Abstract: A dam settlement prediction model integrating BP neural network model and Markov chain prediction was built in this paper. Through emulating the training samples, rolling prediction for the settlement displacement time series was performed by the metabolism-improved BP neural network algorithm. Furthermore, Markov chain was used to correct its random disturbance and the prediction results were improved. This model was applied to the settlement displacement timing prediction of Changzhou dam lock control building. The result shows that the model has high prediction accuracy and good reliability. It improves the long-term prediction ability, and provides an effective method for dam settlement prediction.

Key words: settlement prediction, BP neural network, Markov chain, dam monitoring, Changzhou water power junction