.ipynb
(PyPI-installable) Python package for time series anomaly detection: Documentation – Source code
Contact
Contact: Alessandro Murgia (Sirris), Elena Tsiporkova (Sirris), Mathias Verbeke (KU Leuven) and Wannes Meert (KU Leuven)
Researchers: Mathias Verbeke (PI), Elena Tsiporkova (PI), Robbert Verbeke (Senior Data Scientist), Fabian Fingerhut (PhD researcher), Louis Carpentier (PhD researcher), Alessandro Murgia (Senior Data Scientist)
Research groups: Sirris – EluciDATA and KU Leuven - DTAI