Rolling Mill Vibration Analysis And Fault Diagnosis Methods

In recent years, the increasing demand for steel causes the continuous development ofthe global steel industry.It will cause huge economic losses when rolling mill in the event offailure as the main production equipment.For that,the research on fault diagnosis of rollingmill has the essential and significance meaning.In this paper, characteristics of common faults of rolling mill are presented. Bearingfailure is one of the four major fault of rolling mill, and mainly analyzed the performanceand failure modes of the rolling mill bearings, the characteristics of rolling mill bearing andworking environment.A new fault diagnosis method based on wavelet packet analysis, singular spectrumanalysis and Support vector machine (SVM) on vibration analysis is proposed.The methodadopts the wavelet packet de-noising method for noise reduction processing which suitablefor nonstationary and nonlinear properties. This method can effectively get rid of theinterference signal mixed in the original signal, and reduce the error of the fault diagnosis. Inorder to prove the effect of noise reduction,the paper presents a simulation of the rolling millroll bearing failure driven by DC motor based on the software MATLAB. The waveletpacket de-noising method is used to denoise rolling mill bearings. As the characteristic ofvariable load and variable speed of rolling mill bearings,a simple time series method forbearing fault feature extraction using singular spectrum analysis (SSA) of the vibrationsignal is proposed. The method is easy to implement and fault feature is noise immune. SSAis used for the decomposition of the acquired signals into an additive set of principalcomponents. SSA is adopted to extract fault feature of rolling mill bearings. The comparisonof energy spectrum and singular and the comparison of different loads of singular spectrumare analysed.The singular values (SV) of the selected SV number which can clearlydistinguish the faults of rolling mill bearings are adopted as the fault features.Finally, anoptimal SVM model is established and trained for vibration signals in different workingconditions by using the fault feature vectors,and the input is the selected singulai values.TheSVM is trained by using the SV selected of normal bearing, inner fault, outer fault and bodyfault,and then to diagnoze fault.Cross-validation results show that the proposed method is accurate and effective forrolling mill bearing fault diagnosis.The method proposed in this paper also can be applied to some rotation machinery fault diagnosis.