In this paper, a novel approach is presented to detect dynamic eccentricity in round rotor synchronous motors. For this, an efficient index is introduced based on processing developed torque using time series data mining (TSDM) method. This index can be utilized to diagnose eccentricity fault and its degree. The capability of this index to predict dynamic eccentricity is illustrated by investigation of load variation impacts on the nominated index. Stator current spectrum of the faulty synchronous motor under different loads and dynamic eccentricity degrees are computed. Effects of the dynamic eccentricity and load variation simultaneously are scrutinized on the magnitude of 17th and 19th harmonic components as traditional indices for eccentricity fault diagnosis in synchronous motors. Necessity signals and parameters for processing and feature extraction are evaluated by winding function method which is employed to model healthy and faulty synchronous motors. (C) 2010 Elsevier Ltd. All rights reserved.