Abstract:Abstract Topology optimization is one of the most flexible structural optimization methodologies. However, in exchange for its high level of design freedom, typical topology optimization cannot avoid multimodality, where multiple local optima exist. This study focuses on developing a gradient-free topology optimization framework to avoid being trapped in undesirable local optima. Its core is a data-driven multifidelity topology design (MFTD) method, in which the design candidates generated by solving low-fidelity topology optimization problems are updated through a deep generative model and high-fidelity evaluation. As its key component, the deep generative model compresses the original data into a low-dimensional manifold, i.e., the latent space, and randomly arranges new design candidates over the space. Although the original framework is gradient free, its randomness may lead to convergence variability and premature convergence. Inspired by a popular crossover operation of evolutionary algorithms (EAs), this study merges the data-driven MFTD framework and proposes a new crossover operation called latent crossover. We apply the proposed method to a maximum stress minimization problem in 2D structural mechanics. The results demonstrate that the latent crossover improves convergence stability compared to the original data-driven MFTD method. Furthermore, the optimized designs exhibit performance comparable to or better than that in conventional gradient-based topology optimization using the P-norm measure.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is in structural mechanics, how to avoid getting trapped in local optimal solutions during the topological optimization process, especially when dealing with multimodal optimization problems with high nonlinearity. Specifically, the paper focuses on developing a gradient - free topological optimization framework to overcome the problem that traditional gradient methods may be trapped in low - performance local optimal solutions. The core of this framework is a data - driven multi - fidelity topological design (MFTD) method. It uses design candidates generated by low - fidelity optimization and combines deep - generation models and high - fidelity evaluations to update these candidates. To further improve the convergence stability and diversity - maintaining ability of this framework, the paper introduces a new crossover operation - latent crossover, which is inspired by the crossover operation in evolutionary algorithms. By applying latent crossover to a maximum - stress - minimization problem in two - dimensional structural mechanics, the research shows that compared with the original data - driven MFTD method, latent crossover improves the convergence stability, and the performance of the optimized design is equivalent to or better than that of traditional gradient - based topological optimization methods. ### Summary of the Core Problems in the Paper - **Objective**: Develop a gradient - free topological optimization framework that can effectively avoid getting trapped in local optimal solutions. - **Method**: Propose a data - driven multi - fidelity topological design (MFTD) method and introduce the latent crossover operation. - **Application**: Apply this method to the maximum - stress - minimization problem in two - dimensional structural mechanics. - **Result**: Latent crossover improves the convergence stability of the framework, and the performance of the optimized design is better than or equal to that of traditional gradient - based methods. ### Explanation of Formulas - **Latent Variable Vector \( z \)**: \[ z=\mu+\sigma\odot\epsilon \] where \(\mu\) is the mean vector, \(\sigma\) is the variance vector, \(\epsilon\) is a vector of random numbers drawn from a standard normal distribution, and \(\odot\) represents element - wise multiplication. - **VAE Loss Function \( L_{\text{VAE}} \)**: \[ L_{\text{VAE}} = L_{\text{recon}}+\varrho L_{\text{KL}} \] where \( L_{\text{recon}} \) is the reconstruction loss, usually using the mean - squared error (MSE), \( L_{\text{KL}} \) is the Kullback - Leibler divergence, and \(\varrho\) is a weighting parameter that controls the influence of the KL divergence. - **Kullback - Leibler Divergence \( L_{\text{KL}} \)**: \[ L_{\text{KL}}=-\frac{1}{2}\sum_{i = 1}^{N_{\text{lt}}}\left(1+\log(\sigma_i^2)-\mu_i^2-\sigma_i^2\right) \] where \( N_{\text{lt}} \) is the dimension of the latent space, and \(\mu_i\) and \(\sigma_i\) are the \( i \) - th elements of the mean vector and the variance vector respectively. - **Simplex Crossover (SPX) Algorithm**: - **Select Parent Individuals**: \[ P_0, P_1,\ldots, P_n \] - **Calculate Centroid \( G \)**: \[ G=\frac{1}{n}\sum_{i = 0}^{n}P_i \] - **Calculate Variables \( x_k \) and \( C_k \)**: \[ x_k = G+\epsilon(P_k - G) \] \[ C_k=\begin{cases} 0&(k = 0)\\ r_{k - 1}(x_{k - 1}) \end{cases} \]

Latent Crossover for Data-Driven Multifidelity Topology Design

Enhancing Topology Optimization with Adaptive Deep Learning

Maximum stress minimization via data-driven multifidelity topology design

Data-Driven Topology Optimization With Multiclass Microstructures Using Latent Variable Gaussian Process

Data-driven topology design using a deep generative model

Diffusing the Optimal Topology: A Generative Optimization Approach

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

A New Topology Optimization Methodology Based on Constraint Maximum-Weight Connected Graph Theorem

Algorithmically-consistent deep learning frameworks for structural topology optimization

Data-Driven Multiscale Topology Optimization Using Multi-Response Latent Variable Gaussian Process

A Hybrid Method for Fast and Robust Topology Optimization

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

Gradient-free neural topology optimization: Towards effective fracture-resistant designs

Real-time topology optimization via learnable mappings

Data-driven topology design based on principal component analysis for 3D structural design problems

Differential evolution-based topology optimization using Karhunen-Loève expansion

Structural Topology Optimization with an Adaptive Design Domain

Real‐time topology optimization via learnable mappings

Universal machine learning for topology optimization

Data-Driven Multiscale Design of Cellular Composites with Multiclass Microstructures for Natural Frequency Maximization

A survey of structural and multidisciplinary continuum topology optimization: post 2000