Gearboxes are complex dynamical systems and vibration data from defective gearboxes are typically characterized by nonstationarity and nonlinearity.Consequently

it is difficult to extract fault features from these nonstationary and nonlinear data.Detrended Fluctuation Analysis（DFA）can be employed to analyze nonstationary and nonlinear data and the scaling-law curve obtained by DFA can be used to display the dynamical behavior of complex systems.However

scaling-law curves of original data from fault gearboxes usually assume complex shapes.As a result

it is a difficult problem to extract feature parameters from these complex scaling-law curves.To resolve the problem

the scaling property of increment series from fault gearboxes is studied and that the amplitude component of the increment series mainly conveys the nonlinear information is proved.Next

a novel method for fault diagnosis of gearboxes based on scaling features of amplitude components of time series is proposed.The proposed method consists of four parts.Firstly

the increment series of original series is derived.Next

the amplitude component of the increment series is determined.Thirdly

the DFA algorithm is utilized to acquire the scaling-law curve of the amplitude component.Finally

the fluctuation parameters for the starting and the crossover points of the scaling-law curve are extracted and used as feature parameters to describe the conditions of a gearbox.Subsequently

the effectiveness of the proposed method is checked using gearbox fault data.The results show that the proposed method is robust to noise

delivers a good performance and outperforms the diagnostic methods using the temporal parameters in fault diagnosis of gearbox.