Doctoral Researcher (PhD Candidate) on Machine Learning-based Predictive Maintenance in Industry 4.0

Manufacturing companies in the Industry 4.0 era are increasingly looking for implementing Predictive Maintenance (PdM) to predict failures, classify faults, and optimize maintenance tasks. Artificial Intelligence (AI), and particularly Machine Learning (ML) techniques are applied to build such prediction, classification and optimization models. Although these models yield high performance in static scenarios, their performance becomes questionable during operational use due to two phenomena known as (i) “imbalanced data”: when predicting faults, the majority of the collected data/logs indicates a normal state of operation, whereas only a small fraction indicates faults; and (ii) model drifts (a.k.a, concept and data drifts): when changes happen within the statistical properties of the target class labels or within the independent features (e.g., due to perturbations resulting from a change in hardware, a defective sensor, a network malfunction, wireless interferences, etc.). While these two challenges (imbalance data & model drifts) have received significant attention in image recognition, they have been little investigated in the field of time series analysis and forecasting, which is at the heart of PdM. This PhD thesis will focus on investigating innovative approaches to progress the state-of-the-art in predictive maintenance (PdM) and Machine Learning (ML).

The PhD grant is funded by an industrial partnership program: SnT (University of Luxembourg) and Cebi (aworldwide manufacturer of electromechanical components for the automotive industry), so the PhD candidate must commit to collaborate with and support the industrial partner on a daily/weekly basis.

You Will Be Required To Perform The Following Tasks

• Survey the scientific literature in the relevant research domains
• Investigate new approaches to improve PdM performance when facing imbalanced data & model drifts situations/problems
• Disseminate results through scientific publications
• Communicate and closely work with the partner to collect requirements and report results
• Implement proof-of-concept software tools

The Supervision Team You Will Be Working With Is

• Sylvain Kubler: daily advisor
• Yves Le Traon: head of SerVal