This　paper　aims　to　deal　with　the　fault　monitoring　problems　for　chillers　considering　the　key　performance　indicator　(KPI).　To　reach　this　objective,　the　coefficient　of　performance　(COP),　which　measures　the　energy　utilization　performance　in　the　chiller,　is　first　selected　as　the　KPI.　Then,　the　fuzzy　clustering　method　is　used　to　establish　the　prediction　model　of　the　COP　of　the　chiller.　On　this　basis,　considering　the　membership　matrix　difference　between　the　normal　and　faulty　data　of　COP　after　clustering,　the　chiller　fault　monitoring　is　realized　via　the　changes　of　membership　functions.　Also,　with　the　COP　prediction　data　of　the　chiller,　the　T2　statistic　is　used　to　design　the　evaluation　function,　and　the　kernel　density　estimation　method　is　employed　for　the　threshold　design.　Thus,　an　additional　COP-related　fault　monitoring　algorithm　for　chillers　is　proposed.　Finally,　an　experimental　study　is　carried　out　on　the　chiller　dataset　for　method　demonstration.　©　2024　IEEE.