Remaining Life Prediction Of Aero Engine Rolling Bearing Based On Improved Logistic Regression Model

As the key part of rotating machinery,the rolling bearing is one of the most widely used and easily broken.The performance of the rolling bearing directly restricts the reliability of the machine and the safety of the workers.If the fault diagnosis and the remaining life prediction of the rolling bearing can be carried out in time and make up the timely maintenance plan,the accident can be effectively avoid.Therefore,reliable rolling bearing residual life prediction is an important guarantee for the safely work of rotating machinery.In this paper,the research object of rolling bearing and the fault pattern recognition of the rolling bearing vibration signal is studied,and the life prediction method of the rolling bearing is proposed based on the improved Logistic regression model.The article is organized as follows:First of all,taking the aero engine intermediate bearing(a type of rolling bearing)as the research object,and simulating the working environment of the aero engine,the double bearing rotor test rig and its loading device are designed.The typical failure test and full life accelerated failure test of the intermediate bearing are studied.According to the characteristics of weak vibration signal and complex transmission path,a fault diagnosis method based on rough set and neural network is proposed.The method uses the wavelet packet to diagnose the original signal,and then the noise reduction signal is inputting to the rough set to select the effective measuring point data.Finally,the data of the measuring point is input into the neural network to realize the fault pattern recognition.Secondly,the remaining life prediction method of rolling bearing based on PCA and improved Logistic regression model is put forward.First of all,the Logistic regression model is studied,and it is found that the residual life prediction cannot take into account the degradation trend of bearing and cannot eliminate the influence of random vibration signal on the remaining life prediction value.The Logistic regression model was improved to overcome the above shortcomings.In this paper,PCA is used to reduce the dimension of the useful features in the time domain,frequency domain and time domain,which can represent the degradation of the bearing.The principal component with cumulative contribution rate of more than 95% is chosen as the covariate.The relative characteristic value method is used in this paper,which greatly enhances the applicability of the algorithm.Finally,based on the LabVIEW,the rolling bearing fault diagnosis and residual life prediction system is designed and developed.The system is divided into data acquisition and storage,feature extraction and alarm shutdown,residual life prediction module.According to the structure and working principle of the test rig,the system mainly collects the signals of acceleration,temperature,displacement and speed.According to the needs of this algorithm will extract a large number of time domain,frequency domain,and time-frequency domain features.In the state monitoring,the kurtosis value and the effective value are used as the alarm index.Life prediction module includes PCA dimension reduction,improved Logistic model parameter estimation and residual life prediction.