State Monitoring Of The Seal Face Based On Acoustic Emission

Mechanical seals are widely used in petrochemical,electric power,metallurgy,water supply and drainage,aerospace and other industries.With the rapid development of China's industry,the demand for mechanical seal products,especially high-end mechanical seal products is increasing rapidly.The research object of this paper is diamond coated mechanical seal which is coated diamond through the Chemical Vapor Deposition(CVD).It is a high end mechanical seal and whit high abrasive resistance and long life-span.This kind of seals are usually applied in nuclear power,aerospace and other fields.In this paper,a technology for monitoring the state of such seals is studied.Its main contents are as follows:(1)Studied the failure mechanism of diamond coated mechanical seal coating.the failure mechanism of the diamond coated mechanical seal is summarized as the failure mechanism of the Diamond Like Carbon films(DLC).The failure mechanism of DLC can be divided into three processes: 1)stage of the running in: in this stage the friction coefficient rises quickly and then decelerates.2)stage of stability: in this stage tends to be stable.3)stage of failure: in this stage friction coefficient increases slowly at first and then accelerates rising,(2)Confirmed that the failure of diamond coating can be divided into three periods.Designed a monitoring method of mechanical seal which use acoustic emission and assisted by temperature.The trend of temperature shows the classification of the failure mechanism of DLC is rational.(3)Analyzed and filtered the acoustic emission signals based on the generalized S transform.As the frequency range of acoustic emission signal is wide(frequency range from a few to several hundred kilohertz)and need better frequency resolution in the high frequency range.This paper proposed to use generalized S transform to analysis acoustic emission signal after comparing generalized S transform with the methods of short time Fourier transform,wavelet package analysis and S transform.Then in this paper the acoustic emission signal is filtered by using the method generalized S transform.It is found that the acoustic emission signal after the filtering is closer to the typical acoustic emission signal,which shows the effectiveness of the filter method.(4)Classified the wear status of mechanical seal by multi class GA-TWSVM.TWSVM is a two classification method based on support vector machine.But the parameters of TWSVM are difficult to choose,and only two classifications are limited in TWSVM.This paper proposes a multi classification TWSVM based on genetic algorithm optimization,and classifies the friction state of the seal according to the characteristics of acoustic emission signal which is collected in the mechanical seal monitoring experiment.The classification accuracy rate is higher,and the classification accuracy rate is about 4% higher than that of the SVM.It is proved that this method is effective for identifying the friction state of diamond coated mechanical seals.