Electric Motor Chain Modeling Using Artificial Neural Networks and Semi-Empirical Methods

Abstract

The methodology presented in this paper is dedicated to a user who wants to design a power chain of an electric unmanned aerial vehicle. The power chain includes a brushless direct current (BLDC) motor, a propeller, and a battery. Three models are presented to predict the thrust, electrical consumption, mechanical torque, and rotational speed provided by each component of an electrical power chain. Only public data were used to design these models. The research was developed in order to be used by all people who want to buy electrical power chain components and therefore to model their properties. Neural networks set with MATLAB parameters were used to design BLDC motor output (rotational speed and mechanical torque) models. Empirical methods such as blade element theory (BET) and disk area were used to model propeller thrust. A battery discharge model was designed based on public datasheets. Very good prediction results were obtained for the rotational speed (0.20% of error) and torque (0.45% of error). Small thrust modeling errors of 1.06% were found with the BET method. Models presented in this research could be designed using a wide type of data, such as public data (provided on manufacturer websites) or real test bench measurements.