The multifractal detrended fluctuation analysis（ MFDFA） is developed only for processing one- dimension time series. Consequently,when applied to analyze complex vibration data from a defective gearbox,the MFDFA frequently produces poor results due to disturbance from noise and other factors. As a result,high- dimension data analysis seems effective in solving this problem. To this end,multifractal detrended cross- correlation analysis（ MF- DCCA） is introduced to examine gearbox vibration data and then a novel method for fault diagnosis of gearboxes is proposed based on MF- DCCA. In the proposed method,a cross- correlation function of two- dimension signals is firstly estimated and then the multifractal feature hidden in the cross- correlation function are analyzed. Application to probing realist gearbox vibration data suggested that the proposed method can clearly distinguish between similar and close fault patterns of gearboxes and performs better than the method based on MFDFA in fault diagnosis of gearbox.