Steam turbine is the main system of a steam power plant and critical for power generation. Therefore, there is urgency for maintaining the reliability and availability of a steam turbine. A fast and accurate fault detection and diagnosis (FDD) system should be developed as an integral part to prevent a system from catastrophic disaster due to unhandled failures. Many previous studies applied model-based methods to build the FDD system. However, using those approaches required prior knowledge of the system. The power plant is a complex system, where comprehensive process knowledge is a real challenge. On the other hand, power plants have implemented condition monitoring which resulted in process monitoring data. Therefore, this study proposed a data-driven FDD system in a steam turbine of thermal power plant. The study used the process monitoring data from an Indonesian government owned steam power plant. A neural network based classifier was constructed to detect and diagnose faults as well as normal operating condition based on three scenarios. The result showed that the last two scenarios, with and without PCA approach, outperformed the first scenario which only used selected process parameters. The study demonstrated the superiority of data driven approach in the fault detection and diagnosis area.