| Blades are the key components of large-scale rotating machinery.Blade fatigue failure is the main cause of rotating machinery failures.Online monitoring of blade vibration parameters is an important way to predict fatigue failure and master the failure mechanism.Blade vibration measurement based on the blade tip timing（BTT）is the most effective method for non-contact online monitoring of blade vibration.The BTT monitors blade vibration based on the arrival time of the blade.The BTT has the characteristics of passive sampling,extremely under-sampling,and non-equal interval sampling,and so on.With the reliability,prognostics,and health management（PHM）development of large-scale rotating machinery,it is a higher requirement for blade vibration parameters identification based on the BTT.This paper mainly focuses on the problems of vibration event locating,asynchronous vibration reconstruction based on the multiple signal classification（MUSIC）method,and synchronous vibration parameter identification,etc.It is realized for automatic locating of blade synchronous resonance and asynchronous vibration.Meanwhile,higher accuracy vibration parameters identification is obtained.The main contents are as follows:（1）The basic principles of the BTT measurement and the characteristics of BTT signals were analyzed.Aiming at the requirements of blade vibration parameter identification based on BTT signals,simulation analysis of the anti-noise performance of the blade vibration event locating was carried out,a qualitative analysis of the computational complexity of the MUSIC method due to the scale change before and after the expansion of the snapshot matrix was carried out,the monitoring error of synchronous vibration with equal interval sampling was calculated,and an efficient identification scheme of blade vibration parameters based on BTT signals was proposed.（2）The trend item results in the blade vibration waveform deviating from the zero horizontal line,which seriously affects the accurate locating of blade vibration events and the vibration parameter identification.Aiming at the problem of overfitting trends with fixed windows to lost vibration information,an adaptive adjustment method for the width of the fitting window of the trend was proposed.The threshold of the acceleration is used as the criterion for deciding the width of the fitting window,and the Pearson correlation coefficient of the two columns of data,which before and after removing the trend,is employed as the evaluation index for the trend fitting.In process of trend fitting with the adjustable width of the fitting window,the acceleration threshold is continuously adjusted until the Pearson correlation coefficient is greater than 0.8.Experimental verification was carried out based on the blade vibration data of the aero-engine compressor.The results showed that the adaptive fitting method can ensure that the Pearson correlation coefficient is above 0.8.Compared with the traditional fixed window trend fitting,more original vibration information of the blade can be retained.It is conducive to the accurate locating of blade vibration events and the identification of vibration parameters.（3）Aiming at the problems of weak anti-noise performance and poor locating accuracy in the traditional correlation coefficient method to locate blade vibration events,a blade vibration event locating method based on the fitted ellipse area was proposed.Two BTT probes are used to monitor the blade vibration displacement for ellipse fitting,and the blade vibration event locating is completed based on the fitted ellipse area.Compared with the correlation coefficient method,which can only contain the data of two adjacent revolutions,it can fuse more process information of blade vibration,and the accuracy of locating is significantly improved.Take the number of revolutions corresponding to the blade vibration as the reference position,through simulation analysis,the correlation coefficient method can only locate the asynchronous vibration within 100 revolutions from the reference position on the condition signal-to-noise（SNR）higher than 30d B.While the ellipse-area method is in the range of SNR 10d B～30d B,whether synchronous resonance or asynchronous vibration,the distance between the locating position and the reference position is kept within 10 revolutions,and the locating performance is significantly better than the correlation coefficient method.（4）A method for spectrum estimation of under-sampled blade tip timing signal based on MUSIC method was proposed.When the probe layout satisfies the reconstruction conditions,the MUSIC method can realize the frequency identification of the under-sampling without modifying and expanding the scale of the snapshot matrix.The current MUSIC method reconstructs BTT signals based on the snapshot matrix with expanding,which brings computational complexity increasing problem.Under the condition of analyzing 5 probes to monitor 600 revolutions of data,the calculation elapsed time of the process of frequency identification was reduced to1/60 of the original,and the computational efficiency is significantly improved.In addition,based on the Fourier transform and the remainder theorem,a blade amplitude estimation method was proposed,and the corresponding relationship between the amplitude and frequency under the under-sampling condition is obtained.It can be used as a supplementary method for the MUSIC to be unable to extract the amplitude information.Using Monte Carlo experiments for simulation analysis,under the condition of satisfying reconstruction conditions,the vibration parameter identification of the BTT signals with SNR 30d B,the maximum root mean square error（RMSE）of engine order identification is not higher than 1×10-3,the maximum relative error of amplitude identification is less than 6.26%,which meets the requirements of actual engineering applications.（5）An improved circumferential Fourier fitting（CFF）method was proposed to solve the problem of parameter identification in the non-equal interval sampling of synchronous vibration.Compared with the traditional CFF method,it is based on the non-equal interval sampling mathematical model,which conforms to the actual sampling process of the tip timing signal.The Monte Carlo experiments were used for simulation analysis.The results showed that under the same noise pollution level,the traditional and improved CFF methods have equivalent performance in engine order identification,but in terms of amplitude identification,compared with the traditional method,the relative error of amplitude fitting of improved CFF is reduced about 10%.The accuracy of amplitude identification is further improved,which is of great significance for the high-precision monitoring of the amplitude of synchronous vibration.（6）A small high-speed rotating test rig was built,and an experimental scheme for blade vibration parameter identification based on BTT was designed.The experimental verifications were carried out for the trend fitting method,blade vibration event locating method,improved CFF method,and reconstruction conditions of MUSIC method proposed in this paper.At the same time,the above-mentioned methods were used to analyze and process the blade vibration of the aero-engine compressor,and the engineering applicability test verification of the above-mentioned methods was further completed. |
PDF Full Text Request