Colloquia

Samenvatting

In recent years, rapid advances have been made in the capabilities of unmanned aerial vehicles
(UAVs). Improvements in technologies such as the controller, motors and telemetry systems make
it more feasible to manufacture cheaper drones while increasing functionality [3]. This report
aims to obtain a digital twin that captures the complex dynamic behaviors in the shape of a
simulator. The digital twin incorporates quadcopter dynamics in the shape of equations of motion
(EoM), utilizes the bladed element momentum theory (BEMT) by using neural networks (NN). In
addition, the inflow angle is incorporated in the BEMT to increase the fidelity. Two digital twin
models are developed. The first model incorporates the RPM data of the propellers and based
on this, estimates the drone’s trajectory by using the Runge-Kutta scheme. In terms of stability,
a quaternion representation was used to represent the attitude of the vehicle. Also, to increase
the efficiency of the simulator, a NN that was trained based on BEMT data is used to replace
the computationally demanding BEMT. The second digital twin model uses a trajectory algorithm
in combination with the EoM to compute the total thrust and moments. Here, the BFGS cost
function is utilized and superposition in combination with a NN is used to obtain the propeller
RPMs. Physical experiments are performed with the actual quadcopter and measured by motion-
tracking cameras and the drone’s telemetry system. During the experimental campaign, horizontal
and vertical flight maneuvers is performed and its data is collected. By comparing the experimental
data with the first digital twin model, it is found that the model is capable of matching fairly well
with the data. However, it is not capable of capturing the propeller axis imbalances. For the second
digital twin model comparing it with the first model, the NN overpredicts the propeller performance
as it did not properly invert the BEMT. In general, it proved that the integration of the BEMT
and NN for a quadcopter digital twin is challenging as it leaves plenty of room for improvement in
multiple aspects.