Research On Fault Diagnosis And Life Prediction Technology Of Molecular Pump

Molecular pump is a kind of precision machinery used in high vacuum environment.It is an important part of high-end analytical instruments such as mass spectrometer,scanning electron microscope,etc.In order to improve the vacuum degree,pumping speed and other performance indicators,the molecular pump usually runs at a very high speed.With the operation of the molecular pump,the performance of the internal vulnerable parts gradually degenerates.If the maintenance is neglected,the failure of the parts or even the damage of the whole instrument may be caused.Therefore,it is necessary to study the operation condition monitoring,abnormal detection,fault diagnosis,performance evaluation and life prediction technology of key components of molecular pump.The work of this paper is divided into the following parts:First of all,analyze each subsystem of the molecular pump,determine the vulnerable key parts in the molecular pump,and carry out the extended FMECA analysis of the key parts,determine the fault symptoms and hazards,fault evolution mode,failure mechanism,sensor type and layout.According to the results of extended FMECA analysis,the PHM technology framework of molecular pump is designed,including anomaly detection method,fault diagnosis(performance evaluation)method and life prediction method.Anomaly detection is based on time series modeling technology,fault diagnosis(performance evaluation)is realized by automatic feature extractor and classifier based on deep learning,and life prediction is realized by degradation model and particle filter.Secondly,in order to improve the performance of the fault diagnosis method,a fault diagnosis method is proposed,which uses the integrated autoencoder as the feature extractor and the integrated neural network as the classifier,and is verified in the open data set of bearing and gearbox faults.The results show that this method is superior,not only accurate in diagnosis,but also less affected by the adverse conditions such as noise.The fault diagnosis experiment of the molecular pump is carried out,and the vibration signals under six different fault conditions are collected.The fault diagnosis of the molecular pump is realized under four different conditions by using the proposed method.Finally,in order to master the performance degradation characteristics of the molecular pump,the accelerated life experiment of the molecular pump and the related data acquisition software are designed.The life-cycle data of the molecular pump is collected and used to evaluate the PHM algorithm of this study.The experimental results show that the anomaly detection algorithm used in this paper can accurately detect the boundaries of different degradation stages.The accuracy of performance evaluation algorithm in classifying the four degradation stages reaches 90%.The life prediction algorithm can predict the fault about 90 hours in advance in the middle of degradation,and the relative error is less than 10%.