Application Of Complex Network In Fault Diagnostics Of Centrifugal Pump

In recent years, fault diagnosis as a new technology which crosses variousdisciplines, has developed rapidly and produced huge economic benefits. Because thesignal of failure centrifugal pumps is a non-stationary signal, it is necessary to selectthe appropriate signal processing method which is suitable for non-stationary signal.Complex networks, which provide us with a new viewpoint and an effective toolfor understanding a complex system from the relations between the elements in aglobal way, not only may be a powerful tool for revealing information embedded intime series but also can be used for studying nonlinear dynamic systems that can notbe perfectly described by theoretical model. In this thesis, we focus our research onthe different states of centrifugal. Based on the different state signals measured, wepropose and construct different types of complex networks. Furthermore, throughinvestigating the different state complex networks, we make a systematic study on thecomplex structure and the nonlinear dynamics of the flow pattern transition, and thetheory improves the agility, efficiency and accuracy of the fault diagnosis.First, the concept of complex network is introduced. Through the degree,average shortest path and gathered coefficient analysis, complex network has a betterclassification results. The complex network provides a new approach for uncoveringthe nonlinear dynamic characteristics of different sate of centrifugal pumps.Secondly, a unique method for constructing complex networks from a time seriesbased on phase space reconstruction is proposed. To overcome the drawback of theRQA which only quantifies the recurrence property from the view of recurrence pointand line structure, the RCN is used to obtain more spatial recurrence propertyinformation. After that we calculate the ideal solution in order to obtain theintegration results of node importance in complex networks It is also easy to evaluatethe importance evaluation criteria and then greatly improve the accuracy of the faultdiagnosis. Finally, k-core decomposition of several typical nonlinear series is studied, andthen based on this, a fast algorithm method is employed to divide the communitystructure. According to the definition of k-core, the centrifugal pumps complexnetwork is divided into different cores at router level. Analyzing the maincharacteristic quantities of every k-core, it was found that the method of complexnetwork community structure and k-core decomposition has better diagnosis effect.