A physics-inspired-data-driven model of eVTOL in-ground-effect thrust based on meta-learning and BEMT

Abstract Electric vertical take-off and landing (eVTOL) aircraft with ducted fans have high efficiency, high reliability, and low noise, and have great potential to become the majority of future urban air mobility. Nonetheless, the ground effect will cause fluctuations in ducted fan thrust during take-off and landing conditions. Conventional physical models fail to reflect its transient and non-linear nature. In this paper, a physics-inspired-data-driven (PIDD) model for predicting the in-ground-effect (IGE) thrust of ducted fans is presented. The total thrust is treated as the composition of average thrust and transient thrust. For the average part, hypotheses of ideal twist and exponential function are adopted based on the structure of blade element momentum theory (BEMT). For the transient part, a classification task based on cross entropy is added to the meta-learning algorithm. Altitude related and altitude unrelated items are constructed to obtain the PIDD model and enhance its generalization ability. The features of the model are rotary speed, voltage, and current, while the label is IGE thrust. Few-shot training indicates that the proposed model can predict transient thrust accurately. With mean error on training and testing sets is 1.1% and 0.7% respectively, the PIDD model outperforms the conventional physical model by 4.6% and 1.7% and the data-driven model by 1.5% and 2.7%. The PIDD model constructed succeeds in predicting transient IGE thrust and provides new ideas for relevant research.