The French gas transmission system operator, GRTgaz, conducts regular surveillance of its pipeline network for safety and performance purposes. This activity requires planning and execution of inspections using a multimodal fleet, including cars and planes. The goal, over a year, is to minimize overall costs by determining a weekly schedule and feasible tours for each mode. This problem is framed as a Periodic Capacitated Arc Routing Problem (PCARP) with a multimodal fleet and has not been studied before. To address the problem complexity, we propose a two-phase approach, alternating between tactical (scheduling) and operational (routing) decisions. Moreover, machine learning techniques are incorporated to this approach to provide fast cost approximation. In this talk, we present the overall framework for solving GRTgaz surveillance problem.