Reliability assessment of power converters is extremely important due to the degradation of the converter performance under the thermal and electrical stresses. The space environment is one of the stressful environments for electronic components and equipment due to the range and high speed of temperature changes.The normal or abnormal operation of a converter depends on the quality of the manufacturing process and the environmental and operating conditions. The failure indices usually are obtained based on the previous failures data which are calculated using the history of the main parameter of the converter. These indices are strongly influenced by the aging process. In this article, a new real-time indicator is introduced using the monitoring of the main parameters of the converter. The indicators are calculated using Replicator Neural Network (RNN). In fact, these indicators are obtained based on a comparison between a reference model of the converter in normal conditions and the estimation of abnormal operation of the converter in the future. In the proposed method, a normal distribution function is used to find the limits of error signals. The proposed method has several advantages such as considering all the uncertainties during the process of manufacturing the device, no need for the aging test data, and including all the failure types. In the Electrical power subsystem of a spacecraft, the reliability of power converters can be assessed based on the obtained data from the qualification models, benefiting the proposed method.