Predictive Maintenance Planning For Batteries Of Electric Take- Off And Landing (eVTOL) Aircraft Using State-of-Health Prognostics

Abstract

Electric vertical take-off and landing (eVTOL) aircraft are a futuristic, sustainable transportation mode aimed at reducing traffic congestion. The health management of eVTOL batteries is key for the deployment of such aircraft. In this paper, we consider the continuous monitoring of eVTOL batteries, with streams of measurements related to the charging, discharging, and temperature of the batteries. Based on these measurements, we develop a Convolution Neural Network with Monte Carlo dropout to estimate the distribution of the State-of-Health (SOH) of the batteries, i.e., we develop probabilistic SOH prognostics. The features used for the SOH estimates are selected based on the feature importance quantified by Shapley values. The obtained probabilistic SOH prognostics are further employed for the maintenance planning of the eVTOL batteries. The results show that our approach leads to accurate SOH estimates. Moreover, we are able to identify optimal times for eVTOL battery replacement, as a trade-off between the cost of unexpected failure and the cost of wasted battery life.