| Blade tip clearance(BTC)between rotor blade end and engine casing is one of the key state parameters to influence performance and safety of aero-engine.With the development of modern engine design in the direction of high thrust-to-weight ratio,high pressure ratio,high turbine temperature and high maneuverability performance,blade tip linear velocity is getting higher,and BTC measurement environment is becoming more severe.Therefore,an urgent demand is put forward for on-line BTC measurement with wide bandwidth and high precision.A high-speed acquisition and real-time processing scheme of BTC signal based on FPGA is designed in this thesis.A noise reduction method based on wavelet threshold and a high-precision peak-to-peak extraction method based on Gaussian fitting of BTC signals are proposed.The construction of BTC measurement system and the experimental verification are completed.This thesis mainly completed the following work:1.The mathematical model of capacitive BTC signal was constructed and the relationship between signal bandwidth and blade rotate speed was analyzed.According to the functional requirements of off-line data analysis and online data monitoring of BTC,three data acquisition methods were proposed,including full data,region of interest and on-board.In order to ensure the real-time stable transmission of data packets under the condition of variable rotate speed measurement,a dynamic balanced transmission method of data stream was proposed.Based on the above methods,a high-speed acquisition and transmission system of BTC signal based on FPGA and USB3.0 was designed.2.In view of the dynamic bandwidth,low signal-to-noise ratio and non-stationary signal characteristics of BTC signal,a denoising method based on wavelet threshold was developed.Through theoretical analysis and simulation experiments,the parameters of wavelet threshold denoising model were determined as db5 wavelet basis,sqtwolog threshold,hard threshold function,etc.A 6-layer wavelet threshold denoising signal processing model based on Mallat algorithm was built.By using delay equalization technology,accurate signal reconstruction was ensured,and the real-time wavelet threshold denoising processing of BTC signal based on FPGA was realized.3.A peak-to-peak calculation method for BTC signal based on Gaussian Fitting was proposed.According to the characteristic of signal waveform approximating Gaussian curve,waveform fitting was carried out with the sampling data within the inflection point range,and high precision extraction of signal peak value was realized.Aiming at the baseline drift problem caused by the radial movement of mechanical shaft and other factors,a baseline prediction method of BTC signal based on dynamic mean value was proposed to reduce the error of valley value calculation.Combined with Caruana algorithm,the real-time and high-precision calculation of BTC signal peak-to-peak value based on FPGA was realized.4.The BTC measurement system based on capacitive sensor was built,and the verification experiment of high-speed acquisition and real-time processing of multi-channel BTC signals was completed.The experimental results showed that when the blade rotate speed varied from 1000 rpm to 4000 rpm,the maximum BTC measurement error based on the wavelet threshold denoising method was 21μm,which was 38% and 29% lower than that of the FIR filtering method and the fixed point moving average filtering method,respectively.The maximum BTC measurement error based on Gauss fitting method was 14μm,which was 49% lower than that of the direct peak searching method. |
PDF Full Text Request