Bird strikes represent a major threat to aircraft structures, as a collision during flight has the potential to cause serious structural damage. In order to mitigate this risk, structures have to fulfill the airworthiness specifications prescribed by transportation agencies. The expensive and time-consuming physical certification tests are complemented by numerical simulations that can streamline the number of possible designs for new components. However, bird strike simulations are far from being a trivial task; in a typical impact, the bird behaves as a soft body with fluid-like properties, while the finite element simulation has to deal with complex numerical problems including contact, damage initiation and accumulation, high deformations of the spreading material, failure and removal of elements. The complexity is reflected in simulation time, which, depending on the modeling technique and the number of elements, can be up to hours or even days. This can represent a significant constraint during the design optimization phase, thus limiting the effectiveness of integrated, and automated, simulation processes. A common approach used to speed up design at the cost of some accuracy is the exploitation of surrogate models, numerical models trained on existing data, and used to predict the results for new experiments. Unfortunately, most modeling techniques are ill-suited to deal with the dimensionality and time-dependency of bird strike simulations. One approach that is applicable to this scenario is data-driven, Reduced Order Modeling. This relies on a database reporting the time history of the full-order solutions of the variables of interest, like stress or deformation of the finite elements in the simulation, functions of known values of the design variables, such as the mass of the bird. The database is projected into reduced space by means of singular-value decompositions, and the maps between the original time/space variables and their projections are then used to train a neural network. This offline training creates a compact model that can be applied in reverse to determine the full-order solution generated by new values of the design variables. In this paper, reduced order modeling has been adopted to reproduce the simulation results of a simplified bird strike simulation. The deformation and degradation function of thickness, density and speed, of the elements of an octagonal plate have been used to train a model able to make predictions on the number of damaged nodes in validation scenarios and determine whether the plate sustained the strike, got pierced at the central region, or suffered a failure close to the supported sides. The full paper will include descriptions of the bird strike simulation and reduced order modeling approach, critical analysis and comparison of the simulated and predicted results, and future developments aimed at increasing the complexity of the bird strike scenario to better suit realistic design cases.