استفاده از سري هاي زماني در شبكه هاي عصبي مصنوعي تكاملي به منظور ارزيابي آسيب پذيري در قاب بتني خمشي

Use of Time Series in Evolutionary Artificial Neural Networks to Evaluate the Vulnerability of Concrete Moment Frame

After an earthquake occurred, fast decisions about building safety, the possibility to continue building and exploiting a situation and determine the desired failure rate, is crucial. Today, new techniques using evolutionary artificial neural network model based on artificial intelligence is widely used in various scientific fields, especially structural and earthquake engineering has found. In this article a moment frame with 4 floors and 4 spans under thirty acceleration records from l.Og to 1.5g Nonlinear Dynamic Analysis and the amount damage of total frame in each record any acceleration are calculated then used to determine the damage models of evolutionary artificial neural network was paid. For determining the number and effective delay time data input earthquake artificial neural network models of cross correlation method time series was used. Also genetic algorithm used to optimize structure of EANN models view of point layers number, nodes number in hidden layer, type of transfer function and network learning algorithm. Results show that cross correlation method determine truly influence number and temporal lag of input data. In addition to with, comparing the output of MLP model as the best nonlinear regression model with the index Park - Ang can be concluded that the ability of the MLP model, have respectively more ability and flexibility.