Q-learning based maintenance decision of missile single component

As a precision strike weapon, the strategic position of missile equipment is increasing. Therefore, the reliability of the missile equipment has put forward higher requirements. However, the traditional periodic maintenance methods are increasingly unable to meet the maintenance and security needs of missile equipment, which has resulted in missile weapon systems facing low operational readiness and high security costs. In order to maintain the missile equipment in a good performance state for a long time, and reduce the investment in the maintenance cost of missile equipment during storage, a reinforcement learning-based method to optimize the maintenance decision of missile equipment is proposed. First, the future health state of the missile is predicted by the degradation pattern of missile components. Second, a Markov maintenance decision model of the missile is developed based on the degradation mechanism of missile components. Finally, with the maintenance cost of missile components during storage as the optimization objective, the Q-learning method is used to optimize the maintenance decision of missile components during storage. Through experimental simulations, the optimal maintenance cost of the missile single component—engine during storage and the optimal maintenance decision during storage are obtained.