Evaluation of displacements of steel structures under earthquake effects by artificial neural networks

The design of a building includes a time-consuming process such as dimensioning, soil analysis, structural system selection, material determination, three-dimensional modelling and analysis, revisions and modifications when necessary. The use of artificial intelligence is a technology that can minimise time losses by accelerating this process. The purpose of this research is to investigate the use of this technology in this field. For this reason, firstly, our sample steel structures were determined as steel moment resisting frames. A total of 64 examples were created by using four different floor areas, two different storey heights and eight different number of storey combinations The sample structures were modelled with the SAP 2000 program and linear static analysis was performed under earthquake effect. The maximum displacement values of the sample structures analysed with SAP 2000 were determined as the output set of the artificial neural network (ANN). The number of storeys, storey height, total height of the structure, floor area and weight of the structure were determined as input set as some parameters that will affect the maximum displacement. The ANN model with input and output data set was created with the Matlab program. As a result of the research,comparison and regression analysis were conducted between the predicted values obtained from the ANN model and the values calculated using SAP 2000. In the comparison and regression analyses, it was observed that the values predicted by the ANN were very close to the zero error line, the relative error rates were low and the regression coefficients were very close to 1. The positive results obtained in this research with only one sub-branch of artificial intelligence show the hope of positive results in other sub-branches, and it is predicted that different and innovative scenarios may emerge in the project phase of civil engineering in the future by using more than one sub-branch together.