Condition Prediction Model and Component Interaction Fault Tree for Heat Distribution Systems.

摘要

Predictive maintenance of heat distribution systems (HDSs) is highly desirable, but very difficult to cost-effectively execute. Frequent, detailed inspection is largely impractical, and components are subject to complex, obscure interdependencies that can create seemingly unrelated distresses virtually anywhere in the system. An Engineered Management System (EMS) called HEATER is being developed to give Army utility managers a set of repeatable, inspection-based tools to enhance HDS operation and maintenance programs. Inclusion of a condition prediction function in HEATER would be a great asset for planning and cost-optimizing scenario-based preventive maintenance. The objective of the current study was to develop a practical condition prediction method for incorporation into HEATER. The resulting method introduces two new tools: (1) a component interaction fault tree, which accounts for all HDS components and their major interdependencies, and (2) a condition index (Cl) tree, which assigns weights to various components according to their level of impact on overall system condition. Inspection data are to be processed through the fault tree, then mapped into the EMS via the CI tree. The resulting EMS output is expected to support scenario-based condition prediction that is significantly more reliable than could be achieved through manual tracking of the separate system components.