Fault location of tribo-mechanical systems - a graph theory and matrix approach

Abstract

A procedure for the fault location of tribo-mechanical systems based on graph theory and matrix approach is presented. A component digraph which represents the failure propagation network of the tribo-mechanical system is developed for identification of fault location sources and failure paths. Starting from the nodes that indicate abnormal states, some nodes from the component digraph are selected as the probable candidates of failure origin by back-tracing. These candidates are further screened by using failure propagation probabilities and failure propagation time between adjacent nodes, and back-tracing. Finally, the failure propagation probabilities and failure rates of the devices are used to evaluate the ranking order among the screened candidates for fault location. The faults are located through pre-processing and alarm processing of the digraph. The methodology is applied to a journal bearing oil supply system for a thermal power plant.