Abstract: Aiming at the difficulty of fault feature extraction and fault mode identification in gearbox fault diagnosis, a gearbox fault diagnosis method combining variational mode decomposition (VMD), African vulture optimization algorithm (AVOA) and stochastic configuration network (SCN) is proposed. Firstly, aiming at the problem that the random initialization of SCN network weight and bias will lead to the instability of network prediction results, the AVOA algorithm is proposed to optimize the initialization selection method of SCN network node weight and bias for fault classification and recognition. Secondly, the VMD algorithm is used to decompose the gearbox vibration signal into several intrinsic mode components (IMF), and then the correlation coefficient is used to screen the IMF component and calculate its sample entropy as the feature vector, which is input into the SCN network optimized by AVOA algorithm for classification and recognition. The experimental results show that the proposed method can accurately identify the fault mode of the gearbox, and the recognition accuracy reaches 98.33%. Compared with BP, ELM, RVFL, SVM, SCN and other methods, the proposed method has higher fault recognition accuracy.