Aiming at the difficulty of fault diagnosis and recognition of rolling bearing,a fault diagnosis and recognition method was proposed based on the combination of local mean decomposition (LMD),particle swarm optimization (PSO) and extreme learning machine (ELM)Firstly,the vibration signal is decomposed into a series of product components (PF) from high frequency to low frequency by LMDSecondly,the correlation coefficients between each PF component and the original signal were calculated,and the PF components with high correlation were selected as the feature components,and the feature components were composed of feature vectorsFinally,PSO-ELM network model was used to train the training set and test setThe simulation of bearing data from Western Reserve University verifies that the degree of mode aliasing is lower in LMD than that of empirical mode decomposition (EMD)The method was applied to the rolling bearings of a certain type of linear cutting machine tool,and the rolling bearings of three different states were diagnosed and recognizedThe EMD-PSO-ELM method was compared with LMD-PSO-ELM methodThe experimental results show that LMD-PSO-ELM can not only identify the fault types of rolling bearings,but also greatly improve the accuracy of identification.