Due to the frequent occurrence of spacecraft failures and accidents, it will cause personal injury and economic loss, and will have a huge impact on the aerospace industry. Therefore, it is necessary to enrich the research on satellite constellation fault diagnosis and detection technology. In order to solve the shortcomings of existing research on fault diagnosis and detection of satellite swarms, this paper discusses the neural network fault diagnosis technology, satellite swarm fault diagnosis method and the functional equation of satellite swarm execution fault types. The fault samples and parameter settings of the group fault diagnosis detection application are briefly introduced. In addition, the work design and calculation process of satellite swarm fault diagnosis and detection relying on neural network are discussed. Finally, the application of neural network in satellite swarm fault diagnosis and detection is subjected to experimental comparison and analysis of the correct rate of fault diagnosis. The experimental data show that the algorithm proposed in this paper and Although the correct rate of fault diagnosis of the other two types shows a downward trend after the failure rate of 10%, the correct rate of the neural network algorithm proposed in this paper is significantly better than the other two algorithms, and the algorithm in this paper is in 50% to 200% of the satellites. In the group failure rate, the correct rate of diagnosis is stable at about 96%, while the correct rate of the other two algorithms is gradually lower than 90%. Therefore, it is verified that the satellite group fault diagnosis and detection relying on neural network has high use value.