Research On Planar Array Electrostatic Sensing-Based Monitoring Technology For Exhaust Abnormal Particles Of Gas Turbines

Gas turbines have outstanding advantages such as high power density and smooth operation.They are widely used as the powerplants of major military and civilian equipments,and their operational safety and reliablility are of utmost importance.Therefore,as the main fault source of gas turbines,gas path components such as compressors,combustion chambers,and turbines are of great significance to achieve on-line fault detection.Compared with traditional detection technologies,an electrostatic sensing-based exhaust abnormal particle monitoring technology has a stronger capability to detect early faults,making it widely concerned.However,this technology usually relies on a single electrostatic sensor,which results in non-uniform sensitivity distribution and insufficient useable information,making it difficult to reconstruct the number and charge of the abnormal particles accurately.This is a technical difficulty that needs to be overcome urgently to further improve the performance of the technology.Accordingly,a hemisphere-shaped electrostatic sensors'planar array?HSESPA?is developed,and the compressive sensing?CS?theory is combined with the signal processing of the array,on this basis a planar array electrostatic sensing-based monitoring technology for exhaust abnormal particles of gas turbines is studied in this paper.The main contributions include:?1?An HSESPA is designed and its measurement process is modeled and analized.An HSESPA is designed and its measurement principle is analyzed.On this basis,in order to describe the relationship between the abnormal particles and the array signals,by using the Green function and the method of image charges,a charge-induction model of hemispherical probes is established based on the Poisson equation of electrostatic field,and the measurement model of the HSESPA is further established.Numerical calculations based on the measurement model show that for a specified HSESPA,its sensing units will generate signal peaks when an abnormal particle reachs the monitoring section,and the peaks relates only to the position and charge of the particle.In more detail,the peaks are proportional to the charge when the particle position is fixed,and the ratio between the peaks only relates to the particle position.This content provides a theoretical basis for studying the sensitivity of the HSESPA and related array signal processing methods.?2?The sensitivity of the HSESPA is modeled and analized.To overcome the deficiencies of the commonly used static sensitivity,the dynamic sensitivity of the HSESPA as well as its sensing unit is defined and modeled based on the measurement model.Consider that the accuracy of the dynamic sensitivity model should be improved under actual boundary conditions,finite element modeling?FEM?is performed to simulate the sensing process of hemisphere-shaped probes for abnormal particles,then the theoretical and FEM methods are validated mutually.On this basis,a FEM-based calibration method is proposed for the dynamic sensitivity model,then the calibrated dynamic sensitivity of the sensing units is quantified and analyzed.It shows that the dynamic sensitivity is very non-uniform,which makes it difficult to reconstruct the particle charge accurately.This content provides necessary prior information for using the array and studying related array signal processing methods.?3?CS-based information reconstruction for the exhaust abnormal particles is studied.In order to reconstruct the number and charge of abnormal particles accurately,a CS-based information reconstruction method for sparse charged particles is proposed.It discretizes the measurement model of the HSESPA into a sparse representation form,on this basis,information reconstruction of the particles is abstracted as a mathematical problem of solving sparse solutions of underdetermined equations.Then,by using the basis pursuit algorithm,the sparse solution is convexly solved by constraining l₁-norm.Moreover,a pre-processing method based on singular value decomposition and a result calibration method based on weighted-centroid algorithm are applied to guarantee the the reconstruction quality.Simulations show that by using sufficient sensing units,the charge,number and position of sparse particles can be reconstructed effectively,so that the fault features based on the reconstructed information can more accurately monitor the abnormal particles of early faults than those based on a single signal.In addition,the effectiveness of the CS-based reconstruction method is also validated from the perspective of the array dynamic sensitivity.?4?Based on the analysis of experimental requirements,an HSESPA monitoring system for abnormal particles is developed and experimental studies are carried out.The waveforms of the array signals,the linear characteristic of the sensing units,and the dynamic sensitivity of the sensing unit are tested and analyzed.Different single and multiple abnormal particles are reconstructed by using the CS-based information reconstruction method,and the array dynamic sensitivity is analyzed based on the reconstructed imformation.The correctness of the models and the effectiveness of the methods in this paper have been validated experimentally.