基于机器学习的航空发动机导管CNC弯曲回弹预测及补偿

Aiming at the problems of serious springback and poor geometrical accuracy of aero-engine pipes during computer numerical control (CNC) bending process, the CNC bending experiments were conducted on 1Cr18Ni9Ti stainless steel tube with various specifications. The springback data of pipes during CNC bending were obtained with consideration of different tube diameter D, wall thickness t, ratio of bending radius to tube diameter R/D and bending angle θ. The springback prediction model was established by the machine learning method. The prediction accuracy of three algorithms of BP, RBF and Elman for machine learning was compared and the structure of artificial neural network model was optimized. The springback prediction and compensation of a full-sized aero-engine pipe were carried out according to the prediction model. The maximum angle deviation of pipe is 0.358° after CNC bending, which meets the design accuracy requirements.