General Guidelines for Application of the Extended Subtraction Method in SASSI Soil-Structure Interaction Analysis

The computer program SASSI (1999) is widely used for seismic soil-structure interaction (SSI) analysis of nuclear structures. The program offers two main sub-structuring methods: the Direct method and the Subtraction method. These methods are the basis for computation of the impedance matrix as well as formulation of the equation of motion for typical SASSI SSI models. The difference between the two methods is in the size of impedance problem which is defined by the number of interaction nodes in the free-filed sub-structure model from which the flexibility matrix and subsequently the impedance matrix is computed frequency by frequency. The Direct method, developed in early 1980s as part of the original SASSI program imposes the displacement compatibility between the free-field substructure and the entire excavated soil volume through the common interaction nodes of the two substructures thus the method requires a large group of interaction nodes. However, the strict requirement of the displacement compatibility between the two substructures does make the Direct method capable of producing stable and accurate results within the range of frequencies limited by mesh sizes. In the Subtraction method, the compatibility of the displacement is maintained for a smaller set of interaction nodes at the common boundary of the free-field substructure and the excavated soil. The limitation of the Subtraction method associated with maintaining the compatibility of the displacement at limited number of interaction nodes must be recognized and considered in the modeling and analysis. The accuracy of the solution by the Subtraction method can be improved significantly by extending the Subtraction method to so called “Extended Subtraction” method (ESM). The Extended Subtraction method is based on the experience of SSI analysis for several verification examples that can in effect reproduce the solution of the Direct method with substantially reduced computational effort.