Energy-based fault detection and isolation of a Brayton cycle-based HTGR power conversion unit – A comparative study

Abstract

The amount of data associated with condition monitoring of large scale industrial plants has led to a “big data problem”. Energy-based fault detection and isolation (FDI) has been proposed as a hybrid FDI method to reduce data requirements and improve FDI performance. In this paper, a comparative study is performed on three energy-based FDI techniques, namely: an enthalpy–entropy error-based technique, a residual-based technique and an eigendecomposition-based technique. A representative Brayton cycle-based power conversion unit (PCU) is considered as the case study and simulated under normal and fault conditions. From the simulation data, the PCU is represented in terms of energy, either as an enthalpy–entropy diagram or an attributed graph. The graph representation allows structural information to be captured which contributes towards fault isolation capability. For each FDI technique, operational energy signatures are compared to normal and fault reference energy signatures to detect and isolate fault conditions. This paper contributes by comparing the FDI techniques based on the attributes of quick detection and isolation. The residual-based technique had the best performance in detecting and isolating faults. The results also indicate that a combination of techniques will result in an FDI technique with improved detection and isolation performance over the individual techniques.