Abstract This paper develops a technique based on vibration magnitudes in vibration signals spectrums for the rotor bar breakages detection in squirrel-cage induction motors. The motor model used in the experiments is established and the skewed rotor is fully taken into account due to its effect on magnetic field. The model is used for the computation of time-domain performance characteristics, such as the magnetic field, air-gap flux density and electromagnetic force including the abnormalities. The vibration of the healthy and faulty motors is calculated based on the electromagnetic forces obtained from the transient magnetic field. The features related to rotor fault detection are estimated by Fast Fourier Transform (FFT) of the electromagnetic force and vibration displacement. According to the diagnostic method based on only a vibration sensor, these features can be easily and accurately found in vibration spectrum. In addition, the modal studies are performed by theoretical and experimental methods respectively, and its purpose is to verify that the broken bars failure will not lead to motor resonance. Experimental results, numerical calculation and analytic analysis suggest that it can provide a powerful and a general approach to the broken bars fault detection.