A fault diagnosis model based on multi-kernel extreme learning machine (MKELM) is proposed for the observation-aiming system of armored vehicles. The parameters of the model are optimized by the level-based learning swarm optimizer (LLSO) algorithm, and the simulation experiments are carried out by using the collected historical data. The results show that MKELM has better diagnostic accuracy, and LLSO can solve the problem of slow training speed caused by relatively more parameters of MKELM. Compared with the classical particle swarm optimization (PSO), LLSO-MKELM has a faster optimization speed, which proves that LLSO-MKELM can be used for fault diagnosis of the observation-aiming system, and has a good training speed and accuracy.