Research On Fault Diagnosis And Degradation Trend Prediction Methods For Rolling Bearings

Rolling bearing is one of the key components in the mechanical equipment,the running condition of rolling bearing affect the performance of the whole mechanical equipment directly.Because of long-term work under high speed and high load,there are various faults,therefore,the study of fault diagnosis and prediction of rolling bearing fault is very important to ensure the normal operation of machinery and avoid catastrophic accidents.Taking the vibration signals of rolling bearing as the study case,we studied the early bearing fault identification,intelligent fault diagnosis and bearing degradation trend prediction.The main contents are as follows:1.In the early fault signals of rolling bearing,the useful fault information is eas-y overwhelmed by strong noise,which is difficult to extract the fault feature directly.We studied the principle of maximum correlated kurtosis deconvolution and autocorrela-tion,analysed their disadvantages.The method of maximum correlated kurtosis decon-volution and autocorrelation were used to reduce the noise of bearing early fault signa-Is,identified early bearing faults by contrasting bearing fault characteristic frequency with frequency in envelope spectrum.Experimental results show that:the noise reducti-on effect of this method is more ideal,the early bearing fault can be identified by th-is method.2.The wavelet packet was used as input vector of diagnostic model,the deep belief network was used as a diagnostic model.The effect of wavelet packet decomposition,deep belief network hidden layer combination and the proportion of training samples and test samples on fault diagnosis were analyzed in detail.Comparing and analyzing the three following fault diagnosis methods,deep belief network,support vector machine and BP neural network,the results show that the deep belief networks' diagnostic accuracy is the highest,it can identify the location and extent of damage of bearing fault.3.The index of bearing degradation process was extracted by support vector data description.The advantages of different kernel functions in kernel extreme learning machine are analyzed.Researching the Hybrid kernel extreme learning machine,parameters of Hybrid kernel extreme learning machine were optimized by particle swarm optimization.All vibration data of rolling bearings is analyzed,the analysis results show that the support vector data description algorithm can be used to extract the index of bearing degradation trend.The hybrid kernel extreme learning machine optimized by Particle swarm optimization can predict the bearing degradation trend,and the prediction result of the model is ideal.4.The fault diagnosis and degradation trend prediction system of rolling bearing is designed and developed by hybrid programming with MATLAB and C#.The system is analyzed by using the fault vibration data of rolling bearing,the results show the correct of the system operation,the system is easy to used,and the efficiency is improved.