As　the　multi-way　kernel　entropy　partial　least　squares　(MKEPLS)　method　does　not　make　full　use　of　the　higher-order　statistics　of　the　process　data,　which　will　lose　the　important　information　in　the　feature　extraction,　and　result　in　degraded　fault　identification　performance.　To　solve　this　problem,　a　novel　method　based　on　higher　order　statistics　and　multi-way　kernel　entropy　partial　least　squares　(HOS-MKEPLS)　is　proposed,　in　which　the　raw　data　space　is　projected　into　statistics　space　by　calculating　the　higher　order　statistics　of　the　data　set,　establishing　the　monitoring　MKEPLS　model,　then　adopting　the　contribution　figure　method　on　the　trace　of　the　fault　variables.　Finallay,　the　method　is　applied　to　an　industrial　penicillin　fermentation　process,　and　compared　with　the　MKEPLS　model.　Results　show　that　the　method　has　a　better　monitoring　performance　and　can　detect　and　identify　the　fault.　©,　2015,　Beijing　University　of　Technology.　All　right　reserved.