For the complex components in the signals collected by the bomb supply and transport system in the early stage,and difficulty in extracting and identifying fault characteristics,a kind of early fault prediction method of the bomb supply and transport system based on fuzzy entropy and cuckoo improved extreme learning machine (CS-ELM) was proposed.The improved adjustable quality factor wavelet transform was used to filter the signal and the fuzzy entropy characteristics of each sub-band signal was extracted;the five sub-bands with larger fuzzy entropy value were selected to reconstruct to complete the noise reduction and their fuzzy entropy was incorporated into feature vectors;CS-ELM was used to predict the early failure of the extracted feature vector and the result was compared with the diagnosis result of ELM.The experimental results show that the proposed method is effective,and the prediction accuracy is 90.7%.