An alarm filtering system for the pressurizer of the nuclear reactor(AFSP) is developed based on the rule based deduction method and the independent decision support strategy. This system consists of six parts; knowledge base, inference engine, failure diagnosis system for the pressurizer (FDSP), data transfer system (DTS), data modification system(DMS), and user interface part. The knowledge base consists of rules and facts. The inference engine uses the inherent property of PROLOG, the depth-first searching. FDSP performs failure diagnosis for the pressurizer of the nuclear reactor. DTS is developed to acquire knowledge and determine initial certainty factors. DMS updates database after failure diagnosis is performed. Finally the user interface part utilizes the menu-driven method. The developed system is verified for several sample cases. The system shows the satisfactory capability for the anticipated failures. AFSP is developed on an IBM PC-AT compatible using PROLOG language.