Modeling highly complex entity interactions with high precision is challenging in distributed training simulations. When using existing protocols and standards, most weapons effects are resolved by a probability of a kill or a roll of the dice, reducing accuracy and fair fight. This paper presents the results of a study by the Air Force Research Lab (AFRL) to determine if running physics algorithms using the NVIDIA PhysX SDK with a constructive simulation could more accurately model the damage associated with an aircraft/missile engagement than can be done with the currently used statistics based approach. A statistical approach assumes damage based only on missile proximity at the time of detonation along with a pre-determined probability of kill. This approach typically does not factor the exact aspect, geometry, aircraft material strength, and missile warhead capability at the time of detonation. The study leveraged a Physics-Based Environment Generator (PBEG) running with the Expert Common Interactive Training Environment (XCITE) constructive simulation. The PBEG was programmed to take entity control of the aircraft and missiles from XCITE to determine the level of damage, provide a realistic visualization of the impact and return damage assessment to XCITE. A target aircraft model was constructed rendering multiple sections and facets, each capable of individually sustaining damage. A missile warhead model with a simplified TNT explosive and fragmentation was integrated. The PBEG computed the force of the initial blast shockwave and impact from each high velocity shrapnel piece striking the aircraft on each specific facet. A numerical damage value is assigned for each aircraft part, and damage results are returned to XCITE for training feedback. This approach was shown to increase accuracy in damage assessments over the traditional approach. This paper summarizes the study results and provides recommendations for further investigations using graphics processors for modeling weapons effects.