In commercial airline operations, fleets must execute flight missions while managing complex engine maintenance under tight capacity constraints—such as limited spare-engine inventories and multi-stage repair processes—making maintenance scheduling a challenging combinatorial problem. This paper proposes a maintenance strategy for a 10-aircraft fleet using a distributed reinforcement learning (DRL) framework based on the IMPALA architecture. By decoupling exploration and learning across parallel actors and employing V-trace for bias correction, the DRL agent efficiently learns maintenance policies that respect resource limits and stochastic engine degradation. A custom simulation environment—modeling engine life cycles, unexpected failures, overhaul and minor-service actions, and spare-engine pools—was used to evaluate the approach over a 100-week horizon. Experimental results demonstrate that the proposed DRL strategy maintains high fleet availability, avoids large-scale groundings, and generates stable, automated maintenance schedules without manual heuristics, thereby validating its feasibility for complex, multi-state fleet maintenance problems.