Rapid and optimal design of a tail-sitter VTOL ducted fan using a neural network and PSO algorithm

Considering the optimal performance and new applications of the ducted fans, especially in UAV missions, this paper aims to provide an optimal and rapid method for designing the UAVs based on new mathematical and analytical tools which improved and accelerated many of the long engineered processes. In this design method, an initial design is carried out based on the momentum theory and the first size approximation of different sections of the duct fan, such as the inlet and outlet diameter, the power and the duct thrust, is determined. Then by connecting the MATLAB and a ducted fan design software called the ducted fan design code (DFDC), several optimal design schemes for the duct are extracted by the particle swarm optimization algorithm. The parameters search domain in the algorithm is obtained from the initial design with the Momentum theory method and the various results of DFDC software, in the case. Finally, in order to obtain the final duct design, according to the optimized information, a multilayer perceptron neural network using an error-back propagation algorithm is trained. In the redesign loops, without a time-consuming optimization, the trained neural model can extract the duct parameters very quickly, based on the constraints of structure, control design, and missions targets.