This paper presents a dynamic grouping and routing approach for the maintenance optimization of a geographically dispersed production system (GDPS) consisting of several production sites located far apart from each other. Only one maintenance center is in charge of the preventive maintenance of the system. Maintenance grouping and routing are two interrelated processes but often investigated separately in literature. In this paper, these two processes are jointly studied and integrated in a global model considering economic and geographical dependencies at both component and site levels. The optimal maintenance grouped plan and routes are then determined by a combination of the Local Search Genetic Algorithm (LSGA) and Branch and Bound method (BAB). Moreover, several dynamic contexts impacting the current optimal maintenance grouped planning and routing, which may occur with time, are also studied and integrated in the joint optimization process. Thanks to this consideration, the proposed approach allows updating the grouped maintenance planning and routing to take into account the impacts of dynamic contexts when they occur. The uses and advantages of the proposed approach are illustrated through a numerical example of a GDPS consisting of 15 components located in five different sites.