A deep reinforcement learning method for replacement optimization of aero-engine

Aero-engine is an important part of aircraft, its health state directly determines the safety of aircraft operation. Therefore, how to find the best replacement time for aero-engine before failure has become a crucial problem. This paper proposes an optimization algorithm for aero-engine replacement strategy based on deep reinforcement learning. The Gamma process is used to simulate the natural degradation of the aero-engine, and the homogeneous Poisson distribution is used to simulate the random shock of the external environment. The degradation process of aero-engines in the real service environment is simulated by integrating natural degradation and random shock. The replacement optimization of aero-engines is modeled by Markov decision process (MDP), and the optimal replacement strategy is solved by Double Deep Q-Learning (DDQN). Through a series of experiments, it is proved that the proposed algorithm can effectively avoid the problem of dimension explosion, and find the best timing for aero-engine replacement.