Auto-Design-Based Optimizations of Prestressed Frames Using an Artificial Neural Network (ANN)-Based Hong-Lagrange Algorithm

Structural designs using artificial neural networks (ANNs) are relatively new even though the ANN has been an evolutionary technology in many areas, such as automotive, financial, and medical areas. The ANN-based Hong-Lagrange algorithm shows prospects for real-world applications, providing an auto design and optimizations that substantially increase design effectiveness while considerably reducing human efforts. Holistic designs of prestressed multi-story frames are performed in this study. Big datasets following American standards were generated based on the ABBA (auto-design-based building applications) frame generator, which was developed and verified by commercial software (ADAPT builder and MIDAS). ANNs are trained on big datasets, mapping 38 input parameters to 111 output parameters, to derive weight and bias matrices for the formulation of ANN-based objective and constraint functions of piperack frames. Costs, CO₂ emissions, weights, and energy consumption are minimized by applying sequential quadratic programming (SQP) to ANN-based Lagrange functions constrained by equalities and inequalities, which are imposed by architectural and code requirements. The present study provides an application based on ANN technologies for structural designs, which not only enhances the design efficiencies but also the sustainability of building construction by reducing CO₂ emissions and energy consumption.