Inverter driven electrical machines need resolver for position estimation and efficient electronic commutations. In this paper, the resolver signals are also used for the purpose of mechanical fault diagnosis. Therefore, a comprehensive procedure based on winding function (WF) method is presented to propose a noninvasive index to identify the mechanical fault's type (static/dynamic eccentricity) and even its percentage. In this regard, once the resolver signals are measured and saved in health condition. Then, during the operation of the motor, calculating the proposed index using the healthy and faulty signals leads to predict the fault type and its percentage. The success of the proposed index is verified using transient finite element analysis and experimental measurements on the prototype resolver.