Conceptual Design Method for Bridges Combining Topology Optimization and Deep Generative Model

Key Points

• The aim is to create diverse and structurally sound bridge designs using advanced modeling techniques.
• Implemented level set method for 2D topology optimization to obtain varied bridge designs.
• Created a labeled dataset including maximum displacement for performance assessment.
• Utilized conditional variational autoencoder to learn from the dataset and generate design proposals.
• Some directly generated designs did not meet target performance levels.
• By applying finite element method, designs close to specified maximum displacement were successfully identified.
• The approach demonstrates the ability to produce diverse and structurally valid bridge design options.

Abstract