Parameter Estimation of Joint Models Using Global Optimization

Nonlinear joints and interfaces modeled with a discrete four-parameter Iwan element are defined by parameters that are often unknown a priori or require calibration to get better agreement with test data. While this constitutive model has been validated experimentally, its drawback lies in the difficulty of identifying the correct coefficients. This work proposes a parameter estimation approach using a genetic algorithm to minimize the residual between experimental and model data. Global optimization schemes have the ability to find global minima/maxima of a broad parameter space but require a very large number of function evaluations. This research focuses on decreasing the computational cost of the optimization scheme by developing a simplified model of the structure of interest and defining the objective function with amplitude dependent frequencies and damping ratios. A recently developed quasi-static modal analysis technique is used to determine these amplitude dependent properties of the model at a significantly reduced cost in comparison to solutions obtained with numerical time integration. This technique is demonstrated on a structure termed the Ministack which contains a foam-to-metal interface held together with a press fit joint.