ABSTRACT:
This paper reports on the development of an intelligent algorithm based on ANFIS (Adaptive Neuro-Fuzzy Inference System), BPF (back propagation feed forward) and SOM (Self-Organizing Map - neural network) for the analysis of in-situ permeability in rock masses. Application of the algorithm is demonstrated using the Shivashan dam, Iran, as a case study. In the first analysis stage, it is demonstrated that knowledge and rules about relationships between model variables can be interpreted using ANFIS. In the second stage of analysis, after completing the prediction of lugeon “permeability values and aggregation of the measured and pre-dicted values, it is demonstrated that these data can be classified using SOM. In the prediction stage “Isolugeon” diagrams are generated, and in the clustering stage the rock volume under investigation is divided into three permeability zones.
1 INTRODUCTION In the design of dam structures, one of the most important issues is the detection of permeability variations in different levels of the dam site. However, prediction of permeability using existing data obtained from in-situ tests is a big challenge. This is especially important in the determination of potential water flow paths in the rock mass underlying a potential dam structure, and this has an important impact on the planning of grouting procedures (Houlsby 1990, Andrade 1988). Several different methods for assessing permeability variations in rock masses can be found in the literature; see Arhippainen (1970), Majdi et al. (2004, 2005) and Nakaya et al. (1997) for examples. In this paper, using several branches of computational intelligence (CI) theory, an algorithm to analyse permeability data will be presented and applied to a dam site in North Western Iran.
2 THE INTELLIGENT ALGORITHM TO PERMEABILITY ANALYSIS
Figure1 summarizes a general process to prediction and analysis of “lugeon data.
