Rolling bearings as an important part of rotating machinery,it operates in harsh environments resulting in non-linear and non-smooth vibration signals,which makes it difficult to distinguish between fault signals and normal signals.In view of this,intelligent fault diagnosis method combining multi-modal mutual dimensionless indicators (MMDI) and broad learning system（BLS） was proposed.The optimization complete ensemble empirical mode decomposition with adaptive noise (OCEEMDAN) was used to decompose and preprocess the observable signal of the bearing with wavelet threshold,and the effective intrinsic mode function (IMF) was reconstructed and MDI was extracted to construct the MMDI.The back propagation algorithm and superposition module mode were used to optimize the BLS,which could quickly identify different types of faults.Finally,the proposed method was verified through the datasets provided by Case Western Reserve University Bearing Data Center and some laboratory,the average accuracy was 99.8% and 100%,respectively,the portability and effectiveness of the method were verified.