Research On Calculation Method Of Data-driven Structural Topology Optimization

In this paper,a mechanistic data-driven approach is proposed to accelerate structural topology optimization,employing a finite element convolutional neural network(FE-CNN).Our approach can be divided into two stages: offline training and online optimization.During offline training,a mapping function is built between high-and low-resolution representations of a density field defined on a given design domain.This mapping is expressed by an FE-CNN that targets a common objective function value(e.g.,structural compliance)to be achieved by the density field at the two resolutions.During online optimization,a specific design domain with specific boundary condition with a high-resolution density field is reduced to a lowresolution field via the trained mapping function.The original high-resolution density field is thus optimized through computations that are only performed on the low-resolution field,which is then followed by an inverse mapping back to recover the high-resolution optimal density field.Our numerical examples demonstrate that this approach can successfully accelerate 2D compliance-minimization(optimization)by up to an order of magnitude in computational time.Our proposed approach shows promise in high dimensional density-based structural topology optimization.The present limitations of our approach are also herein discussed.