This paper aims to present a new systemic modelling approach using the unified modelling language (UML) and Multi-Agents Systems (MAS) to model the complexity and analyze the risks of maintenance activities within a Liquefied Petroleum Gas (LPG) supply chain. This approach is a two steps corresponding to Modelling & Simulation and Risk Analysis. Based on a multi-agent simulator, the objective is to consider the complexity of maintenance tasks of supply chain equipment to better understand risks and to identify abnormal and accident prone situations. The case study is LPG supply chain located in France. AnyLogic simulator is used to assess safety of this supply chain. Through simulation, this proposed approach provides a whole view to ease understanding of maintenance activities within LPG supply chain and to lead to effective analysis of its safety. The final goal is orienting maintenance actors to the best risk minimizing decisions. (C) 2017 Elsevier Ltd. All rights reserved.