Optimizing UAV Electric Propulsion Systems using Artificial Neural Networks: A Data-Driven Approach
Please login to view abstract download link
In the rapidly evolving field of unmanned aerial vehicles (UAVs), the evaluation of electric propulsion systems plays a crucial role in determining overall vehicle performance. Accurate evaluation of these propulsion systems is essential for optimizing UAV design and operational effectiveness. Traditionally, such evaluations have relied on extensive experimental testing, which, while effective, can be time-consuming, costly, and limited in scope. To overcome these challenges, the integration of artificial neural networks (ANNs) has emerged as an innovative approach, enabling more efficient and accurate performance predictions. This study explores the application of an ANN-based model for evaluating UAV electric propulsion systems, highlighting its potential to revolutionize the way propulsion performance is analyzed. The ANN architecture consists of an input layer that receives data related to various propulsion parameters, a hidden layer that processes these inputs, and an output layer that delivers predictions, such as thrust force. One of the key contributions of this study is the exploration of the hidden layer configuration. By varying the number of neurons, the study identifies the optimal architecture that closely matches experimental results. This process allows for the fine-tuning of the neural network, ensuring that it can learn the complex, nonlinear relationships inherent in UAV propulsion systems. The training data used in this study comes from a comprehensive set of experimental tests performed on various UAV propulsion systems. By feeding this data into the ANN, the model learns to predict thrust force under various scenarios. Once trained, the neural network is tested on unseen data to evaluate its predictive accuracy. The results of this study demonstrate that the ANN-based model can accurately predict propulsion system performance, closely matching experimental data even for scenarios that were not part of the training set. The neural network's high accuracy validates it as a reliable tool for UAV propulsion evaluation, reducing the need for physical testing and streamlining optimization of performance metrics.
