To solve the unsatisfying fault predicting effect of hydraulic pumps based on traditional algorithms, a novel fault fusion predicting method based upon the improved Newman-Watts(INW) is proposed. First, the structure of the reserve pool in traditional echo state network (ESN) is modified, and the INW-ESN fundamental predicting model is established. Neighboring matrix elements are redefined to improve the network performance. Furthermore, the fault fusion prognostic model is proposed. By the employment of Dezert-Smarandache theory (DSmT), the INW-ESN prognostic information and degradation model information are fused to increase predicting accuracy. Finally, the proposed method is verified by the application on the hydraulic pump performance degradation experiment.