The development of various topologies and configurations of axial-flux permanent magnet machine has spurred its use for electromechanical energy conversion in several applications. As it becomes increasingly deployed, effective condition monitoring built on reliable and accurate fault detection techniques is needed to ensure its engineering integrity. Unlike induction machine which has been rigorously investigated for faults, axial-flux permanent magnet machine has not. Thus in this thesis, axial-flux permanent magnet machine is investigated under faulty conditions. Common faults associated with it namely; static eccentricity and interturn short circuit are modelled, and detection techniques are established. The modelling forms a basis for; developing a platform for precise fault replication on a developed experimental test-rig, predicting and analysing fault signatures using both finite element analysis and experimental analysis. In the detection, the motor current signature analysis, vibration analysis and electrical impedance spectroscopy are applied. Attention is paid to fault-feature extraction and fault discrimination. Using both frequency and time-frequency techniques, features are tracked in the line current under steady-state and transient conditions respectively. Results obtained provide rich information on the pattern of fault harmonics. Parametric spectral estimation is also explored as an alternative to the Fourier transform in the steady-state analysis of faulty conditions. It is found to be as effective as the Fourier transform and more amenable to short signal-measurement duration. Vibration analysis is applied in the detection of eccentricities; its efficacy in fault detection is hinged on proper determination of vibratory frequencies and quantification of corresponding tones. This is achieved using analytical formulations and signal processing techniques. Furthermore, the developed fault model is used to assess the influence of cogging torque minimization techniques and rotor topologies in axial-flux permanent magnet machine on current signal in the presence of static eccentricity. The double-sided topology is found to be tolerant to the presence of static eccentricity unlike the single-sided topology due to the opposing effect of the resulting asymmetrical properties of the airgap. The cogging torque minimization techniques do not impair on the established fault detection technique in the single-sided topology. By applying electrical broadband impedance spectroscopy, interturn faults are diagnosed; a high frequency winding model is developed to analyse the impedance-frequency response obtained.