Regression Models for Validating Cyclic Triaxial Test Results

Abstract

Abstract Cyclic triaxial tests are commonly used to evaluate a soil’s potential for liquefaction during seismic events. Following testing, it is necessary to evaluate the validity of the test results. This can be a difficult task unless the engineer has previously acquired a reasonable amount of experience with this type of testing. In order to help evaluate the validity of the test results, a series of eight predictive models have been developed for sands and soils with non-plastic silts. These models predict a mean value of the number of cycles required to cause liquefaction as well as 50 and 95 % prediction intervals. The models were separated by the soil type (sands and soils with less than 30 % non-plastic fines versus soils with 30 % or more non-plastic fines), the definition of liquefaction used (initial liquefaction versus 5 % double-amplitude strain), and the method of specimen preparation (moist tamping, air-pluviation, or slurry-deposition). These models were developed using the data from over 1000 tests collected from the author’s files and the literature. Following development, the validity of each model was assessed by examining the statistical parameters of the model, analyzing each model’s residuals, and by predictions made using additional data obtained from the literature.