The Graph Of Rotor System Fault Aided Decision Making Method

Under the background of industrial modernization transformation and upgrading,mechanical equipment is increasingly large,complicated and intelligent,and the state data generated by it is also increasing exponentially,which brings great trouble to the current fault diagnosis methods of vibration signal analysis,because the massive data is mixed with a lot of redundant information.This diagnosis method which only relies on vibration signal to identify fault mode has a high requirement on vibration fault data set and has not made a significant breakthrough in intelligent research.One of the main problems is that we humans are much more likely to accept and understand textual data than a bunch of numbers with no discernable pattern.This phenomenon is more deeply reflected in the design of fault diagnosis software.Generally speaking,most of the fault diagnosis software can only rely on fault data to realize fault identification function,and the intelligent system with fault decision function expected by people still remains at the theoretical level.Notably,in other fields,researchers have made breakthroughs in human-computer interaction using knowledge mapping technology.Based on the above reasons,on the basis of studying the fault identification results obtained by dimensionality reduction algorithm,this paper applies the knowledge graph model to the field of rotor fault visualization and auxiliary decision making.The main work overview is as follows:(1)Aiming at the difficulty of fault classification caused by the redundancy of feature attributes in high-dimensional fault data sets of rotating machinery,a dimension reduction method of fault data sets based on local Fisher Principal component discriminant analysis(LFPCDA)was proposed.The Laplacian score algorithm was used to filter the redundant features of the high-dimensional fault feature set,and the principal component calculation was integrated into the local Fisher discriminant analysis(LFDA).The principal component that can best reflect the fault nature could be selected adaptively to form the projection matrix.The reliability and generalization of the proposed algorithm are proved by using the rotor fault data set simulated by a two-span rotor bench.(2)Aiming at the problem that the neural network model has a "black box" feature in the fault identification process,the rotor fault data and fault mechanism diagram are constructed.The diagram clearly shows the internal relationship between various kinds of faults,which can be used for operators to trace the root cause of equipment faults and the direction of equipment maintenance.(3)Aiming at the problem that advanced rotor fault diagnosis technology is difficult to be applied to practical engineering,a new method based on.NET rotor fault diagnosis and decision aid desktop application,the program embedded in the proposed LFPCDA data dimension reduction algorithm and rotor fault map.The program consists of rotor fault identification module and fault assistant decision module.The feasibility of the program is verified by using the rotor fault data set simulated by the double-span rotor test platform.