This paper presents a new method which is derived from the classical viscous plus Coulomb friction model to identify mechanical parameters (moment of inertia(J), viscous friction(B) and coulomb friction (C)) of switched reluctance motor (SRM) in real time. It is important to know mechanical parameters in model based control and higher precision simulation of SRM. Total torque and rotor speed must be precisely obtained since mechanical parameters dependent on total torque and rotor speed. In addition, mechanical parameters have very small value according to the motor rated torque. Therefore, total torque is obtained torque estimator based on artificial neural network (ANN). The motor is driven at three different speed cases. Rotor speed and phase current have been measured while the motor drive system has been run at three different speed cases. The measured values are used in the developed ANN based on torque and the new method to find mechanical parameters. Experimental results show that the new method gives precise and accurate mechanical parameters for both under low and high level reference speed of SRM.