HGA: A genetic algorithm method for direct estimation of paleostress states from heterogeneous fault-slip data

Abstract

Field data on fault-slip observations is commonly heterogeneous. Paleostress estimation from such data sets is, in general, carried out in two steps: (i) the classification of the heterogeneous data set into homogeneous subsets and (ii) an inversion of each homogeneous subset. This study gives a new approach, the HGA, that combines the two issues in a single step process and gives the stress tensors directly. The given heterogeneous data are directly operated upon by the genetic algorithm operators, initialization, elitism, selection, encoding, crossover and mutation. These operations simulate such a guided search that finds successively fitter solutions, the stress tensors, until the globally fittest solution is obtained. We first explain the basic steps of the algorithm on a working example and then demonstrate its veracity using several synthetic and two natural examples.The proposed genetic algorithm method obviates the necessity of having first to classify the heterogeneous data into homogeneous sets. It directly estimates different stress states by inversion of the given heterogeneous fault-slip data. In contrast to the existing linear methods, the method is not vulnerable to entrapment of the solution in a local optimum. Although the method requires an a priori estimate of the maximum number of expected homogeneous sets in a given population, this estimate does not control the final results. Like any other method, the genetic algorithm method too has its merits and limitations and these are discussed.