The Spectral Kurtosis (SK) enhances non-Gaussian behavior associated to deviations from the nominal operation of the cranes machinery. This fact easies fault detection, with the subsequent prevention of dramatic malfunction. In this paper the rotor of a container gantry crane is monitored to get the kurtosis of its normal operation. Then, two types of rolling bearings faults are modeled, according to the design of the rotors crane. These signals are added to the real normal operation recordings, and processed under an estimator of the SK. The experience allows the conformation of a higher-order statistical fault-pattern data base, without the need of stopping huge machinery, and with the subsequent saving, settling the basis of an automatic surveillance system.