Aiming　at　the　nonlinear　and　multi-stage　features　in　batch　production　process,　a　new　method　of　improved　sub-stage　multi-way　kernel　entropy　component　analysis　(ISMKECA)　based　on　improved　multi-stage　expansion　is　proposed.　In　this　method,　three-dimensional　historical　data　is　preprocessed　using　the　proposed　data　expansion　strategy,　which　solves　the　model　error　problem　caused　by　data　filling;　and　then　through　kernel　mapping,　the　data　is　mapped　from　low-dimensional　space　into　high　dimensional　feature　space,　which　solves　the　nonlinear　characteristic　of　the　data.　In　high　dimensional　feature　space,　the　stage　division　of　the　data　is　carried　out　according　to　kernel　entropy　and　angle　structure　information,　and　the　ECA　monitoring　models　are　established　in　the　divided　stages,　which　solves　the　multiple　modal　problem　of　the　data.　Finally,　the　proposed　algorithm　was　applied　to　the　online　monitoring　of　industrial　penicillin　fermentation　simulation　system,　which　verifies　the　effectiveness　of　the　proposed　method.