Abstract:

The prediction of structural reactions to big earthquakes is vital in giving warnings for potential damages early enough to minimize losses of life and properties. In the current study we describe buildings by fixed construction and environmental related parameters. Our models are based on real data of damaged buildings collected after the occurrence of three big earthquakes in Turkey. We extend our previous work to include the soil type for damaged buildings. We employ different techniques, namely neural networks (NN) and support vector machines (SVM) to improve the prediction accuracy. The results show that support vector machines, and in particular support vector regression gives better results compared to neural networks. Although we only used averages of soil type for each region, we observed that adding soil type has improved accuracy of predictions for building damages. It is to be noted that these types of predictions are important to ensure the serviceability and safety of existing structures. Our models are vital for the authorities to make fast and reliable decisions and can be also used to improve the development of new constructions codes. 

Keywords: