Uncertainty Quantification Working Group
July 25, 11:30 AM, CNLS Conf. Room, TA-3, Bldg. 1690

Using random fields to quantify the geometric shape uncertainties of spherical marine floats

Jason E. Pepin (ESA-WR) and Ben H. Thacker (Southwest Research Institute, San Antonio, TX)


To better understand both the predictive capabilities of engineering analysis codes, and enhance the analyst's confidence in those predictions, an integrated experiment and analysis project has been developed. The focus of this project is to generate precise probabilistic structural response simulations using numerical models of commercially available, spherical, stainless steel marine floats, under collapse loads, and compare with experimental results. The spherical marine float geometry was chosen because of its simple shape, yet highly complex nonlinear deformation behavior, leading to complex states-of-stress. There is also a wide variability associated with geometry and mechanical properties of commercially available (i.e., off-the-shelf) marine floats. The wide variability is not uncommon, and principally due to numerous forming processes, different operators, etc., which bulk production operations employ for a single material lot.

Actual collapse (i.e., buckling load) data from a set of 24 samples will be compared with numerical predictions using a full three-dimensional (3D) finite element model and the nonlinear structural dynamics code DYNA3D. Inherent anomalies and variations in geometry and material properties, which were collected from the set of test specimen, are included in the numerical model.

The probabilistic validation of the current work is performed using NESSUS. Variations in geometric shape parameters as observed from test data are characterized using random fields. Uncertainties in material properties (i.e., stiffness, strength, and flow) are also included in the probabilistic model based on statistical distributions derived from tensile and compression tests. Finally, the probabilistic numerical model is validated by comparison to the predicted and observed variation in collapse load and the probabilistic sensitivity measures are analyzed.

e-mail: jpepin@lanl.gov