Detection of lubricant starvation in ball bearings is essential to maximize their lifetime. The lack of lubrication introduces impulses in the vibration signal related to fundamental train and ball spin frequencies. The phenomenon is attributed to the lack of a damping to dissipate the impacts of the balls when entering to the load zone together with the gap increase between bearing components, previously filled by lubricant. This work presents an experimental vibration analysis to study the effect of preload as a mean to decrease the extra clearance in dry conditions and eliminate the source of impulses that characterize lubrication starvation. Furthermore comparison between RMS, Crest Factor and Kurtosis reveals that an increase in preload indeed reduces the signal impulsivity and kurtosis proves to be an effective method to detect lubricant starvation. Finally, the application of fast-kurtogram algorithm is used as a pre-processing method to improve the performance of kurtosis in detecting of lubrication starvation.