Abstract:We present a new framework for solving general topology optimization (TO) problems that find an optimal material distribution within a design space to maximize the performance of a structure while satisfying design constraints. These problems involve state variables that nonlinearly depend on the design variables, with objective functions that can be convex or non-convex, and may include multiple candidate materials. The framework is designed to greatly enhance computational efficiency, primarily by diminishing optimization iteration counts and thereby reducing the solving of associated state-equilibrium partial differential equations (PDEs). It maintains binary design variables and addresses the large-scale mixed integer nonlinear programming (MINLP) problem that arises from discretizing the design space and PDEs. The core of this framework is the integration of the generalized Benders' decomposition and adaptive trust regions. The trust-region radius adapts based on a merit function. To mitigate ill-conditioning due to extreme parameter values, we further introduce a parameter relaxation scheme where two parameters are relaxed in stages at different paces. Numerical tests validate the framework's superior performance, including minimum compliance and compliant mechanism problems in single-material and multi-material designs. We compare our results with those of other methods and demonstrate significant reductions in optimization iterations by about one order of magnitude, while maintaining comparable optimal objective function values. As the design variables and constraints increase, the framework maintains consistent solution quality and efficiency, underscoring its good scalability. We anticipate this framework will be especially advantageous for TO applications involving substantial design variables and constraints and requiring significant computational resources for PDE solving.

Machine Learning to Aid Tuning of Numerical Parameters in Topology Optimization

Enhancing Topology Optimization with Adaptive Deep Learning

A Novel Application Framework for Self-Supporting Topology Optimization

Universal machine learning for topology optimization

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

Machine Learning Powered Sketch Aided Design Via Topology Optimization

Algorithmically-consistent deep learning frameworks for structural topology optimization

Accelerating Large-scale Topology Optimization: State-of-the-Art and Challenges

Real-time topology optimization via learnable mappings

Real‐time topology optimization via learnable mappings

A Novel 3-D Topology Optimization Methodology Based on the Min-Cut Theorem

Problem-Independent Machine Learning-Enhanced Structural Topology Optimization of Complex Design Domains Based on Isoparametric Elements

Machine Learning-Driven Real-Time Topology Optimization under Moving Morphable Component-Based Framework

A Fourier neural operator-based lightweight machine learning framework for topology optimization

Problem-independent Machine Learning (Piml)-Based Topology Optimization—a Universal Approach

Neural Level Set Topology Optimization Using Unfitted Finite Elements

Machine Learning based parameter tuning strategy for MMC based topology optimization

Discrete Variable Topology Optimization Using Multi-Cut Formulation and Adaptive Trust Regions

Topology Optimization Through Deep Neural Network for Different Mechanical and Thermomechanical Problems

Deep learning accelerated efficient framework for topology optimization

A dynamic topology optimization design method based on multiple domain variable parameters