The fourth industrial revolution involves more complexity. This research effort focuses on decision‐making in helicopter engine maintenance activities. Such a decision‐making task is difficult and relies on a variety of experts who only have partial knowledge and incomplete situation awareness, due to the great diversity of everyday operational practices. In this paper, we propose a digital twin multi‐agent approach to collaborative decision‐making in prescriptive maintenance.

This research work aims at improving collective decision-making and learning through a digital twin of the organization in the context of a complex industrial activity such as helicopter engine maintenance. Field and bibliographic studies allowed to determine that the digital twin should be based on a multi-agent system model for reasons of flexibility and modularity necessary in this constantly changing environment. The digital twin is intended to adapt to the organization but also to enhance it by including missing information flows. This paper presents the agent model chosen and inspired from reinforcement learning and how it allowed to identify these missing flows. The importance of interfaces in the digital twin and what they should contain to integrate agents is shown, as well as the psychosocial aspects to be considered for humans to handle their design.