CKI-25: SEISMIC RESPONSE OF UNDERGROUND STRUCTURES IN SOFT SOILS - CYLINDRICAL SHAFTS IN DRY SAND NUMERICAL SIMULATIONS OF CENTRIFUGE TESTS
• R. F. Scott
It had been originally proposed in January 1991 to meet Kajima's requirements for studies on cylindrical shafts embedded in soft ground that Caltech perform centrifuge tests on an instrumented model shaft and follow these with a limited number of numerical evaluations of the test data. However, when Kajima examined the proposals for the current fiscal year, they decided that in view of the fact that their own centrifuge was close to operating, they would prefer to carry out the centrifuge tests themselves, and give a contract to Caltech to provide numerical simulations of the test data and to evaluate these simulations for correspondence with the actual cylinder.
The cylinder was to be embedded in dry sand in a box mounted on the centrifuge and operated at 50 g. The box would be subjected to a horizontal base motion simulating the northsouth component of the El Centro 1940 earthquake. The shaft was to be made of aluminum and constructed so that the bending properties of the shaft (EI) simulated a full-scale reinforced concrete shaft 6 meters in diameter. It was to be instrumented with strain gauges, pressure transducers, and accelerometers; accelerometers were also to be placed in the sand at the base, inside the soil mass and at the soil surface to record the motions of the model during the input test.
The tests were carried out in the period July to October 1992 and the data were communicated by Dr. Honda of Kajima at a meeting at Caltech with R. F. Scott and B. Hushmand. During this period also, work was performed at Caltech on examining numerical models which would be used to represent the test data when they became available. An "engineering model" was constructed composed of a number of masses, springs and dashpots to represent the basic characteristics of the cylinder in sand problem, and a variety of finite element codes was selected from those available, either on the commercial market or from universities, in order to simulate the test motion. The finite element models consisted of ABAQUS, a commercially available program which has the ability to incorporate nonlinear soil behavior, the commercial program SAP90, which, in the form employed, has no nonlinear capability but can be used for plane strain equivalent linearized versions of the test. The third finite element program used was one named DYSAC2 which has been developed at the University of California at Davis and is a two-dimensional but fully nonlinear finite element code incorporating a soil constitutive model of the bounding surface type developed by Dafalias and Herrmann.
Computations have been performed with all of these models, with the engineering model being used for parametric studies to achieve the best match with the experimental data. The three-dimensional ABAQUS model presented many numerical difficulties in representing the dynamic soil behavior; they seem to be attributable to defects in the original program. Even linear solutions required long computational times. The SAP90 model was used for simulating a number of tests and proved an economical way of approaching the modeling, while the DYSAC2 representation also involved a great deal of computer time, and it was only possible to study the model behavior during a limited duration of input motion for both the low and high level earthquakes applied to the deep foundation.
2. Selected Experimental Results
3. Numerical Models
4. System Identification
5. Comparison and Results
6. Other Models
7. Conclusions and Recommendations
January 15, 1993