Behaviour model for dam displacement derived by an evolutionary algorithm

Data recorded by the behaviour surveillance system of a large dam are used,among other things,to develop some forecast models for various state parameters which are important for the structural safety.In this paper,a nonlinear behaviour model of regression type is derived using a genetic algorithm based method.The upstream-downstream displacement at the top of the dam is selected as an example of state parameter and the measured pendulum deviations are considered as output data.The reservoir water level and air temperature are accepted as external influence factors,but theirs averaged values on different previous time-periods(7,15,30 and 60 days)are well used.Data sets with 332 values for each input /output parameter were prepared starting from the recorded data at Herculane dam,a 58 m tall and 190 m long arch type dam,between January 2000 and June 2007.The data with even numbers of these series were used to obtain the regression coefficients and all data were then considered for validation.The same procedure may be easily adapted to obtain behaviour models for any other state parameters.These models can then be used to interpret the data recorded in operational surveillance and to quickly identify some situations with potential risks for the dam safety.