Session: DAC-04-03: Data-Driven Design

Paper Number: 116574

116574 - On the Connectedness of the Topology Optimization Predictors 

Deep learning-based topology optimization predictors have been shown to be effective in generating optimal designs. However, these predictors are prone to topological errors, particularly for high-resolution domains. Although various methods have been developed to enhance the accuracy of predicted structures, such as using large training datasets, complex networks, and physic-based loss functions, these methods do not include topological information in deep learning models. Similar issues arise in other applications, such as vessel, neuron, or road segmentation from images, and several modifications to typical loss functions have been proposed to improve the topological validity of the predictions. However, these topological loss functions have not been explored in-depth for topology optimization models.
In this study, we evaluate and compare four distinct topological loss functions to explore their influence on the performance of deep learning-based topology optimization predictors. We provide a detailed analysis of the advantages and limitations of each loss function, including factors such as computational complexity. We also evaluate the effectiveness of each loss function in generating structures that are both accurate and connected. Our findings offer insights into the advantages and limitations of these modified loss functions and provide a basis for future research and development toward improving the accuracy and efficiency of deep learning predictors in topology optimization.

Presenting Author: MohammadMahdi Behzadi University of Connecticut

Presenting Author Biography: Mohammad Behzadi is a Ph.D. candidate in the Department of Mechanical Engineering at the University of Connecticut, where he works in the Computational Design Lab under the supervision of Professor Horea Ilies. He completed his Bachelor of Science in Mechanical Engineering at the Sharif University of Technology in Iran before joining UConn in 2018.
Mr. Behzadi's research lies at the intersections of topology optimization and artificial intelligence. He is primarily focused on developing data-driven models that use AI techniques to efficiently synthesize high-quality designs for topology optimization.

Authors:

MohammadMahdi Behzadi University of Connecticut
Horea Ilies University of Connecticut