The broader aim of this research is to develop fundamental understanding of ductile fracture process in structural steels, propose robust computational models to quantify the associated damage, and provide numerical tools to simplify the implementation of these computational models into general finite element framework. Mechanical testing on different geometries of test specimens made of ASTM A992 steels is conducted to experimentally characterize the ductile fracture at different stress states under monotonic and ultra-low cycle fatigue (ULCF) loading. Scanning electron microscopy studies of the fractured surfaces is conducted to decipher the underlying microscopic damage mechanisms that cause fracture in ASTM A992 steels. Detailed micromechanical analyses for monotonic and cyclic loading are conducted to understand the influence of stress triaxiality and Lode parameter on the void growth phase of ductile fracture. Based on monotonic analyses, an uncoupled micromechanical void growth model is proposed to predict ductile fracture. This model is then incorporated in to finite element program as a weakly coupled model to simulate the loss of load carrying capacity in the post microvoid coalescence regime for high triaxialities. Based on the cyclic analyses, an uncoupled micromechanics based cyclic void growth model is developed to predict the ULCF life of ASTM A992 steels subjected to high stress triaxialities. Furthermore, a computational fracture locus for ASTM A992 steels is developed and incorporated in to finite element program as an uncoupled ductile fracture model. This model can be used to predict the ductile fracture initiation under monotonic loading in a wide range of triaxiality and Lode parameters. Finally, a coupled microvoid elongation and dilation based continuum damage model is proposed, implemented, calibrated and validated. This model is capable of simulating the local softening caused by the various phases of ductile fracture process under monotonic loading for a wide range of stress states.  Novel differentiation procedures based on complex analyses along with existing finite difference methods and automatic differentiation are extended using perturbation techniques to evaluate tensor derivatives. These tensor differentiation techniques are then used to automate nonlinear constitutive models into implicit finite element framework. Finally, the efficiency of these automation procedures is demonstrated using benchmark problems.