Artificial neural network based hole image interpretation techniques for integrated topology and shape optimization

Homogenization based and density based topology optimization seeks the best conceptual structural configuration on a predefined design domain with specific boundary and loading conditions. Such structural configuration is most often the minimum-compliance design under a fixed material usage constraint. Shape optimization must be subsequently executed so as to ensure the satisfaction of other practical design constraints such as stress and displacement, and attain the detailed definition of the structure configuration with a smooth circumference and interior hole contours. Complicated procedures involved in connection between topology and shape optimization are major obstacle for most design engineers to overcome. A fully automated configuration optimization system was developed [C.Y. Lin, L.S. Chao, Automated image interpretation for integrated topology and shape optimization, Structural and Multidisciplinary Optimization 20(2) (2000) 125–137] to execute the entire configuration design process automatically with room of improvements in the hole representation templates and hole interpretation reliabilities. In response, this paper proposes two-stage artificial neural networks based hole image interpretation techniques with improved template variety and recognition reliability.