Monitoring of wind turbines under variable-speed operating conditions has become an important issue in recent years. The gearbox of a wind turbine is the most important transmission unit; it generally exhibits complex vibration signatures due to random variations in operating conditions. Spectral analysis is one of the main approaches in vibration signal processing. However, spectral analysis is based on a stationary assumption and thus inapplicable to the fault diagnosis of wind turbines under variable-speed operating conditions. This constraint limits the application of spectral analysis to wind turbine diagnosis in industrial applications. Although order-tracking methods have been proposed for wind turbine fault detection in recent years, current methods are only applicable to cases in which the instantaneous shaft phase is available. For wind turbines with limited structural spaces, collecting phase signals with tachometers or encoders is difficult. In this study, a tacholess order-tracking method for wind turbines is proposed to overcome the limitations of traditional techniques. The proposed method extracts the instantaneous phase from the vibration signal, resamples the signal at equiangular increments, and calculates the order spectrum for wind turbine fault identification. The effectiveness of the proposed method is experimentally validated with the vibration signals of wind turbines.