Combining multi-field information fusion theory with the ammunition feeding system, a multi-level information fusion method for fault detection and prediction is proposed in this paper. Firstly, it is studied that the occurrence and development mechanism of damage and failure of complicated ammunition feed system under rapid impact. The wear damage and stop shooting faults are diagnosed by measuring the power loss characteristics at different scales. Using modern signal processing and intelligent algorithm, the weak damage and wear stuck failure characteristics are extracted, and the fault positions are identified. On this basis, and combining the theories of impact vibration and wear damage, the degree of wear stuck failure, the law of damage development and expansion, and the remaining life of the system after the emergence of multi-damage fault, can be predicted. Finally, multi-field information fusion theories and methods for fault monitoring are established, and a monitoring platform with embedded system is developed.