Predictive model for the horizontal displacement of a dam using autoregressive neural network

The interpretation of data gathered from dam monitoring directly influences the detection of abnormal behaviors. Using previously recorded data, predictive models can be developed, so that the signs of a possible failure are detected as early as possible. The paper presents a multi-step ahead predictive model to generate the values for the horizontal displacement of a dam, using previous values of the displacement, water level and temperature. The model is based on an autoregressive neural network that was trained and tested using historical data. The results show a good prediction accuracy (maximum 2.63% relative errors), especially for up to 8 months ahead prediction.