The gearbox of Volvo Construction Equipment vehicles may be considered as a critical mechanical component. Gear failures may cause costly down time while the gearbox is repaired. To improve uptime and reduce service costs, reliable and robust real time condition monitoring and diagnostics of gearboxes based on vibration data may be considered. This concerns the utilization of adequate feature extraction methods and classification methods. In this thesis project feature extraction methods have been considered for the purpose of enabling the detection of the slightest abnormal gear vibration in a gear test rig and a transmission test rig at Volvo CE. In particular a Time-Frequency Domain Averaging (TDFA) algorithm has been implemented for the extraction of gear box vibration features enabling the detection of abnormal gear vibration. Vibration has been measured for both test rigs during a test period of 2500 hours and a great number of vibration records were produced. Fluctuation in vibration could be observed as the experiment progressed using the TDFA method. The TDFA provide information on periodic components in a vibration signal related to the rotation speed it is synchronized and variations of them over time.