Integration model of monitoring dam safety based on rough wavelet network

It is very difficult that a single model fits and forecasts accurately the relation between effects and the loads of a dam system, and monitors the safety of the dam. This paper proposed a new approach implementing the nonlinear integration of single models with wavelet networks. The calculations of all submodels were regarded as the inputs of wavelet networks, and the observations were the outputs of wavelet networks. Based on the time-frequency information contained in trained data, the original structure of wavelet network was determined by time-frequency analysis. Rough sets theory was used to minimize the redundancy existing in wavelet networks. Forgetting factors method was adopted to train the weight parameters of wavelet network. Finally, this paper implemented the optimization integration for the displacement models of one dam with the trained wavelet network. The numerical examples show that the proposed integration model has strong universalized capabilities and the ability to adapt for the changes of function.