Prediction of Building Condition after Earthquake

Abstract

 Artificial neural network (ANN) is one branch of artificial intelligence that mimics the workings of the human brain nerve. ANN is able to predict of output from the input data set that previously had done the learning process. In the event of an earthquake, a building will suffer damage either safe, moderate, or collapse. After an earthquake, a building need to be audited to determine its status or damage index if it is safe (Immediate Occupancy), moderate safe (Life Safety), or damage (Collapse Prevention). In the manual system, it takes a long time because the building experts will gather data before determining the condition of the building that is it still worth or not. This paper explains the design and testing of neural networks application to predict the condition of the building using the data obtained from the simulated earthquake on a building using Finite Element Analysis software. The ANN architecture was designed using feed forward and back propagation algorithm to decrease the error and set the weight in specific iteration, furthermore it is trained with 835 data. The results show that the prediction accuracy of ANN application to the 100 testing data is 92 % with 16 input neurons, 32 hidden neurons, 1 output and 0.1 learning rate. 

Keywords: ANN, Building, Prediction, damage index