Vibration analysis of machines operating under non stationary load conditions requires special attention. This refers to the necessity of use of order tracking algorithms together with additional advanced signal processing methods, especially when faults indicators of interest are located in resonance frequency ranges. Nowadays cyclostationarity analysis becomes more popular in the field of mechanical systems signal processing. It has proven its functionality in diagnostics of rotating machinery in numerous cases. One of the most important tools of cyclostationarity is the Instantaneous Power Spectrum (IPS). The authors investigated a selected cases of wind turbine drive train faults as typical examples of machinery operating under non stationary load conditions. Instantaneous power spectrum was applied to vibration signals components originated from rolling element bearings. The capabilities and limitations of IPS calculated in angular domain have been presented with the main focus on variable operational conditions.