Special topics on soil-structure interaction

Abstract

In the design of nuclear power plants, the topic of soil-structure interaction has traditionally been one of great theoretical and practical controversy for several reasons. In the beginning of the era of nuclear power, the science of soil-structure interaction was in its infancy and very little theory or observational data were available to the engineering profession. At the same time, the state of the art of structural dynamics, however, was already developed and engineers had a considerable amount of data on actual dynamic performance of structures, so that a great amount of confidence in the ability to perform a dynamic structural analysis was already in existence. It was only natural to extend the basic fundamentals of dynamic structural modeling to soil-structure interaction. Consequently, many ‘approaches’ and ‘theories’ of how to account for soil-structure interaction were erroneously developed without providing real consideration of basic fundamentals of interaction of a foundation with an elastic continuum, commonly referred to as elastic half-space or continuum solutions. At the same time, while great confusion was developing on how to apply elastic half-space theory to earthquake analysis problems, the dynamic finite element approach was born and was thought by many to be the final panacea to the problem of soil-structure interaction. Of course, as is usually the case, there is no perfect or single best approach to any problem and the end result is that each of the many approaches to account for soil-structure interaction, has its strong and weak points depending on site or structural conditions. This paper is intended to discuss some of the important details of soil-structure interaction theory to provide a common means of comparison and to introduce some new approaches to simplify the solutions of deeply embedded foundations. A review of recent literature on soil-structure interaction reveals several important facts. First, conflicting comparisons between the lumped parameter and finite element solutions apparently exist. In some cases, both approaches give similar results while for others the results vary widely essentially because different models are used for comparison of both methods. Secondly, fundamental errors are committed on finite element mesh size parameters and boundary conditions. Thirdly, misunderstandings on soil mass and foundation or structural modal damping lead to gross errors in the lumped parameter approach. Finally, the limitations of the various approaches are not always understood. It is noted that both the finite element and lumped parameter approaches should yield similar results if they are appropriately used to solve the same problems. A summary of the advantages and limitations of both approaches are presented and discussed with a short presentation regarding the state of the art in the determination of soil stiffness and material damping characteristics. Furthermore, the paper will illustrate one type of analysis technique which uses a hybrid approach of both finite element results and lumped parameter solutions. Such an approach is developed to account more accurately for the influence of embedment. Results using both pure finite element solutions and lumped parameter models show that the influence of embedment can be accurately considered even for deeply embedded structures (depth to width ratio equal to 1.0). The key to the approach lies in establishing the coupling between horizontal and rocking modes of foundation vibration. Once the coupling parameter is accounted for, it is then possible to develop a lumped parameter models that account for the variation of soil motions below the ground sufrace. Previous lumped parameter models have not accounted for these variations from the surface to the depth of the foundation. Details of the lumped parameter approach for embedded foundations and an illustration with numerical examples are provided. Recommendations are then presented on a procedure for soil-structure interaction of deeply embedded foundations. The primary advantages are that: (1) the model is more easily generated than a finite element model, (2) the model is less susceptible to modeling errors, such as mesh size, model size, and boundary influences, (3) parametric studies may be easily conducted since parameters such as damping may be more directly controlled, and (4) the computer costs for analysis are significantly reduced.