For fault diagnosis of rolling bearings in industrial equipment,the measured signal is often disturbed by high frequency noise and intermittent noise,resulting in low accuracy of signal decomposition and feature extraction.To solve this problem,a method based on total variation denoising (TVD) and improved local mean decomposition (LMD),was proposed.The total variation method was adopted to reduce noise,and appropriate regularization parameter was selected,so that the noise reduced signal had a high signal-to-noise ratio (SNR) and a low root mean square error (RMSE).The local mean decomposition was performed on the noise-reduced signal,and the best PF components were selected according to the interrelationships value and the kurtosis value,and the envelope analysis was performed to realize the extraction of fault features.Experimental verification of the measured signals was performed.The results show that by using the method,an effective noise reduction effect can be achieved and the fault features in complex vibration signals can be accurately extracted.