To　effectively　extract　the　weak　fault　features　of　core　components　of　a　large-scale　power　equipment,　a　new　fault　diagnosis　method　based　on　noise-assisted　EMD-1.5　dimension　spectrum　was　proposed.　The　interval　characteristic　of　signal　extreme　value　points　affected　the　appearance　of　mode　mixing　in　the　EMD　method.　According　to　this　status,　an　evaluation　method　of　the　interval　characteristic　of　signal　extreme　value　points　was　proposed,　the　principle　of　Gaussian　white　noise　helping　to　avoid　mode　mixing　was　analyzed.　The　noise-assisted　EMD　method　was　obtained　by　adding　Gaussian　white　noise　to　the　original　signal　to　improve　the　capability　of　signal　anti-alias　decomposition.　Meanwhile,　the　noise　assisted　EMD　method　combined　with　a　1.5　dimension　spectrum　to　get　a　new　fault　diagnosis　method　having　the　capability　of　effective　anti-alias　decomposition　and　extracting　the　nonlinear　coupling　feature　in　order　to　extract　the　weak　fault　features.　Finally,　the　noise　assisted　EMD-1.5　dimension　spectrum　was　effectively　verified　by　simulation　experiments　and　engineering　examples　of　electric　locomotive　rolling　bearing　fault　diagnosis.