Author(s): S. M. Ali-Zomorodian; S. Alinaghizadeh; M. Derogar
Linked Author(s):
Keywords: Embankment dams breach; BREACH mathematical model; Artificial neural networks; Peak breach outflow; Breach time; Mollasadra dam
Abstract: There are a number of models have been developed and employed over the years to estimate the effects of a dam breach and the resultant flood. These have included physical scale models and mathematical or computer models. Among the more widely used dam breach computer models over decades is the dam-break flood forecasting model (DAMBRK). DAMBRK simulates the outflow from a reservoir and through the downstream resulting from a developing breach in a dam. It is based on erosion, soil mechanic equations, hydraulic laws and sediment transportation. As such, the model is not one that can easily be employed this difficulty led to use other powerful methods. In this study a new method has been developed for prediction peak breach outflow and breach time by Artificial Neural Networks (ANNs). For this purpose synthetic breach parameters of about 115 dams were developed by DAMBRK model and they were used to train and test the neural networks. The performance of the network models is investigated by changing input parameters. The most efficient and global model for assessing dam breach potential and the most significant input parameters affecting dam breach are summarized. Best results were found with back propagation neural networks using multiple hidden layers. A forecast study is performed for the Mollasadra dam. Comparisons between the artificial neural network results and dam DAMBRK model are made. The results indicate that neural networks are useful for predicting dam breach parameters.
Year: 2005