A deep learning method for fast predicting curing process-induced deformation of aeronautical composite structures

Abstract

Continuous fiber-reinforced composites are increasingly used in civil aviation for their superior mechanical properties and light weight. However, the process-induced deformation (PID) of composite structures has always been one vital problem during manufacturing. In particular, the PIDs of composite parts are significantly affected by the stacking sequences in the autoclave process. This work proposed an efficient prediction method to elucidate the PID cloud maps of composites with different stacking sequences by combining the finite element (FE) method and convolutional neural network (CNN). The accuracy of FE simulation was experimentally verified using rectangular laminates with 12 types of stacking sequences. Moreover, this deep learning method was employed to study the PID contours of the rectangular laminates and the tail rudder composite structure of civil aircraft. The results demonstrated that the accuracy of the PIDs assessment of the tail rudder structure is 96%, taking 9.67 s to complete the rapid prediction by the CNN model, compared with 6 min by only using FE simulation. Furthermore, the present method has universality for the cloud map prediction of composite parts with different design parameters such as geometric shapes, laying angles, and stack layers.