Against the weak vibration signal of gear crack fault and the input variation under nonlinear condition, which challenge fault diagnosis'' accuracy, we proposed a whole system method that used 2nd order Volterra kernel function to analyze the inner contact between crack fault and nonlinear change. After identifying kernel function with time series, we analyzed the nonlinear information indicated in frequency spectrum. Analysis results show that gear crack 2nd order kernel function, taking into consideration the influence on diagnosis accuracy caused by input elements, is sensitive to nonlinear changes caused by fault condition, and has solved the fuzziness and the indeterminacy problems caused by traditional gear sideband theory, thus presenting its application foreground in gearbox nonlinear fault diagnosis.