Research On Damage Location Based On Ultrasonic A-scan Signals From Ultrasonic Testing

High-strength composites possess certain stiffness,long service life,flexural strength,etc.,and have a mass advantage under the same conditions of strength and stiffness,thus reducing the use of energy and indirectly meeting the requirements of energy conservation and emission reduction.Despite the material advantages,composite materials are prone to complex defects during fabrication and assembly due to the complexity of the process,and the continued development of these defects can lead to structural integrity failures.To prevent catastrophic failures and due to the efficiency and portability of ultrasonic inspection technology,it is often used for material quality assessment and damage monitoring localization.Ultrasonic detection is based on the piezoelectric effect to transmit and receive ultrasonic waves and determine the presence and location of defects based on the echo amplitude and time of flight of the echo signal.Based on this principle,this paper first proposes a flexible piezoelectric fiber sensor for an ultrasonic signal receiving unit,the piezoelectric fiber consists of a piezoelectric elastomer shell and an enameled wire containing a metal core,the piezoelectric elastomer shell is prepared from a mixture of PVDF and DMF on the enameled wire through an electric wetting assisted dry spinning method,as the carbon fiber composite material has good electrical conductivity,without affecting the structural properties of the material The piezoelectric fiber sensor is encapsulated in the carbon fiber composite material using epoxy resin to form the signal receiving sensor unit without affecting the structural properties of the material.The piezoelectric fiber has good piezoelectric response performance in the dynamic response performance test of low-speed impact experiments,with corresponding charge response to different energy levels of dynamic impact force,and the response pressure difference of this sensor increases with the increase of impact energy level.Secondly,based on the prepared flexible piezoelectric fiber sensor,which is used in combination with PZT to form a one-transmitter-two-receiver type sensor network,combined with ultrasonic detection A-scan signal damage monitoring and localization imaging method,the ultrasonic waves are excited with a 5-peak modulated signal at 100 k Hz,and the remaining sensing units receive the ultrasonic signals in the non-destructive and damage states,respectively,and the received signals are characterized by.Signal amplitude and Hilbert energy peak with the increase of the damage prefabricated distance and gradually reduce to small,using PZT-PZT,PZT-piezoelectric fiber two different combinations,in the simulated damage distance of 50 mm and 100 mm,the average positioning error of the two combinations are 10 mm,13.75 mm,but in the simulated damage distance of 200 mm,PZT-piezoelectric fiber The positioning error is 15 mm,which is closer to the simulated damage position than the positioning error of 33 mm for the PZT-PZT combination.Since acquiring ultrasonic A-scan signals by sensor networks is a tedious and timeconsuming process,using phased-array ultrasonic detection to acquire certain orders of magnitude of ultrasonic detection signals,a hybrid deep learning model CNN-LSTM is proposed for the identification of non-visible damage depth locations inside carbon fiber composites based on the classification and identification of A-scan signals.By obtaining artificially prefabricated A-scan signals of defects at different depth locations as the training set of the deep learning model,the laminates after low-velocity impact experiments are thus classified and recognized for the delamination damage generated inside them.Compared with the traditional LSTM model and CNN model,the hybrid model combines the feature extraction advantage of the CNN model and the time memory advantage of the LSTM model,respectively,to improve the model’s computational speed and recognition accuracy.The depth measurement of the defects inside the carbon fiber laminate by optical microscopy shows that the hybrid deep learning model CNN-LSTM can effectively reduce the depth error of the ultrasonic detection of defects,and the average depth relative error is reduced to 8%.