Symplectic Geometry Mode Decomposition And Its Application In Fault Diagnosis Of Gearbox

As a key running part of mechanical equipment,gearboxes are widely used.Once a certain part of gearbox fails,it may cause economic losses and even casualties.Therefore,in-depth research on the fault diagnosis technology of gearboxes is of great significance for ensuring the safe operation of mechanical equipment and preventing accidents.At present,the common signal processing methods for fault diagnosis based on vibration signals have certain defects,such as lack of adaptability,mode confusion,and poor robustness.Therefore,it is difficult to process complex noisy signals in actual engineering.As an emerging adaptive signal decomposition method proposed in recent years,symplectic geometry mode decomposition(SGMD)method has the advantages of keeping the inherent characteristics of time series unchanged and suppressing modal confusion,and is suitable for processing vibration signals of mechanical faults.However,SGMD also has theoretical and practical problems that need to be improved.Therefore,based on the research and improvement of SGMD,taking gearbox as the main research object,this paper uses SGMD method to extract gearbox fault features,and makes a series of in-depth research on gearbox fault diagnosis methods.Firstly,the decomposition ability of SGMD method is studied in depth.Three different types of simulation signals were established to verify the decomposition ability and noise robustness of SGMD.By analyzing the decomposition ability of SGMD under different signal-to-noise ratios(SNR),the applicability of SGMD for signals with different degrees of noise is verified.Secondly,to avoid the weak fault feature being concealed by strong background noise and decomposition error,eigenvalue decomposition(EVD)and calculus operators are used to improve SGMD,and the enhanced symplectic eigenmode decomposition method(ESCMD)was obtained.The experiments have proved the feasibility of the proposed improvement ideas and the noise robustness and feature enhancement performance of ESCMD,and the proposed method is applied to gearbox fault diagnosis.Subsequently,for the recognition problem of multi-sensor signal feature tensors,the nearest neighbor convex hull classification(NNCHC)is extended in tensor space to form the nearest neighbor convex hull tensor classification(NNCHTC)method.For the recognition problem of single vibration signal feature vectors,combined with the idea of convex hull and deep learning,the deep stacking l1-norm center configuration convex hull(DSl1C3H)method is proposed.Finally,ESCMD is combined with NNCHTC and DSl1C3H to form two kinds of intelligent fault diagnosis methods,and apply them to gearbox fault diagnosis.The two have different application scopes.In the case of collecting multi-source vibration signals through mulit-sensor,NNCHTC has high recognition rate and excellent robustness.However,when only single vibration signal can be collected due to limited conditions,DSl1C3H has more advantages than NNCHTC,and has accurate recognition rate and excellent robustness.