Research On Rough Set Attribute Reduction Method For Bearing Fault Knowledge Discovery

With the progress of science and technology,all kinds of rotating machinery are developing in the direction of intelligent,high quality and efficient operation.Comprehensive monitoring of rotating machinery will collect a large number of vibration signals,which promotes the field of fault diagnosis into the era of big data.However,a large number of data contains a lot of invalid redundant information.How to remove the redundant information and mine the most real and valuable information to help realize intelligent fault knowledge discovery has become an urgent problem.As a data analysis tool that can describe uncertainty,rough set theory is used in intelligent decision-making,which promotes the rapid development of artificial intelligence.The key problems of rough set theory are knowledge acquisition method and attribute reduction algorithm,so the theory is widely used in the field of mechanical fault diagnosis.In this paper,rough set theory is used as a knowledge discovery tool,and the attribute reduction of bearing fault feature set is discussed,combined with intelligent classifier for fault recognition.The main research contents and achievements are as follows:(1)In order to solve the problem that the influence of different granularity weights on the reduction results is not considered in attribute reduction of neighborhood multi-granularity rough set model,a neighborhood weighted multi-granularity rough set model is proposed to reduce the attributes of fault data sets.In the decision information system,the model gives a weight to each condition attribute,redefines the approximation set,dependency and importance under the model,and establishes an attribute reduction algorithm of fault data set based on neighborhood weighted multi-granularity rough set;then the sensitive feature subset obtained after reduction is input into KNN classifier for pattern recognition;finally,the bearing fault data set is used for fault diagnosis Data sets verify the effectiveness of the method.(2)Aiming at the problems of uncertain parameters and too many parameters in decision rough set model,a method of parameter determination is proposed 。and a neighborhood single parameter decision rough set model is constructed.At the same time,a fault diagnosis method combining neighborhood single parameter decision rough set model and non naive Bayesian classifier is designed.This method uses neighborhood single parameter decision rough set model to reduce the attributes of bearing fault feature set,and then inputs the extracted low dimensional sensitive feature subset into non naive Bayesian classifier for pattern recognition.The designed fault diagnosis method integrates the advantages of neighborhood single parameter decision rough set in attribute reduction and non naive Bayesian classifier in pattern recognition.The effectiveness of the method is verified by constructing bearing fault feature set.(3)Aiming at the scientific management of fault data in the development of industrial big data intelligent decision-making technology,a set of rotating machinery fault diagnosis system is designed based on C #.The system is composed of database module and fault diagnosis function module,which is verified by bearing fault data set.The application shows that the system can effectively realize the storage of fault data and complete the knowledge discovery of fault data,so as to verify the effectiveness of the fault diagnosis system.Based on rough set theory,this thesis applies it to attribute reduction of fault data set.Under the background of industrial big data,it provides a new idea for eliminating redundant information,acquiring sensitive features of faults and improving the accuracy of fault identification,which makes intelligent fault decision-making technology have a better development.