Indoor fire simulators have been used to analyse building safety in the events of emergency evacuation. These applications are primarily focused on simulating evacuation behaviors for the purpose of checking building structural problems in normal time rather than in real time situations. Therefore, they have limitations in handling real-time evacuation events with the following reasons. First, the existing models mostly experiment the artificial situations using randomly generated evacuees while real world requires actual data. Second, they take too long time in operation to generate real time data. Third, they do not produce optimal results to be used in rescueing or evacuation guidance. In order to solve these limitations, we suggest a method to build an evacuation simulation system that can be used in real-world emergency situations. The system performs numerous simulations in advance according to varying distributions of occupants. Then the resulting data are stored in DBMS. The actual person data captured in infrared sensor network are compared with the simulation data in DBMS and the querried data most closely is provided to the user. The developed system is tested using a campus building and the suggested processes are illustrated.