Multi-objective optimization of aviation dry friction clutch based on neural network and genetic algorithm

To improve the flexibility and stability of high-power aviation dry friction clutches, aiming at engagement time and impact torque, a clutch model was constructed using UG software. Orthogonal experiments were designed and dynamic simulation was conducted using Adams software to generate sample data. The effects of friction coefficient, logarithm of friction plate, pressing force, and loading time of pressing force on the coupling time and impact torque were also analyzed. Particle Swarm Optimization was used to optimize a Back Propagation neural network to establish a predictive model with the clutch engagement time and impact torque as targets. Then, the Non-Dominated Sorting Genetic Algorithm II was used to find Pareto solutions to obtain the optimal design parameters for the clutch. The optimized engagement time was reduced by 1.2 s (24%) and the impact torque was reduced by 75.6 N·m (18%). The simulation results show that the average relative errors between the predicted values and the actual values of the engagement time and impact torque are 0.61% and 2.76%, respectively. The average errors of the engagement time and impact torque optimized by NSGA-II are 1.00% and 5.12%, respectively, indicating that this method can be effectively applied to the design of aviation friction clutches.