Deep Deterministic Policy Gradient (DDPG) is a deep reinforcement learning algorithm designed to solve decision-making problems in complex systems. DDPG has been extensively used for decision-making in various complex systems. However, DDPG has not been widely applied to the aircraft maintenance decision optimization, which is a crucial aspect of the aircraft Prognosis and Health Management (PHM). Evaluating the performance of DDPG first is essential, as it directly impacts the effectiveness of PHM systems and the determination of maintenance actions. This work investigates the performance of DDPG and compares it with Ant Colony Optimization (ACO) and Deep Q Network (DQN) algorithms within an aircraft maintenance simulation environment. Two types of typical aircraft systems, simple and complex, are defined and simulated in the environment as maintenance objects. We find that DDPG has a unique advantage in terms of training efficiency and solving efficiency when dealing with complex aircraft systems, which have both complex state spaces and action spaces. Besides, the accuracy of the DDPG algorithm’s solution and the rationality of its decisions are verified.
A Comparative Evaluation of Deep Deterministic Policy Gradient on Aircraft Maintenance Decision Optimization
Lect. Notes Electrical Eng.
International Conference on Aerospace System Science and Engineering ; 2024 ; Shanghai, China July 17, 2024 - July 18, 2024
Proceedings of the International Conference on Aerospace System Science and Engineering 2024 ; Chapter : 44 ; 448-457
2025-03-09
10 pages
Article/Chapter (Book)
Electronic Resource
English
Wiley | 2022
|