During the last twenty years, many innovative control architectures of manufacturing systems have been developed and promoted in literature. One of the main attributes, in correlation with the aims of Industry 4.0 paradigm, is to define control architectures where both the actors and the interactions between these actors could cope with an evolution of the environment. To do so, dynamic architectures are being recently developed, where the hierarchy of decision can be jeopardized at any time during the normal behaviour of the system. However, the deployment of such architectures faces major software development issues, that a proper initial modelling could help solving. The objective of this paper is to exhibit good practices in the modelling of dynamic architectures in order to enable an automatic reconfiguration when needed.