In this thesis a systematic approach to fault detection problem in electrical rotating machines based on magnetic field analysis in machine is presented. The aim of this thesis is a contribution in the field of faults and failures identification in rotating machines, that is, improvement of monitoring systems aimed for rotating machine condition monitoring. This thesis deals with detection of inter-coil short circuits in coils of the synchronous machines armature and field winding. Applying FEM (finite element method) to the generator 2D and 3D models, magnetic field wave forms in the machine were calculated, on the points of interests, under both, faultless and faulty conditions. Using the models, flux density and magnetic flux were calculated on the points accessible for measuring with Hall’s sensors and measuring coils in the actual machine. When results of calculations were compared, under faultless and faulty conditions, changes in wave forms caused by the analyzed faults were identified and two new methods of fault detections have been made. New methods are based on analysis of higher harmonics in frequency spectrum of flux density and measuring coils induced voltages wave forms. To confirm calculation results extensive measurements have been made on real synchronous machine, specially adapted in a way that simulated failures can be analyzed. Techniques and technical solutions usually present in monitoring systems for rotating machines where applied to perform measurements. Measuring data are presented. Also comparison between calculated and measuring data is made, in order to confirm theses of this work. Comparison of the data acquired by measurements and calculation for faultless conditions are insignificant, while for faulty conditions deviation is present but in amounts that do not challenge credibility of two new fault detection methods. On the bases of calculation and measurement results fault classificators are made. Within this work system, based on Bayesian networks, for fault detections in rotating machines is made.