Fault Prediction Of Rotating Machines Based On Vibration

In order to ensure safety operation of large rotating machinery,it is essential to conduct condition monitoring and failure prediction for key-parts.Based on prediction results,a reasonable arrangement for maintenance schedule of large rotary machine is available.In this way,maintenance cost can be reduced and security incidents can be avoided.A fault prediction method of rotary machine based on vibration is researched.Using this method remaining useful life of rotating machinery parts can be obtained.Firstly,main failure modes and fault types of rotating machine key parts are illustrated.Different fault mechanisms are introduced respectively.Failures of rotating machine can be divided into gradual changing failure and abrupt failure.The gradual changing failure can be predicted using some methods.Secondly,implementation procedures of machinery fault prediction method and the operating method for every individual procedure are studied.The main procedures of fault prediction include initialization of fault prediction,signal selection,feature extraction and selection,trend prediction and calculation of estimated failure time.The situations of project and relevant standards are taken into account in the whole process of the implementation process of failure prediction method and the design of operating methods of each procedure.Furthermore,various trend forecast methods for short-term prediction and long-term prediction are studied and discussed.A multiple forward rolling average forecasting way is applied to optimize BPNN(BP neural network)forecasting method and CSVR(combined support vector regression)method.After optimizing process,prediction performance become better.Long-term forecast method based on curve fitting is studied,too.Results of long-term prediction using different fitting functions are compared.Finally,a wind turbine gearbox bearing failure prediction case is studied.The fault prediction methods based on vibration of rotating machine has been verified using case data.The results suggest that methods studied in this paper can effectively predict rotating machinery failure.The prediction results are of good accuracy.The methods can be applied in practical projects.