Gear Fault Diagnosis Using Wavelet Packet Transform For Feature Extraction And Decision Using Flow Graph

Gear is an important transmission component in the production of modernindustrial, used in all areas of production and life, its stable and reliable work hasgreat social significance. But the failures of the gear are inevitable, since that inorder to reduce the loss of production caused by gear failure; we need to real-timemonitor of the gear, to diagnosis gear fault, replaced the fault gear before gearfailure to avoid serious production losses.In this paper proposed a gear fault diagnosis based on wavelet packet for faultfeature extraction and flow graph decision-making method for gear fault detectionand diagnosis.This paper analysis the variations of gear fault vibration signal, and thencompare the vibration signal extracted from experimental with the theory analysis,to verify the accuracy of theory analysis; using time-domain and frequency domainsign attributes to characterize these gear vibration signal.Through theoretical analysis drawn gear vibration signal characteristic, thenuse of wavelet packet to extract fault sign attributes. This paper introduce a kind ofnew method for wavelet de-noise, eliminating the problems of wavelet de-noisedecompose level and de-noise threshold value selection, at the same time analysisa kind of wavelet packet transform method, eliminating the frequency andfrequency band confusion, reducing the error in fault sign attributes extraction.According to the vibration theoretical analysis, combined with the actual signal toextract gear’s fault sign attribute information.In order to analysis the extracted sign attributes’s importance and relationshipwith the final decisions. This paper proposed a flow graph representation method,for sign attributes reduce, to eliminate redundant information. After the reductionto generate the decision rules, analysis the relationship of sign attributes with thegear fault. Meanwhile introduces the incremental learning algorithm of flow graphto improve the adaptability of flow graph.At last, using fault simulation platform to simulate different conditions anddifferent gear fault vibration signals. Combined with the first to four chapterstheoretical analysis, using wavelet packet analysis to extract sign attributes, usingflow graph for sign attributes reduction and extraction of decision rules, obtainedthe decision rules of different conditions and gear fault. The results demonstratethat this method can accurately and reliably detect failure modes in a gearbox.