Monitoring the health of large electrical machines, especially power station generators, is now an integral part of their operation to maintain and extend life. This work studies the use of electromagnetic sensors to detect inter-lamination insulation faults in the stator cores of large generators before they propagate to a level that can lead to catastrophic failure. The work develops a deeper understanding of the electromagnetic behaviour of core faults so that condition-monitoring sensors can be more specific about the location and severity of the fault. The study develops two new three-dimensional (3D) analytical models, one for predicting the fault current distribution in a stator core fault and the second for predicting the varying detection of such current by air-cored sensors. This further analysed the 3D detection efficiency of typical short fault lengths to compare with the two dimensional (2D) default of infinite-length faults. Different fault positions were modelled so that a clearer understanding of the location and severity of the fault is possible. These were validated on a specially constructed experimental test core that can impose controlled fault currents. The study also demonstrates how small core faults can escalate then self-limit radially, but may propagate axially into longer more damaging faults.

Interlamination insulation faults in the stator cores of large electrical machines can damage both winding insulation and stator core, thus confidence in electromagnetic test results is important. They may be validated by FE methods, however the 3D models required for short faults are computationally challenged by laminated structures, requiring approximations. A homogenised 3D FE model was used to model faults buried in the teeth and yoke of the core, with a new experimental methodology developed to calibrate fault currents. Limitations were identified in modelling just a core section due to images and the constraint of axial packet air gaps on fault flux dispersion. A system of transverse 2D FE models of the principal fault flux paths in the core were constructed to estimate the differential impact on fault signals by the air gap presence and applied to the 3D FE model. Together with corrections for images this gave close predictions of experimental results, supporting the validity of the model. The verified electromagnetic test results now permit assessment of the threat that a detected buried fault presents.

Faults in the stator cores of large electrical machines can both damage local winding insulation and propagate to catastrophic failure. This study develops three-dimensional finite element models of inter-laminar insulation faults in order to obtain a deeper understanding of the electromagnetic behaviour of core faults and the sensitivity of sensing systems. The problem of developing a model that adequately reflects the laminar constraints of the structure, while remaining computable is addressed, together with eliminating images from boundaries. The model was validated by experimental measurement and results shown to be closely matched, with the fault current distribution also predicted. The sensitivity profiles for various fault positions and lengths were determined, which enables condition-monitoring sensors to be more specific about the location and true threat that a fault signal may pose to the machine.