Rolling　bearings　are　important　parts　of　rotating　machinery.　When　the　early　failure　occurs,　it　is　difficult　to　effectively　extract　the　weak　fault　features.　Aiming　at　this　problem,　an　early　fault　diagnosis　method　of　variational　mode　decomposition　(VMD)　of　optimizing　the　parameter　K　value　was　proposed.　First,　the　instantaneous　frequency　mean　judgment　method　was　used　to　determine　the　value　of　modal　number　K,　and　then　the　fault　diagnosis　signal　was　processed　by　VMD　method.　By　analyzing　the　intrinsic　modal　function　components　obtained　by　decomposing　the　fault　signal　of　the　bearing,　the　sensitive　components　were　obtained　for　the　envelope　demodulation　analysis　to　judge　the　fault　type　and　severity　of　the　bearing.　Finally,　the　results　obtained　by　the　EMD　and　VMD　algorithm　were　compared.　Results　show　that　the　optimized　VMD　algorithm　can　successfully　extract　the　early　fault　features　of　the　bearing　and　achieve　the　diagnosis　of　early　bearing　failure.　©　2019,　Editorial　Department　of　Journal　of　Beijing　University　of　Technology.　All　right　reserved.