Inverse design of cellular structures with the targeted nonlinear mechanical response

Area of Science:

• Materials Science
• Mechanical Engineering
• Computational Design

Background:

• Advanced additive manufacturing allows complex cellular structures with tunable mechanical properties.
• Modulating unit cell topology influences material behavior, like stress-strain response.
• Designing printable structures for specific nonlinear responses is computationally challenging.

Purpose of the Study:

• To develop a data-driven generative framework for inverse design of cellular structures.
• To enable precise control over nonlinear mechanical responses through topology optimization.
• To create a computational tool for mapping desired material behavior to feasible 3D geometries.

Main Methods:

• Utilized a conditional variational autoencoder (cVAE) architecture.
• Trained the cVAE on a dataset of structure-property pairs.
• Explored decoder-only and encoder-decoder generation modes for design inference.

Main Results:

• The cVAE successfully learned a latent space for efficient structure-property mapping.
• Generated designs exhibited structural plausibility and mechanical accuracy.
• Predicted stress-strain curves closely matched target responses.
• The framework balanced geometric fidelity with functional performance under joint conditioning.

Conclusions:

• The proposed data-driven framework enables inverse design of cellular structures with targeted nonlinear mechanical properties.
• The cVAE approach offers an efficient method for exploring design spaces and generating printable, high-performance materials.
• This work advances computational materials design for additive manufacturing.