Time series model for the permeability of concrete dams and their bedrocks using a neural network optimisation method

摘要

On the condition of specific hydrogeology and circumstance, the complicated seepage flow field occurs in the concrete dam and its foundation, and it shows a gradual transformation process under the function of interior and exterior factors. Among the factors, interior ones act as a deterministic role on the dam seepage flow safety, and the exterior ones take the positive and stimulative action on the formation and evolvement of complicated seepage flow field. Because of the significance of permeability of dam foundation to the seepage flow field, it is necessary to establish its time series model in order to reflect the evolvement process and evaluate the safety status of seepage flow field. To do this, firstly the relationship of seepage flow ingredients varying with permeability coefficients and water level is demonstrated. Then, combing with seepage prototype observation data, using the dam seepage flow conditions in different operation periods, the permeability coefficients in different time are inversed of impervious structure in dam and its foundation by way of finite element method and artificial neural network optimization method. Accordingly, the time series model of permeability coefficients, which can implicitly characterise the evolvement rule of complicated seepage flow field, is established. Finally, the feedback analysis of seepage flow field is made through the model. The practical analysis results show that the time-serial model is reasonable and feasible for analysing the evolvement rule of permeability coefficients, and has higher precision.