Research On Intelligent Fault Diagnosis Based On Cluster Analysis For Rotating Machinery

Rotating machines have been widely used in modern manufacturing,their reliability is thus important for the industrial processes.With the rapid development of science and technology,mechanical equipment in modern industry is large,precise and complex,which cause great challenges for maintenance and automatic fault diagnosis.With the rise of the novel generation of artificial intelligence and big data technology,some new research ideas on the fault diagnosis have been emerged.Based on the new techniques of cluster analysis,modern signal processing and feature extraction,several novel intelligent approaches for fault diagnosis of machines have been developed in this work,which are able to reliably and accurately identify different faults with less prior knowledge and artificial involving operations and greatly improve the level of intelligence,compared with traditional techniques.First,the framework of intelligent fault diagnosis approach as well as cluster analysis methods for fault diagnosis have been simply reviewed.It is found that cluster analysis technique is an effectively and reliably way for intelligent fault diagnosis and has great potential for future development.Then,three kinds of modern signal processing and feature extraction technology are expounded,including Wavelet Packet Transform(WPT),Ensemble Empirical Mode Decomposition(EEMD),and Variational Mode Decomposition(VMD).Meanwhile,clustering analysis technology has been also briefly summarized.Particularly,Affinity Propagation(AP)clustering and Density Peaks Search(DPS)clustering methods are both introduced in detail.Their advantages will be illustrated separately.Next,a novel intelligent fault diagnosis method for bearings based on adaptive feature selection technique and AP clustering is proposed for a non-expert to carry out diagnosis operations.The proposed original self-weight algorithm is adopted to automatically select some sensitive features(SFs)from the raw feature sets.Subsequently,redundant features adaptively eliminated by AP method.The proposed intelligent method is then applied to the bearing fault diagnosis.Results demonstrate that the proposed method is able to reliably and accurately identify different fault categories and severities of bearings with less time,compared with other methods.Finally,in order to automatically identify fault categories for online applications and further enhance the diagnosis accuracy,a novel adaptive density peaks search(ADPS)clustering algorithm and a novel VMD feature denoising technique have been developed in this thesis.It is the first time to employed VMD for feature denoising in feature domain.An example is used to verify that VMD-based feature trend extraction technique can effectively improve the clustering performance.Self-weight algorithm is then also adopted herein to achieve sensitive features for fault diagnosis.Results of bearing and gear fault diagnosis have well demonstrated that the proposed method is able to reliably and accurately identify different faults with less prior knowledge and diagnostic expertise.Moreover,the proposed technique can be adopted to adaptively monitor different conditions using unlabeled bearing run-to-failure testing data,which also shows that it is well suitable for online applications and thus has broad application prospects.