ARTIFICIAL NEURAL NETWORKS FOR CONCRETE DAM MONITORING

The relatively high cost of installation and conservation of these systems force the designers to provide the minimum required number of instrumentations. However minimum requirements exist due to the redundancy especially for embedded instruments because of the cost of repair or retrofit. Artificial neural networks (ANN) as general tools for nonlinear system identification can help the designer to increase the redundancy and also to cover the total behavior of the structure. The ANN can simulate the global structural behavior and compensate for the lost data, damaged instrumentation and even little amount of these equipments. In this paper the application of ANN in the simulation of the structural and thermal behavior of large concrete arch dams is considered. The field data of pendulums and thermometers of two large dams in Iran for several years are employed to construct the neural networks. The networks are then used to predict the dam behavior for other periods of time. The results have been compared with the target field data and have shown a good accuracy and adaptation with them. Due to the high flexibility and simplicity it is recommended to use such tools as a complementary for dam monitoring and instrumentation.