The constant requirement of aerospace industry to enhance the structural efficiency has driven to the usage of high-performance composite materials, either monolithic or sandwich. However, aerospace composite structures are prone to damage due to high-velocity impact events such as bird strike, hail impact, etc. These impact events can result in extensive damage including structure perforation, which will eventually degrade its post-impact residual strength. Therefore, the early detection of damage in composite structure is imperative to avoid catastrophic failure. This paper develops the computational models which predict the dynamic behavior of a helicopter composite sandwich structure undergoing a bird strike. The models are aimed to be used as virtual tools for a future digital-twin-assisted fault detection technique. Firstly, a high-fidelity (HF) FE/SPH model was developed in LS-DYNA, and it was validated against the soft body impact experiments. Afterwards, a computationally efficient low-fidelity (LF) model was developed and correlated with the high-fidelity model. It was concluded that the high-fidelity model can sufficiently accurately predict the strain history experimentally recorded by the FBG sensors, and that size of the predicted delamination area at the front face of the sandwich structure agrees very well with the experimentally observed delamination area. It was also shown that the LF model can rapidly predict the global dynamic response of sandwich panel under the impact loading, through the good agreement between the numerical strain histories with the FBG measurements. Consequently, the LF model can be used as a quick numerical guide for the identification of the loading condition, whereas the HF model can be used as virtual damage detector and estimator of damage extension before the scheduled inspection.