Artificial Neural Networks for the Assessment of Structural Response

The following video highlights the procedure utilized in the work of Bakas, Makridakis and Papadrakakis, Torsional parameters importance in the structural response of multiscale asymmetric-plan buildings.  This work’s approach constitutes the creation and utilization of a large database of multiscale structures in order to derive statistically reliable numerical results for the various torsional criteria, structural response parameters, and their modeled relationships. Thus, in the initial step of this study, the structure’s database was created using a random number of floors, bays in X and Y direction, and columns sizes. In order to perform the structural analysis and design for the generated population of structures (several hundreds of instances) and to derive necessary response parameters such as eccentricity of each one design, a random structures generator algorithm was implemented and linked to the structural analysis and design software ETABS. For each step of the random geometry generation algorithm, the dimensions of the vertical elements were changed as well as the number of floors and bays, and the corresponding structural response parameters were calculated and saved. Afterwards, the derived results of the structural analysis were utilized to construct the regression model with linear relationships as well as artificial neural networks.


%d bloggers like this: