This paper describes the development of a microstructure-based finite element model to simulate fracture cutting of bovine cortical bone. A cohesive zone-based (CZ) approach is used to simulate fracture cutting of both the haversian, and plexiform components. This model is an improvement over existing bone cutting models in this category that have relied only on parametric inputs for the material properties. A key novelty of this work lies in the fact that the cohesive zone parameters associated with each of the failure modes are heuristically extracted using their specific acoustic emission signatures. This approach ensures that the CZ parameters capture the specific failures observed in the cutting experiment. The cohesive tractions were first extracted using the experimental results for the +20° rake tool. These tractions are then used to predict the results for the 0° rake tool. The tool design utility of the validated model is demonstrated using parametric studies. While the heuristic approach has its limitations given the property variations seen across different bone types and age, the maturation of such experimental data sets combined with machine learning techniques, can yield an effective mapping between the heuristically obtained CZ parameters and the local microstructural property variations in bone. This avenue presents a key opportunity to model the cutting of natural composites whose individual phases cannot be separated for characterization.