Fault detection in rotating machines is generally based on vibration signal analysis. However, the sound radiated by a structure and its vibration are closely linked. We can therefore imagine that acoustic measurements could be useful for diagnostic improvement. In this paper, a fault diagnostic method based on acoustic imaging is proposed. Beamforming is used to describe the acoustic field generated by an operating machine. Usually, the source strength is mapped in order to identify the radiating areas. In this paper, fault detection features are plotted instead. The spectral kurtosis is mapped as a function of space and frequency to separate and localise impulsive sources.