Font Size: 

Aiming at reducing the computational time of structural topology optimization, a non-iterative topology optimization algorithm is proposed by deep learning. A generative adversarial network (GAN) is developed to create the high-dimensional mapping from the noise to the optimized configuration. Furthermore, a convolutional neural network (CNN) is utilized to evaluate the predicted configuration generated by GAN. Training dataset composed of the optimized configuration and the corresponding boundary condition is obtained after simulating a large number of topology optimization procedures based on independent, continuous and mapping (ICM) method. Numerical experiments show that the networks can accurately estimate the size of topology configuration and constraint parameters with negligible computational time. The method provides a new solution for topology optimization design to improve the computational efficiency of topology optimization and broaden the application of ICM method.

Topology optimization, Deep learning, ICM method