The coupling is the unique mechanical connection between the high-speed rotating machinery gearbox and the generator in the wind turbine. Looseness and skid faults of coupling directly threaten the stable and reliable operation of wind turbines， and even cause major accidents or damage to other equipments. After coupling loosening， the observed signals are also contaminated by strong background noises and harmonic interferences. The fault characteristics are difficult to identify. To address those problems， adaptive ensemble empirical mode decomposition（EEMD） with the collaborative signal-to-nosisratio （CSNR） index is proposed. The method in this paper is used to analyze the numerical simulation， and components of the simulation signal are separated accurately. This new method is applied to the looseness fault diagnosis on coupling of wind turbines， and the fault characteristics of weak signals under the background of strong noise of looseness are extracted effectively， verifying the effectiveness and practicability of the method in practical engineering applications.