<![CDATA[Indonesia is one of the countries that located in quake zone. The potential of earthquake that could cause the damage to the building should be considered into the design of the building. Therefore, one of the solutions to analyzing the structural responses and the damage level promptly and easily when the earthquake occurred is with using Artificial Neural Network (ANN). The building model is a reinforced concrete building with 10 floors and height between floor is 3.6 m. Model building receives a load of earthquake based on earthquake time history of New Zealand, San Francisco, Cape Mendocino, El Centro, Loma Prieta, San Fernando, Kocaely, Nenana and Danieli. Each time history scaled to 0,5g, 0,75g, and 1,0g. In addition, this earthquake time history also scaled to the PGA of Pekanbaru. Artificial Neural Network (ANN) are designed in 4 architectural models using MATLAB program. Model 1 is ANN with the input of the displacement, velocity and, Model 2 is ANN with the input of displacement, Model 3 is ANN with the input of velocity, and Model 4 is JST with the input of acceleration. Output of the ANN is the damage level of building with the category of Safe, Immediate Occupancy (IO), Life Safety (LS) or in a condition of Collapse Prevention (CP). Artificial neural network trained with 8 data sets of earthquake time history, and the remaining 1 data set of earthquake time history was used for simulation. Artificial neural network models has the prediction rate to predict the damage level between 85%-95%. The results from simulation with the 913 data from time history of Danieli’s earthquake for all scales is 93,32% data could be predicted correctly by ANN Model 1. For the Model 2, Model 3, Model 4 in a row could provide prediction correctly up to 92,11%,91,46% and 87,62%. This indicates that artificial neural networks can predict the damage level of building with average accuracy 90,13%.Keywords: Artificial neural network, structural response, time history, damage level.]]>
