In order to solve the real-time fault diagnosis problem of complex system, a multi-objective optimization strategy of test set based on simulated annealing algorithm (SAA) is designed. Starting with optimal multi-criteria such as fault isolation rate, number of tests and cost, each objective is sorted in descending order consistent with its importance and realized in accordance with its priority. The example of Apollo system state detection before launch is introduced to validate its ability. The simulation results show that the SAA principle is simply and better precision, which can give effective direction to the test selection of real-time fault diagnosis system.