Experimental modelling of propulsion transients of a brushless DC motor and propeller pair under limited power conditions: A neural network based approach

This paper presents an experimental framework to describe the dynamic behavior of brushless direct current (BLDC) motors, which are frequently used in unmanned aerial vehicle (UAV) applications. Typically these applications require varying the angular speeds of motors which have to be precise parts of the dynamic model. The experimental setup is a motor/propeller pair equipped with an electronic speed controller. The substantial contribution of the paper is to incorporate the voltage drop caused by the consumed power and the alleviation of the modulation effect on the measured battery voltage. The consequence of the voltage drop is to obtain different angular speeds under same excitation inputs. A Neural Network (NN) based approach is chosen to handle this modeling issue. Levenberg-Marquardt algorithm is used to tune the adjustable parameters of NN, which is trained offline using the data observed through a set of experiments. Some experimental validation results are presented to justify the model.