Research On Feature Extraction Technology Of CFRP Defect Detection Based On Time-Frequency Analysis

Carbon fiber composite material(CFRP)is the material basis of modern industrial production.It is widely used in various fields and has broad application prospects.During the manufacturing and use of carbon fiber composite materials,due to factors such as the preparation process and external impact,it is easy to produce internal defects,resulting in a decrease in its overall mechanical properties and strength,resulting in a reduction in the usability and reliability of the material.Therefore,to ensure the integrity and reliability of the material,non-destructive testing should be performed before and during its use.Using ultrasonic testing technology as a means,this paper conducts research on carbon fiber composite material defect detection and feature extraction technology based on time-frequency analysis.Aiming at the aliasing phenomenon of structure noise and near-surface defect echoes in ultrasonic detection of layered composites,this paper proposes a near-surface defect feature extraction technology based on time-frequency analysis.The short-time Fourier transform,the Wegener transform and the Cui-Williams transform are used to analyze the echo signal respectively,and the influence of the form and width of the window on the characteristics of the time-frequency image is analyzed.For the problem of resolution and cross-interference,an improved pseudo-smoothed Wegener transform analysis method is adopted,which solves the conflicting problem of cross-interference and time-frequency resolution.The characteristic parameters of defects are identified,and the centroid distance between the energy gathering areas is used as the characteristic parameters to identify the depth information of near-surface defects;the V-detector real-valued negative selection algorithm and real-valued negative selection algorithm in the artificial immune algorithm are applied to carbon fiber composites In defect identification,through comparative analysis,it can be seen that the V-detector real-valued negative selection algorithm can effectively identify defects,and its recognition rate has been effectively improved compared with the real-valued negative selection algorithm.The research shows that the signal processing method based on time-frequency analysis can effectively realize the identification of near-surface defects in layered composites.