Abstract: The launching decision-making system is an expert system which can diagnose the faults of the launch vehicle during the pre-launch testing and give the auxiliary repair strategy and launch decision. In order to achieve the quick fault diagnosis, the launching decision-making expert system should use a fast inference algorithm because there are massive test data need to be processed and the time for diagnosis is quite short. Therefore, a fast inference algorithm including the fact-matching and rule-matching was presented. The fact-matching network was borrowed ideas from the rete algorithm. And the rule-matching network was performed based on the vectors calculation. Finally, a fast inference algorithm suitable for pre-launch FDD (fault detection and diagnosis) of launch vehicle was given. The presented algorithm has been applied in a distributed FDD expert system for a real of launch vehicle. Preliminary results indicate that the algorithm reasons more quickly and more accurately with less system resource consumption, and can meet the real-time requirements.