The Application Of FARIMA Model In The Fault Diagnosis Of Complicated Mechanical System

In a long time, for complex mechanical system, people want to discover the fault, judgethe degree of fault, analyze and predict the behavior of the system, accurately and timely, somore and more attention was paid to fault diagnosis technology. At the same time, it is avaluable subject in industrial and application of signal detection field. Time series analysismethod is a classical analysis method, and has a unique advantage in fault diagnosis. At mosttimes, vibration signal is established to ARMA model and analyzed, but this modeling me-thod has some limitations. Noticing that vibration signal has long-time memory, this papertries to use FARIMA model to analyze vibration signal.The purpose of this paper is proved that when modeling vibration signal FARIMAmodel is much more accurate than traditional ARMA model, by actual instance, on the basisof introducing the characteristics of long-memory. This paper further analysis the characte-ristic of long-time memory and fractal difference of the FARIMA model, illustrates the con-dition of FARIMA model and its advantage. This paper accounts the steps of modeling ofstationary and non-stationary time series model, what’s more, it summarizes several methodsof parameter estimation, all show that FARIMA model is different with ARMA model,thought the former is the promotion of the latter. Through the analysis of the results of Bent-ly experiment about the steam turbine rotor vibration signal, and the results of simulationand parameter estimation by MATLAB and SAS software, this paper show that this model-ing method is much more effective than traditional ARMA model. When analyzing FARIMAmodel, we consider S α S-FARIMA model and T-V-FARIMA model whose parameters arevarying with time, these special circumstances validate the flexibility and effectiveness ofthe FARIMA model when modeling some actual data, indicate the direction of fixing thismodel.