Sternal deflection is an injury criterion used in current regulatory and consumer tests worldwide to assess thoracic injury risk. However, this criterion has some serious limits when applied to the Hybrid-III dummy:​ the risk curve based on the criterion is restraint dependent, and it does not allow discrimination between some advanced restraint systems. The THOR dummy, despite its better biofidelity, is confronted with similar limits. This paper presents a study aiming at identification of more robust injury criteria. A human body FE model-based approach was used to achieve this objective. First, an existing human model was updated and validated for frontal impact simulation, not only in terms of its gross motion response, but also in terms of its capability to predict rib fractures. It was then submitted to a wide range of loading types:​ impactor, static airbag, belt only restraint, airbag only restraint and combined belt and airbag restraint. For each loading type, different loading severities were applied to generate different levels of rib fracture:​ from the absence of fractures to numerous fractured ribs. Based on these simulations, bending was identified as the main loading pattern for rib fracture, and two injury criteria were formulated:​ the Combined Deflection (Dc) and the Number of Fractured Ribs (NFR). The Dc is a deflection-based criterion which takes into account not only sternal deflection, but also the effect of asymmetrical loading. This effect can be characterized by L-R differential deflection (difference of thoracic deflections measured on the left side and the right side of the thorax). The NFR is a rib strain-based criterion which intrinsically reflects the injury level of ribs. The simulations showed that the maximum peak strain of all ribs does not correlate with the number of fractured ribs. The NFR can be calculated by measuring dummy rib strain and by fixing a strain threshold beyond which a dummy rib is considered fractured. A possible approach to apply the NFR to mechanical dummies was proposed. However, based entirely on numerical simulations, the findings of this study need to be evaluated by physical testing. A preliminary study on THOR rib strain measurement showed positive signs for implementation of the NFR on the THOR dummy.