Application Of Data Mining Technology On Fault Diagnosis For Rotary Machinery

The rotating machinery has a wide application in the engineering area. In recent years with the devolopment of science and technology, there is a enormous rise in the complexity level of the relating devices. If the malfunction occurs, it will not only r educe the productive efficiency, but also trigger off a industrial incident in which seve re casualties are involved. So it requires the researchers to dig deeply into the fault d iagnosis technology to find new methods that fit/follow the devolopment of machinery.As a newly-emerging technology, data mining integrates all kinds of smart algorithms which can deal efficiently with large data sample. This paper is exclusively concerne d with these smart algorithms of the data mining technology system which is applied in the pattern recognition of the rotating machinery fault diagnosis processing.Afterwards more studies have been done on the algorithms of data mining applie d in the fault diagnosis technology to definitize the process of data mining and deter mine that the object of algorithm study is Fuzzy c-Means. It’s supposed for the study on the data preprocessing of the data signals to firstly complement the missing data by applying algorithm of grey prediction.The vibration is one of the most basic mechanical motion which happens in the movements of the rotating machinery. When the machinery malfunctions, it will effect the stationary signal of the normal vibration. Consequently the fault state and type c ould be directly reflected through the detection, collection and feature extraction of the rotating machinery’s vibration signals. This paper selects as its object of data study t he feature data samples which are obtained through the feature extraction of dimensio nless time-domain feature of the vibration signals collected from two different experim ental platforms.This paper aims at analysing the disadvantages of the FCM algorithm,which is applied i n the data classification experiment of rolling bearing faults.To reduce the effect of outliers,th e concept of relative fuzzy index is proposed,and the update value of normalized degree of me mbership is set.Moreover,the author proposes a unsupervised algorithm that combines the FC M algorithm with the S2 FCM algorithm,and applies it in the data classification experiment o f rolling bearing faults.After using the data of both researching platforms to verify its efficienc y,the results show that this new algorithm could raise the convergence rate of the merit functio n efficiently,and thus is evidently superior to the traditional algorithm in terms of accuracy rat e of the results of cluster analysis.