Data-based modeling of vehicle crash using adaptive neural-fuzzy inference system

Vehicle crashes are considered to be events that are extremely complex to be analyzed from the mathematical point of view. In order to establish a mathematical model of a vehicle crash, one needs to consider various areas of research. For this reason, to simplify the analysis and improve the modeling process, in this paper, a novel adaptive neurofuzzy inference system (ANFIS-based) approach to reconstruct kinematics of colliding vehicles is presented. A typical five-layered ANFIS structure is trained to reproduce kinematics (acceleration, velocity, and displacement) of a vehicle involved in an oblique barrier collision. Subsequently, the same ANFIS structure is applied to simulate different types of collisions than the one which was used in the training stage. Finally, the simulation outcomes are compared with the results obtained by applying different modeling techniques. The reliability of the proposed method is evaluated thanks to this comparative analysis. © 2013 IEEE.