This paper presents a model of the dc motor with an eccentric rotor. The winding function theory shows the effect of eccentricity fault on the motor inductances and the simulation is done using a nonsymmetric air-gap function. A modified equation is presented to show the existence of rotor slot harmonies in the dc motor current. To detect the eccentricity fault, a pattern recognition technique is utilized. The proposed algorithm works at steady state and uses armature current as input. The rotor speed is needed in order to provide the appropriate feature for the classifier. Therefore, rotor speed is estimated from the armature current using the commutation harmonics. The experimental results obtained from a 1/3-hp shunt dc motor verifies the proposed method. In order to cover different motor conditions, data are collected at different shaft speeds for both a healthy dc motor and a dc motor with an unbalanced load which exhibits static eccentricity.