A state classifier based diagnosis of abnormal condition in centrifugal pumps are proposed to enhance integrity of nuclear power plants. Data management and diagnostic platform was built based on OSA-CBM (Open Systems Architecture-Condition Based Maintenance) and OSA-EAI (OSA for Enterprise Application Integration), international guide on maintenance of industrial facilities. Signals of various fault conditions of a main feed water pump simulator were measured and manipulated by the proposed CBM platform and diagnostic algorithm using ANN (Artificial Neural Network). The platform can successfully estimate conditions of a system.