برآورد مشخصات ديناميكي سدهاي خاكي با استفاده از شبكه عصبي مصنوعي

Estimating the Dynamic Characteristics of Embankment Dams Using Neural Networks

In this paper, different kinds of neural networks have been described. Among these nets, Perceptron network have been presented in more detail due to its successful applications in various complex engineering problems. This study have been continued with brief description of using Perceptron nets in predicting response and material behavior of an earth dam during an earthquake. In order to identify the capabilities of Perceptron nets in function approximation, first fundamental period of 50 homogeneous earth dams was extracted from experimental Oner equation. These data have been used in training 5 different neural networks. Results emphasize the high ability of nets in functions approximation even with few data. Besides, a Perceptron net have been trained with first modal frequencies of 24 different earth dams, which are obtained from forced vibration tests, ambient vibration tests and earthquake records. Despite the shortage of available data, results of this study indicated the sufficient accuracy of net in predicting the first natural frequency of embankment dam. It should be mentioned that in some cases, the estimate natural frequency of dam bodies had more accuracy than some experimental equations