The risks faced by modern energy systems are increasing, primarily caused by natural disasters. As a new form of multi-level energy complimentary utilization, integrated energy systems are attracting more and more attention for their high-efficiency and low-cost. However, due to the deep coupling relationship between systems, they are more susceptible to natural disasters, resulting in a cascading failure. To enhance the resilience of the integrated electricity-gas system, this paper proposes a failure restoration strategy after a natural disaster occurs. First, the temporal constraints of the dispatching model are considered, and the failure restoration problem is molded into a multi-period mixed-integer linear programme, aiming to recover the interrupted loads as much as possible. Second, since the uncertain output of distributed generation sources (DGs) such as wind turbines and photovoltaic systems will threat the reliability of restoration results, the robust formulation model is incorporated to cope with this problem. Third, we propose a new modeling method for radial topology constraints towards failure restoration. Moreover, the Column and Constraints Generation (C&CG) decomposition method is utilized to solve the robust model. Then, the piecewise linearization technique and the linear DistFlow equations are utilized to eliminate the nonlinear terms, providing a model that could be easily solved by an off-shelf commercial solver. The obtained results include the sequence of line/pipeline switchgear actions, the time-series dispatching results of electricity storage system, gas storage system, and the coupling devices including the gas-fired turbine, power to gas equipment. Finally, the effectiveness of the proposed restoration strategy is verified by numerical simulation on a 13-6 node integrated energy system.