In this work a new methodology for monitoring reactor systems is introduced and its application on pressurized water reactor (PWR) data is presented. The methodology implements a synergistic framework of various relevance vector machines and fuzzy inference. The goal of the framework is to fuse the sensor readings and provide them in a solid form to the operator: this solid form is identified as the state of the reactor. Initially, each of the relevance vector machines fuses the data taken from a specific set of sensors, which have been assigned to it. In the next step, the fused data in the form of a nominal value are forwarded to a fuzzy inference system that takes the values of all the relevance vector machines and further fuses the data by providing a single value that matches the state of the reactor. This proposed methodology is applied on a set of real-world data taken from the LOFT Facility, which is a setup to simulate a pressurized water reactor. Results demonstrate the efficiency of the method in identifying the correct reactor state, while reducing the volume of processing data.