Predicting the collision history of basaltic asteroids from parametrized shapes with an artificial neural network

Asteroids, remnants of our solar system, offer significant insights into planetary formation and evolutionary history through the investigation of their mutual collisions. Previous studies have focused on understanding how the impact conditions and material properties determine the collisional results, e.g., remnant shapes, utilizing laboratory experiments and simulation techniques. However, can we make use of the post-collision information to deduce the impact scenarios, such as the impactor size and velocity? This question remains poorly explored. In this work, we employ a hybrid SPH-DEM method, combining smoothed particle hydrodynamics and discrete element method, to model a wide range of impact scenarios focusing on basaltic spherical asteroids with diameter of tens of kilometer. The simulation results are then used to create a dataset for training an artificial neural network (ANN), which is specifically designed to predict the impact conditions from spherical-harmonic representations of asteroid shapes. The developed ANN model exhibits a good performance and robustness, which further suggests a probable sub-catastrophic origin for Eros-shaped asteroids, as validated by independent numerical simulations. We also find that the impact-induced asteroid shapes are highly sensitive to the material friction, while other material properties can be negligible. Our results suggest a promising future for using ANN to better understand the asteroid collision process and the diversity of impact outcomes.