Research On Detection Method Of Blade Weight Gain Defect Based On Time Frequency And Fractal Theory

With the rapid development of social economy,the application of some large scale complex mechanical equipment such as axial flow compressor in industry is increasing.Blade in axial compressor belongs to the high speed rotating parts.The gas flow in the compressor,not along the radius direction,but along the axial direction,resulting in its bad operating conditions and complex load so the risk of failure of the blade is greatly increased.If an effective way to diagnose blade defect can be researched,detecting the position and degree of defect will be found early.Then timely maintenance replacement of blade will avoid more damage.This thesis takes the Level III blade of axial compressor of an enterprise as the research object.Partial defect of blade is simulated with magnet to locate and quantify the defects.The main work is as follows:(1)The local weight gain of the blade is simulated with different mass of magnet blocks,and the vibration test bench is set up to develop the test plan and set up the relevant test parameters to collect the vibration signal before and after the defects of the blade.(2)By using short time Fourier transform(STFT)to analyze the vibration signal,by selecting the time window function to detect changes with reasonable narrow range of blade vibration signal,and the detection result is extracted to maintain the degree and frequency shift based on two parameters,The least square method was used to fit out the degree of maintenance and defect as well as the function relation between frequency shift and defect distance thereby,quantify the degree of weight gain defect of the blade and locate the defect position.(3)Db wavelet packet is used to decompose the defect vibration signal by 3 layers.The energy ratio of the defect signal in each frequency band is calculated.As a training sample,the BP neural network with different nodes and different hidden layer is trained by Levenberg-Marquarat neural network.Finally find out the best network and use the training of the network for different locations,different degree of weight gain localization and quantitative detection.(4)Because the blade vibration signal has fractal character,the dynamic characteristic of the actual system can be reflected by the fractal dimension of the vibration signal.By analyzing the defect signal calculate the box dimension then use the box dimension as the characteristic parameter of the defect signal to identify the degree and position of the defect.(5)Based on MATLAB/GUI,the detection system of the weight gain of the axial compressor blade was developed.Its main functions include:data reading,short time Fourier transform analysis,wavelet packet-neural network analysis,fractal analysis,extraction of characteristic parameters and display of analysis results.Through the research on the detection method of the blade weight gain defect,it has a certain practical significance and value of blade defect detection.