winding short circuit faults are recognised as one of most frequent electric machine failure modes. Effective on-line diagnosis of these is vital but remains a challenging task, in particular at incipient fault stage. This research reports a novel technique for on-line detection of incipient stator short circuit faults in random wound electrical machines based on in-situ monitoring of windings thermal signature using electrically non-conductive and electromagnetic interference immune fibre-Bragg grating (FBG) temperature sensors. The presented method employs distributed thermal monitoring, based on the FBG multiplexing feature, in a variety of points within windings, in proximity to thermal hot spots of interest that arise from fault. The ability of the proposed method to enable fault diagnosis through identification of fault induced localised thermal excitation is validated in steady-state and transient operating conditions on a purpose built inverter driven induction machine test facility. The results demonstrate the capability of unambiguous detection of inter-turn faults, including a single shorted turn. Furthermore, the winding thermal and electrical characteristics at the onset of inter-turn fault are examined and correlated enabling better understanding of fault diagnostics requirements.