The method of joint set identification using genetic algorithm was introduced. For handling of orientation data, the basic genetic algorithm was modified. We used real encoding scheme for the representation of candidate solutions and the orientation matrix for calculating mean direction of joint sets. The selection, crossover and mutation operations using real encoded chromosome were also implemented. Davies-Bouldin index and variance were used for cluster validity criteria. Finally, we developed GAC (Genetic Algorithm based Clustering), a FORTRAN program based on above algorithms and applied it to 3 different joint data sets. It is found that the results of joint set identification using GAC were acceptable for engineering design. From the application of GAC, we found that cluster validity index based on variance is more efficient in finding the number of clusters than Davis-Bouldin index. In addition, the genetic algorithm based clustering was proved to be a fast and efficient method for the joint set identification task.
The stability of structures in rock-mass, such as tunnel or rock slope is critically dependent on various characteristics of discontinuities. Therefore, it is important to survey and analyze discontinuities correctly for the design and construction of structures in rock-mass. One standard procedure of discontinuity survey and analysis is a joint set identification from a large unprocessed orientation data. Joint set identification is a tedious procedure, however is fundamental to rock engineering design such as rock mass classification, key block analysis, rock slope stability analysis and discrete fracture network modeling. Conventionally, manual method using contour plot had been widely used for this task, but this method has some short-comings such as subjective identification results, manual operations, and so on.
For these reasons, researchers introduced automatic joint set clustering techniques. Mahtab & Yegulalp (1982) developed clustering algorithm based on Poisson probability but the clustering shape is dependent on the initial value. Hammah & Curran (1998) and Jung & Jeon (2003) developed joint set clustering method using fuzzy k-mean algorithm which doesn't guarantee global or optimal solution.
In this article, joint set identification using genetic algorithm (GA) was introduced because GA provides higher probability to yield optimal solution. Basic genetic algorithm was modified for handling of orientation data and Davies-Bouldin index (DBI) and variance (VI) were used for cluster validity criteria. Finally we developed GAC, a FORTRAN program based on above algorithms and applied it to 3 different joint data sets.
GA belongs to a class of search techniques that simulates the principles of natural selection to develop solutions of large optimization problems. GA operates by maintaining and manipulating a population of potential solutions called chromosomes. Each chromosome has an associated fitness value which is a qualitative measure of the goodness of the solution encoded in it. This fitness value is used to guide the stochastic selection of chromosomes which are then used to generate new candidate solutions through crossover and mutation. Crossover generates new chromosomes by combining sections of two or more selected parents.