The determining earthquake risk status of existing buildings with neural network

Turkey, with 95% of its land in seismic active zone has faced earthquakes that results in serious damages and casualties. It is an inevitable necessity to determine the earthquake risk status and secure of existing reinforced concrete building. Detailed analysis of existing RC buildings is time consuming. The fact that the importance of rapid assessment methods has increased in our country depends on the reasons mentioned above and the size of our building stock.There are many rapid assessment method applied in our country and these methods call for a great deal of experience in both application and assessment stage. This thesis aims to develop faster, more simple and alternative method that helps us determine the earthquake safety of existing buildings. For this purpose, ANN-based model was proposed with neural networks (ANN) approach, widely used in many areas in recent years and the reliability of this model has been checked by existing methods.ANN model's training and test sets have been established with the sample of buildings models that determined the seismic performance by using P25 rapid assessment method. Verification of the ANN model were made with a reference set of existing buildings have been exposed to 2003 Bingol earthquake.In some cases, to determine by entering into the existing buildings seems impossible. Therefore, the performance evaluation of existing buildings may be required with the data of buildings, which can be determined with street observation. For this purpose, two different ANN model is proposed. The proposed ANN based performance evaluation models are trained, tested and confirmed thanks to the reference buildings and the input data obtained through scanning the street.In conclusion, it points out the fact that ANN based performance assessment model presents fast and reliable results and can be used as an alternative method in determining the seismic performance of buildings.