Deep Neural Network Based Optimization of Individual Blade Pitch Control Input Scenarios for Helicopter Rotor Vibration Reductions
摘要
This study attempts to minimize vibration while maximizing the aerodynamic performance of a medium-class utility helicopter rotor at 140 knots, using optimized multiple harmonic Individual Blade pitch Control (IBC) inputs combining 2P and 3P actuations. A rotorcraft comprehensive analysis code, CAMRAD II, is used to develop the aeromechanics model of the main rotor using IBC. For the design optimization to search for the best IBC input scenarios, a deep neural network (DNN) model is constructed based on the parametric results using CAMRAD II to predict the rotor Vibration Index and rotor power. The best IBC pitch amplitudes and the control phase angles of the 2P and 3P actuations are determined by a design optimization framework combining the DNN and NSGA-II. The optimized IBC input of θIBC = 2P/1.12°/276.21° + 3P/0.77°/172.73° (Case 1) minimizes the rotor vibration by 63.44% while slightly increasing the rotor aerodynamic performance by 0.03%. The optimized IBC input of θIBC = 2P/1.03°/240.84° + 3P/0.50°/182.69° (Case 5) maximizes the rotor aerodynamic performance by 0.31% while reducing the rotor vibration by 31.36%.