Validation framework for assessment of numerical predictions using leap experiments

Thumbnail Image
Goswami, Nithyagopal
Issue Date
Electronic thesis
Civil engineering
Research Projects
Organizational Units
Journal Issue
Alternative Title
The Liquefaction Experiments and Analysis Projects (LEAP) are a series of international collaborations to obtain high quality (centrifuge) experimental data of saturated soil systems and to use this data to validate and assess the performance of constitutive models and numerical tools used in soil liquefaction analyses. The collaboration involved researchers in the US (Rensselaer, George Washington University, UC Davis), Canada, China, Colombia, France, Italy, Japan, Korea, Taiwan, and UK. A validation effort was undertaken in 2015 and 2017 (referred to as LEAP-2015 and LEAP-2017) using a benchmark centrifuge model of a sloping deposit tested in a rigid-wall container. The deposit consisted of saturated Ottawa F-65 sand and was subjected to ramped sinusoidal excitations of 1 Hz frequency (including a Motion2 of 0.15g and Motion4 of 0.25g). LEAP-2015 consisted of 6 centrifuge tests (each in a different facility) whereas LEAP-2017 consists of 23 tests repeated more than once at each of the centrifuge facilities. Numerical simulations were conducted by a number of predictors. The predictors used different constitutive models for their blind predictions that were compared with the measured responses of sloping deposit specimens for calibration and validation. In this thesis, the LEAP-2015 centrifuge test accelerations were used to estimate and analyze the associated shear stress-strain responses using a simple system identification technique and compared with those of the numerical predictions (including an assessment of the effects of the rigid boundaries). It was verified that the soil stresses and strains within the central vertical zone of the deposit may reasonably be estimated using a vertical array of lateral acceleration and a simple shear beam model. The cyclic component of the stress-strain responses were found to be in fair agreement for most of the predictions and compared reasonably with experimental results. A difference metric was developed and used to quantitatively assess the differences between state variable such as accelerations, stresses, strains, etc. The differences among the target motion and achieved base acceleration during the 23 LEAP 2017 tests were quantified. These discrepancies (among the input motions) were expressed as a unique aggregate of four discrepancy measures associated with phase, frequency-shift, amplitude at 1 Hz, and amplitude of frequency components higher than 2 Hz (2+Hz). The amplitude of the 1 Hz frequency component and of the components with frequencies higher than 2 Hz (2+Hz) were concluded to be the main factors that varied among different centrifuge facilities. A sensitivity analysis was conducted for LEAP-2017 centrifuge tests to assess the effects of variability in input motion and soil properties on the observed response. The sensitivity of the acceleration response to differences in relative density of the tested deposits and input motion amplitude at 1Hz and 2+Hz, and was obtained using a Gaussian-process based kriging. The sensitivity was also obtained using the tip resistance of a CPT test (instead of relative density). The deposit response was found to be more sensitive to discrepancies in motion amplitude at 2+Hz. The discrepancy measure and sensitivity analysis was utilized to develop a new metric to assessment and ultimately validate numerical predictions.
School of Engineering
Full Citation
Rensselaer Polytechnic Institute, Troy, NY
Terms of Use
PubMed ID