Study Of Gear Fault Diagnosis Method Based On Flow Graphs With Incomplete Information

As an important part of the rotating machinery,gears has been widely used in power plants,helicopters,heavy trucks and other equipment.Due to bad operating environment,complex conditions,gear state information is incomplete.Besides,the human practice is always limited by the objective environment and conditions.Therefore,the obtained diagnosis information of fault modes is often somewhat incomplete.In this thesis,valued character relation is used to process gear incomplete diagnosis information.Then,on the base of flow graph theory,a knowledge representation method and a knowledge discovery method of incomplete information for gear fault diagnosis are put forward,and a mode identification method of gear fault is proposed.Finally,a gearbox fault diagnosis experiment was conducted to verify these methods.This thesis includes the following parts:Only one semanteme is utilized to understand,analyze and deal with incomplete information in most of the existing generalized indiscernibility relations.They can not be used to understand,analyze and process gear incomplete fault diagnosis information that is brought about by a variety of causes.To solve this problem,a similarity calculation method between two samples is presented on the basis of character relation.Then,character relation is improved and valued character relation is presented.Finally,an incomplete information processing method of gear fault diagnosis based on valued character relation is proposed to process gear incomplete fault diagnosis information.The case of incomplete information system processing for fault diagnosis of automatic gearbox is used to validate the practice,effectiveness,and accuracy of this method.The existing knowledge representation methods of incomplete information for gear fault diagnosis can not intuitively represent fault attribute values and the dependent relationship between fault condition attribute values and fault decision attribute values.Moreover,these methods are hard to describe dependence degree among attributes.To address this problem,incomplete flow graph is firstly defined according to flow graph.Then,a construction algorithm of incomplete flow graph is presented on the basis of the definitions of incomplete flow graph.Finally,a knowledge representation method of incomplete information for gear fault diagnosis based on flow graphs is proposed to solve the problem of knowledge representation of gear incomplete fault diagnosis.A knowledge representation case validates that this method can intuitively represent an incomplete fault diagnosis information that simultaneously contains three unknown attribute values,and quantitatively represent dependence degree among attributes to facilitate users’ to understanding and analyzing.Most of the existing knowledge discovery methods can be utilized to aquire fault diagnosis knowledge from incomplete information that only contains one kind of unknown attribute values.A certain method can be utilized to aquire fault diagnosis knowledge from incomplete information that simultaneously contains two unknown attribute values.However,the knowledge discovery process of the method is cryptic and difficult to understand.To address this problem,an assignment reduction algorithm based on valued character relation is presented on the basis of valued character relation.Then,an attribute reduction algorithm of incomplete flow graph is presented according to the assignment reduction algorithm based on valued character relation.Finally,a knowledge discovery method of incomplete information for gear fault diagnosis based on flow graphs is proposed to address knowledge discovery problem of gear incomplete fault diagnosis information.A case validates that this method can be used to directly discovery fault diagnosis knowledge from incomplete information system that simultaneously contains three unknown attribute values.Furthermore,the knowledge discovery results are intuitive,easy and clear.Nowadays,most mode identification methods of gear faults can not intuitively represent fault condition attribute values and dependent relationship among attribute values.The identification models of some methods are very complex,and the identification process is still obscure and hard to understand.To address this problem,the definitions of priori probability,posterior probability,and conditional probability in flow graph are given to realize flow graph probabilistic.Then,a mode identification algorithm based on flow graphs is presented on the basis of flow graph probabilistic.Finally,a mode identification method of gear faults based on flow graphs is proposed.Experimental results show this method can accurately identify gear fault modes.Besides,the structure of mode identification model is simple and the strategy of identification process is clear.Gearbox experiment assembly developed by Spectra Quest Ltd Company is used to verify the proposed gear fault diagnosis method based on flow graphs with incomplete information.Vibration analysis methods and oil sample analysis methods are utilized to extract gear incomplete fault diagnosis information.Firstly,an incomplete information system for gear fault diagnosis is constructed.Valued character relation is used to process this information system.Then,incomplete flow graph for fault diagnosis of gears is constructed to discovery fault diagnosis knowledge.Finally,a mode identification algorithm based on flow graphs is utilized to determine the fault mode of test samples.Experimental results show that the method pocesses a very high accuracy.Moreover,the fault diagnosis process is intuitive and the fault diagnosis strategy is clear.