The micro-mechanics based approach to the study of ductile fracture has successfully overcome many of the limitations (such as large scale material yielding, cyclic loading, and size/scale dependence of J) of traditional fracture mechanics approaches (i.e. K, J and CTOD's). A number of the currently available micro-mechanics models (i.e. SMCS, Hancock and McKenzie, 1975; VGM, Kanvinde and Deierlein, 2006) predict fracture accurately under high triaxiality and axisymmetric conditions; however, the mentioned conditions do not encompass the full range of stress states (including low-triaxiality or non-axisymmetric conditions) which are relevant to the structural, mechanical and aerospace industries. As such, the primary objective of the work presented in this dissertation is to inform the development of a more general damage model which is applicable to a broader range of stress states and seismic (i.e. cyclic) loading which can result in ultra-low cycle fatigue (ULCF) failures.;New model development is realized through a collaborative multi-scale approach which combines the results of an extensive test series (Smith, 2014) and a series of computational void simulations. To probe the full range of practical stress/loading conditions, a more general finite element (FE) framework for simulating the response of micro-voids is developed. The new void cell framework and the results of the 146 void simulations comprise the primary body of work presented in this dissertation.;The void simulations can be divided into two groups: (1) those which effectively simulate an array of voids while modeling a representative void cell, and (2) those which explicitly model an array of voids. Void growth rates measured from the single void model (SVM) are used to inform the selection of a new functional form for the damage model presented in this dissertation while the multi-void model (MVM) provides qualitative and quantitative insights regarding localized deformation between neighboring voids. Findings from the MVM simulations are (1) in agreement with observations obtained from sectioned images (Smith, 2014) of fracture coupons that expose undergrown voids in the near vicinity of the failure surface and (2) are used to develop a strain-based indicator for localization initiation that shows strong agreement with failure strains observed from coupon scale tests (Myers, 2009). Moreover, the trends observed from both model types indicate that there is minimal void growth and that localization does not occur at low triaxialities. Both finding suggest that an alternate fracture mechanism than the traditionally excepted 'growth to coalescence' mechanism is active under these conditions.;Despite the power of micro-mechanics based models, the ability to arrive at accurate fracture predictions is contingent on the calibration of the parameters which define the material constitutive response. The capability for complementary FE simulations to reproduce the force-displacement response obtained from physical tests (which is typically relied upon for model calibration) provides a false sense of security and neglects issues (i.e. non-uniqueness of the model parameter set) associated with model over-fitting. To investigate the susceptibility of typical calibration approaches to result in non-unique fits, a simple example is employed. Results of the example demonstrate that (1) multiple (and therefore non-unique) parameter sets may adequately reproduce the force-displacement response of typical calibration specimen and (2) that local plastic strains (often used to evaluate local fracture criteria) can result in error more than 65% despite agreement with the calibration metric. Thus, selection of parameter sets based solely on qualitative agreement between test data and complementary simulations can lead to erroneous results when evaluating material resistance to fracture.
展开▼
