In this paper, a comprehensive method for eccentricity fault diagnosis in switched reluctance machines (SRMs) during offline and standstill modes is presented. This sensorless method is able to detect occurrence, location, direction, and severity of the eccentricity fault in SRM. First, the theory of the method is presented, and then its background formulation is explored and discussed. In order to model the eccentric SRM and apply the test method, a 3-D transient finite-element method (3-D TFEM) is employed, which takes into account all physical and geometrical characteristics of the motor components. Next, a test rig is established to carry out various measurements on a faulty motor with different degrees of eccentricity. Then, a procedure of diagnosis method with achieved data is described. In this regard, this method investigates how the fault location as well as the faulty phase in the SRM can be detected. Furthermore, the strategy of the fault direction detection based on induced voltage variations is stated. In addition, a proper index will be proposed in order to detect the fault severity. To demonstrate the merits of this indicator in the eccentricity detection and estimation of its severity, the correlation between the index and frequency and voltage variations are evaluated.