Open PhD Defense Naim Al Khoury

Naim's open PhD defense

We are pleased to invite you to the public defense of Naim Al Khoury's doctoral dissertation on "Degradation-driven Spare Parts Inventory Control for Multi-Machine Systems with Lead Times Uncertainty".

🗓️ Thursday 30th of April 2026.
🕔 Start at 16:00, followed by a reception in the ISyE meeting room department.
📍 Lokaal 0.2, building Magnel, ground floor, Technologiepark-Zwijnaarde 60, 9052 Gent.
🗣️ Spoken language: English.

Attendance registrations open until the 20th of April via this registration link.

There will also be a livestream for those who cannot attend in person via this Teams link.

We warmly welcome colleagues, researchers and anyone interested to join us for this milestone moment.

Get directly involved on this topic and share your thoughts on Naim's open PhD defense on LinkedIn. You can also read more about Naim's history and his work.

Shorty before the public defense, Willem van Jaarsveld, one of the jury members, will present about his research in deep reinforcement learning. You are all cordially invited as well.
- Time: start at 14:15.
- Location: vergaderzaal 1.1 - Hermann von Helmholtz, iGent, Campus Ardoyen.

Degradation-driven Spare Parts Inventory Control for Multi-Machine Systems with Lead Times Uncertainty

Managing spare parts inventory is a critical challenge for service providers maintaining multi-machine systems, particularly under the uncertainty of replenishment lead times. While Condition-Based Maintenance provides predictive insights into future demand, effectively integrating these insights into inventory control remains an open question. This research addresses this challenge by developing data-driven spare parts policies for systems with multiple machines and stochastic lead times.

The study first introduces a Proactive Base Stock Policy, which leverages real-time degradation data to order spare parts, minimizing inventory levels while maintaining service requirements. By exploiting structural properties, an intelligent algorithm is developed to find its optimal parameters. This heuristic policy achieves significant savings compared to traditional policies that do not leverage data.

Given these potential savings, a development framework is established to enable benchmarking of policies from different methodologies. The problem is further extended to include inventory capacity constraints and batch ordering. Within this framework, the Proactive Base Stock Policy is adapted to operate under these realistic constraints. Subsequently, the research develops advanced data-driven policies using three Deep Reinforcement Learning algorithms complemented by domain knowledge from existing policies. The results demonstrate that these DRL-based methods outperform traditional policies, scale to larger problem, and are explainable.