Problem
Research questions and friction points this paper is trying to address.
shape optimisation
thin-shell structures
differentiable geometric representation
gradient-based optimisation
structural compliance
Neural parametric representations for thin-shell shape optimisation
🤖 AI Summary
This work addresses the challenge of shape optimization in thin-shell structures, which has been hindered by the lack of flexible and differentiable geometric representations. The authors propose a neural parametric representation (NRep) based on multilayer perceptrons with periodic activation functions, mapping parametric coordinates to physical coordinates. Treating the network parameters as design variables, the method minimizes structural compliance under a volume constraint via gradient-based optimization. This approach yields a compact yet highly expressive mid-surface geometry, enabling end-to-end differentiable optimization and extensibility to complex lattice-skin architectures. The framework successfully reproduces known optimal solutions across multiple classical benchmark cases, demonstrating its effectiveness and robustness.
📝 Abstract
Shape optimisation of thin-shell structures requires a flexible, differentiable geometric representation suitable for gradient-based optimisation. We propose a neural parametric representation (NRep) for the shell mid-surface based on a neural network with periodic activation functions. The NRep is defined using a multi-layer perceptron (MLP), which maps the parametric coordinates of mid-surface vertices to their physical coordinates. A structural compliance optimisation problem is posed to optimise the shape of a thin-shell parameterised by the NRep subject to a volume constraint, with the network parameters as design variables. The resulting shape optimisation problem is solved using a gradient-based optimisation algorithm. Benchmark examples with classical solutions demonstrate the effectiveness of the proposed NRep. The approach exhibits potential for complex lattice-skin structures, owing to the compact and expressive geometry representation afforded by the NRep.
Innovation
Methods, ideas, or system contributions that make the work stand out.
neural parametric representation
thin-shell optimisation
differentiable geometry
periodic activation functions
gradient-based optimisation
🔎 Similar Papers
No similar papers found.
