Design of neural network modeler in reduced beam section based on optimized GA data
Nowadays, engineering solutions, that guarantee the safety of residents and at the same time pay attention to economic concerns, are proposed in structural engineering.structural Researchers have investigated connections with smaller cross-section in metal structures, demonstrating satisfactory performance in steel bending frames under various cyclic loads. A significant shortcoming of previous research has been the insufficient focus on cost-effectiveness. In addition, many studies have predominantly concentrated on specific types of connections or relied heavily on trial-and-error methods instead of optimizing the dimensions of these connections. This study presents the analysis and optimization of various joint types. Specifically, Reduced Beam Section (RBS) joints were modeled using ABAQUS software through dynamic analysis and numerical methods. The outputs from these analyses were then processed using a genetic algorithm (GA) in MATLAB. The optimal data generated by the GA served as suitable input for neural network modeling, facilitating design refinements. Utilizing two neural networks, the optimal length and cross-section of RBS connections were determined, resulting in more accurate design outcomes, enhanced design efficiency, reduced project execution time, and ultimately, cost savings.