TY - JOUR JF - mdrsjrns JO - MCEJ VL - 17 IS - 5 PY - 2017 Y1 - 2017/12/01 TI - Sensitive index of damage in reinforcement concrete frames under seismic sequence using artificial neural networks TT - شاخص حساس به خسارت در قابهای بتن آرمه در معرض توالی لرزه ای با استفاده از شبکه عصبی مصنوعی N2 - In seismic active zones, large mainshocks usually follow by numerous aftershocks. Because of the short time intervals between consecutive shocks, additional damage due to the accumulation of inelastic deformations from all sequences is increased and the structures that has been already damaged by the preceding shock collapse before any repair is possible. Moreover, despite the importance of seismic sequence phenomena on increased damage and the evidence of structural damage caused by the recent multiple earthquakes such as Nepal and Hindu-Kush (2015), most structures are designed according to the modern seismic codes which only apply a single earthquake on the structure in the analysis and design process. In this case, the structure may sustain damage in the event of the "Design earthquake", and this single seismic design philosophy does not take the effect of strong successive shocks on the accumulated damage of structures into account. For this reason, the effect of various parameters such as Peak Ground Acceleration, Magnitude, Shear Velocity Wave, Effective Peak Acceleration, Peak Ground Velocity, Epicentral distance, the time gap between first and second earthquakes, Period of reinforced concrete frames and etc, is examined on the damage of reinforced concrete frames under single and consecutive earthquakes. At first, six concrete moment resisting frames with 3, 5, 7, 10, 12 and 15 stories, are designed according to the Iranian Code of Practice for Seismic Resistant Design of Buildings (i.e. Standard No. 2800 guideline) and analyzed under three different databases with/without seismic sequences phenomena. For each database, single and consecutive earthquakes are selected according to Peak Ground Acceleration (PGA), Effective Peak Acceleration (EPA) and Peak Ground Velocity (PGV) criteria from Pacific Earthquake Engineering Research (PEER) and United States Geological Survey’s Earthquake Hazards (USGS) centers. At next step, in order to train the multilayer artificial neural networks with back-propagation learning algorithm, period of reinforced concrete fames (T) and some of earthquake features including PGA, PGV, EPA, magnitude (M), shear wave velocity in the station (Vs), epicentral distance (Epc) and time gap between consecutive earthquakes (Tg) as artificial neural network inputs and Park-Ang (1985) damage index - as the results of nonlinear dynamic analysis in OpenSees software and neural network target – are selected. For each database, 400 neural networks are designed with a different number of neurons in each hidden layer from 1 to 20 and ideal neural network is determined with the least value of Mean Square Error (MSE) and maximum value of regression (R) among all networks. Then, for considering the effect of input parameters on structural damage (Park – Ang 1985) caused by single and consecutive seismic scenario, the range and reference values for each group of input parameters – single and consecutive cases in each database – are chosen to be close to the median values and introduce to ideal neural networks and damage indexes are determined. The results show that structural damage caused by with/without seismic sequence scenario is more sensitive than other parameters to Magnitude and Acceleration for single earthquakes and the ratio of these parameters in the second shake to first for consecutive shocks. SP - 169 EP - 180 AU - Ghodrati Amiri, G. AU - Rajabi, Elham AD - KW - Artificial Neural Network KW - Damage Index KW - Seismic Sequence KW - Nonlinear Dynamic Analysis UR - http://mcej.modares.ac.ir/article-16-6604-en.html ER -