Increasing the reliability and the downtime of wind turbines is critical to minimise the cost of energy (COE) in the wind sector, especially for offshore wind turbines. Due to the high impact that gearboxes and generator downtimes create on wind turbines, reliable and cost-effective condition monitoring systems (CMS) for the drive train are a great concern to the wind industry. This manuscript presents an approach for condition health monitoring and fault diagnosis in wind turbine gearboxes and generators by means of analysing the power dependant vibrations gathered. This methodology is based on the establishment of the normal operation boundaries for carrying out the identification of deviations related to a defect. The validity of the baseline is studied using q-factor and probability of detection (POD) concepts. Given the nonlinear and nonstationary nature of the faulty vibration signals, envelope analysis is proposed as a demodulation technique to be applied to the signals, prior to the frequency response being extracted. The methodology is validated by field trials in a WINDMASTER300 wind turbine. Baselines for the generator and gearbox were produced as a tool to detect future faults developed within the turbine. Envelope analysis makes the identification of the vibrational frequencies representative of failure very likely. (C) 2017 Published by Elsevier Ltd.