Online task of diagnosing disturbances and planning remedial actions are very important maintenance works in Nuclear Power Plant (NPP). A general disadvantage of conventional computer support system is that diagnosis results are not acceptable by human operators since the diagnostic knowledge is hidden in the support system. In this study, an application of semiotic analysis method applying Multilevel Flow Models (MFM) is introduced into real-time diagnostic system of PWR plant. The study shows that MFM based alarm analysis method has a good time characteristic and can provide more comprehensible diagnosis result.