MODELING OF STORE SEPARATION BEHAVIOR BASED ON A NEURAL NETWORK AND UNSTEADY FLOW SOLUTIONS

Erinc Erdogan and Ismail H. Tuncer

In this study a neural network based method is developed for the prediction of separation characteristics of external store weapons carried under aircraft wings. The method is based on an artificial neural network trained by high fidelity unsteady flow solutions. The unsteady flow solutions as the store separates from the carriage and the resulting six degrees of freedom motion of the store are computed conditions by a commercial flow solver for various flight conditions. The trajectory of the store and the unsteady aerodynamic loads acting on it are used to train a neural network. A simulation is used for evaluating the 6 degrees of freedom motion of a store based on the aerodynamic loads predicted by the neural network as a function of the flight conditions and the instantaneous position of the store. The efficiency and the accuracy of the method developed are determined by comparing the predicted and computed trajectories of the store for flight conditions not used in the training set. It is shown that the method developed successfully can predict the store separation characteristics.

Fulltext (PDF)