Motor current signature analysis (MCSA) has been accepted in the field as a reliable means of detecting rotor cage faults in induction motors. However, there are many limitations to MCSA-based detection for other types of faults including rotor eccentricity and load defects. Recently, there has been increasing interest in airgap or leakage flux monitoring as a low cost alternative to replace or complement MCSA. Flux monitoring can provide a reliable means of detecting rotor faults since anomalies in the rotor magneto-motive force or airgap can be directly measured. In this paper, a new method based on monitoring the attenuation of rotor rotational frequency sidebands in the flux spectrum is proposed for reliable detection and classification of rotor cage and eccentricity faults. It is also shown that flux monitoring can provide reliable detection of rotor faults independent of load defects. An experimental study under controlled broken bar, eccentricity, misalignment, and load unbalance conditions are given to support the claims. The test results show that rotor cage, eccentricity, and load defects can be detected and distinguished for cases where MCSA alone is ineffective. A comparative evaluation between conventional and proposed flux monitoring methods is also given.