A hierarchical cluster approach for forward separation of heterogeneous fault/slip data into subsets

A new simple method of stress inversion uses hierarchical cluster analysis for forward separation of heterogeneous fault/slip data into subsets. Fault/slip data are classified into homogeneous fault classes, and a clustering routine classifies these into subsets. The method includes a way of discarding some residual data at the first stage that makes it fairly easy to recognize and eliminate some spurious fault data. However, this method is a type of hard division that overlooks the indeterminate nature of fault data. The more heterogeneous the data, the larger the calculation needed to find from a K-data set the homogeneous fault class that agglomerates a pair of 5-data subsets, sampled in a binomial distribution, with the maximum similarity in estimated stress vector between them. The K-data set is a working data group successively taken from the whole data. Given P phases of different stress state, the minimum value of K is 5P+1. Results from applying the method to two examples, artificial and real, demonstrate the feasibility of the method.