Neural prediction of water absorption in the sealing process of a dam grout curtain

Marek Slonski and Zenon Waszczyszyn

Cracow University of Technology, Department of Computer Methods in Civil Engineering, ul. Warszawska 24,31-155 Krakow, Poland 
e-mail: {mslonski.zenwa}@twins.pk.edu.pl

Abstract:

The problem is formulated as a prediction of unit water absorption Q in secondary boreholes on the base of measurements of the unit water and cement absorption in primary boreholes. Having a predicted water absorption Q the injection of cement grout can be evaluated. The paper is associated with measurements carried out on the earth dam Klimk6wka in the South-Eastern Poland. The input and output data correspond to average values of measured unit absorption in three curtain layers. Back Propagation Neural Networks (BPNNs) and Fuzzy Weight Neural Networks (FWNNs) were used in each layer. It was stated that satisfactory crisp approximation was achieved in the deepest layer C. Application of FWNN s gave interval type predictions in all layers.

Keywords: dam grout curtain, layer of curtain, unit water absorption, Back Propagation Neural Network, Fuzzy Weight NN