Abstract:Topology optimisation can facilitate engineers in proposing efficient and novel conceptual design schemes, but the traditional FEM based optimization demands significant computing power and makes the real time optimization impossible. Based on the convolutional neural network (CNN) method, a new deep learning approximate algorithm for real time topology optimisation is proposed. The algorithm learns from the initial stress (LIS), which is defined as the major principal stress matrix obtained from finite element analysis in the first iteration of classical topology optimisation. The initial major principal stress matrix of the structure is used to replace the load cases and boundary conditions of the structure as independent variables, which can produce topological prediction results with high accuracy based on a relatively small number of samples. Compared with the traditional topology optimisation method, the new method can produce a similar result in real time without repeated iterations. A classic short cantilever problem was used as an example, and the optimized topology of the cantilever structure is predicted successfully by the established approximate algorithm. By comparing the prediction results to the structural optimisation results obtained by the classical topology optimisation method, it is discovered that the two results are highly approximate, which verifies the validity of the established algorithm. Furthermore, a new algorithm evaluation method is proposed to evaluate the effects of using different methods to select samples on the prediction performance of the optimized topology, and the results were promising and concluded in the end.

Non-iterative Structural Topology Optimization Using Deep Learning

Enhancing Topology Optimization with Adaptive Deep Learning

Evolutionary Discrete Multi-Material Topology Optimization Using CNN-Based Simulation Without Labeled Training Data

Deep Super-Resolution Neural Network for Structural Topology Optimization

Generating Three-Dimensional Structural Topologies Via a U-Net Convolutional Neural Network

Accelerated topology optimization design of 3D structures based on deep learning

Deep learning-based topological optimization for representing a user-specified design area

A Multi-Objective Topology Optimization Methodology Using Deep Learning and Its Application to Electromagnetic Devices

Algorithmically-consistent deep learning frameworks for structural topology optimization

TO‐NODE: Topology optimization with neural ordinary differential equation

A Two-Stage Network Framework for Topology Optimization Incorporating Deep Learning and Physical Information

Real-time Topology Optimization Based on Convolutional Neural Network by Using Retrain Skill

Deep learning accelerated topology optimization with inherent control of image quality

Cross-resolution topology optimization for geometrical non-linearity by using deep learning

Investigation into the Topology Optimization for Conductive Heat Transfer Based on Deep Learning Approach

Deep learning accelerated efficient framework for topology optimization

Convolution Hierarchical Deep-Learning Neural Network Tensor Decomposition (C-HiDeNN-TD) for high-resolution topology optimization

A Deep Learning Approach for Efficient Topology Optimization Based on the Element Removal Strategy

Deep learning driven real time topology optimisation based on initial stress learning

Machine Learning for Topology Optimization: Physics-based Learning Through an Independent Training Strategy

Real-time Stress-Based Topology Optimization Via Deep Learning