Damage recognition is important to asses the sstructural system and provide safety during their service life. Change in dynamic characteristics from the undamaged state indicates significant damage in the structures. The concept of change in natural frequency due to earthquake has been employed to identify the damage in the structure. A 2D reinforced concrete building frame is analyzed for different levels of damage using structural engineering software STAAD.Pro V8i. Further, an artificial neural network using backpropagation algorithm is trained to develop the correlation between the damage in the frame with its known dynamic characteristics. Efficiency of artificial neural network to predict the damage for untrained parameters is studied.