Application of kriging as a geostatistical method and anns to permeability analysis at ag-chaie Dam, Iran

In the design of dam structure, one of the most important issues is detection of permeability variation in different levels of the dam site. Several different methods for assessing permeability variations in rock masses can be found in the literatures. In this paper potential of both artificial neural network (AANs) and geostatistics has been used to predict permeability whit regards to Lugeon test results at Ag-chaie dam site on northwestern of Iran. ANNs are computer systems composed of processing elements that are interconnected in particular topology which is the problem dependent (complexity of problem). In other hand, geostatistics is type of statistics that consider spatial dependence of values, hence regional variables of that case is used. we use Kriging as a geostatical method for prediction. The focus of this paper is on the Back Propagation (BP) network to predict permeability. Finally, the results of the both approaches are compared with each other. The comparison shows that the geostatistics has better prediction results.