Current improved Empirical Mode Decomposition(EMD)methods enhance the accurate identification of peak and valley points in mechanical signals through noise-assisted filtering techniques,thereby improving the mode deco...Current improved Empirical Mode Decomposition(EMD)methods enhance the accurate identification of peak and valley points in mechanical signals through noise-assisted filtering techniques,thereby improving the mode decomposition performance,which is of great significance in extracting fault features from mechanical signals.However,noise-assisted filtering leads to the loss of critical features in mechanical signals and introduces a large amount of residual noise into Intrinsic Mode Functions(IMFs)that obscure signal features.To address these issues,a Precise Identification-based Mode Decomposition(PIMD)method is proposed.This method directly enhances the ability of EMD to precisely identify peak and valley points by using a proposed precise identifi-cation approach,which improves mode decomposition performance and avoids the negative impacts of noise-assisted filtering,thus benefiting the extraction of more mechanical fault features.Simulation results show that the proposed PIMD method can precisely identify peak and valley points of signals with noise of different signal-tonoise ratios and perform a highly rigorous high-low frequency decomposition,significantly outperforming EMD.Finally,mechanical fault diagnostic experiments on four bearing cases and two gear cases demonstrate that,compared to four mainstream methods,the PIMD method exhibits the best mode decomposition perfor-mance and can extract more and clearer mechanical fault features.展开更多
基金Supported by National Natural Science Foundation of China(Grant No.52075236)Opening Foundation of Intelligent Manufacturing Technology(Shantou University),Ministry of Education(Grant No.STME2024002)+1 种基金Hundred Doctor and Hundred Enterprise,Science and Technology Project,Ji'an City(Grant No.42064001)Guangdong Provincial University Innovation Team Project(Grant No.2020KCXTD012).
文摘Current improved Empirical Mode Decomposition(EMD)methods enhance the accurate identification of peak and valley points in mechanical signals through noise-assisted filtering techniques,thereby improving the mode decomposition performance,which is of great significance in extracting fault features from mechanical signals.However,noise-assisted filtering leads to the loss of critical features in mechanical signals and introduces a large amount of residual noise into Intrinsic Mode Functions(IMFs)that obscure signal features.To address these issues,a Precise Identification-based Mode Decomposition(PIMD)method is proposed.This method directly enhances the ability of EMD to precisely identify peak and valley points by using a proposed precise identifi-cation approach,which improves mode decomposition performance and avoids the negative impacts of noise-assisted filtering,thus benefiting the extraction of more mechanical fault features.Simulation results show that the proposed PIMD method can precisely identify peak and valley points of signals with noise of different signal-tonoise ratios and perform a highly rigorous high-low frequency decomposition,significantly outperforming EMD.Finally,mechanical fault diagnostic experiments on four bearing cases and two gear cases demonstrate that,compared to four mainstream methods,the PIMD method exhibits the best mode decomposition perfor-mance and can extract more and clearer mechanical fault features.
