post thumbnail

AI-PROFICIENT PhD student at UL, Van Thai NGUYEN, defended his PhD on the 12th of June 2023

Van Thai NGUYEN, PhD student at Université de Lorraine defended his PhD entitled “AI-based maintenance planning for multi-component systems considering different kinds of dependencies” on the 12th of June 2023.

The jury was composed of Anne BARROS (CentraleSupelec, reviewer), Chi-Guhn LEE (University of Toronto, reviewer), Antoine GRALL (Université de technologie de Troyes), Khanh NGUYEN, (Université de Toulouse), Phuc DO, (Université de Lorraine, supervisor), Alexandre VOISIN (Université de Lorraine, Co-suprvisor)

His PhD work based on AI-PROFICIENT results can be summarized as follows:

The maintenance optimization of complex industrial systems remains a significant challenge due to the various dependencies among components, including economic, stochastic, and structural dependencies, along with the large number of maintenance decision variables to optimize.

To address this challenge, this thesis aims to propose an artificial intelligence-based maintenance optimization approach that considers different types of component dependencies. Specifically, the proposed maintenance approach incorporates a prediction model based on neural networks for estimating system-level maintenance costs without the need for individual component-level costs. This is carried out within the framework of multi-agent deep reinforcement learning, which can be applied to large-scale sequential decision-making to optimize maintenance decisions. Additionally, a novel model for component state dependencies is developed and integrated into the proposed maintenance approach. Several numerical studies are conducted on systems with different configurations under various observability scenarios to investigate the performance, advantages, as well as the limitations of the proposed maintenance approach.