Detection And Classification Of Impact-Induced Delamination In Fiberglass Pre-Impregnated Laminated Composites From Ultrasonic A-Scan Signal Using Artificial Intelligence

• Main Author: Mahmod, Muhammad Faisal
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: T Technology; TJ1-1570 Mechanical engineering and machinery
• Online Access: http://eprints.usm.my/46418/1/Detection%20And%20Classification%20Of%20Impact-Induced%20Delamination%20In%20Fiberglass%20Pre-Impregnated%20Laminated%20Composites%20From%20Ultrasonic%20A-Scan%20Signal%20Using%20Artificial%20Intelligence.pdf; http://eprints.usm.my/46418/

Impact-induced delamination (IID) in fiberglass pre-impregnated laminated composites (FGLC) is an important failure mode. Besides affected the material strength and structural reliability, this failure mode normally present minor damage on the surface but the internal damage may extensive. Existing detection method using static and dynamic load response have limitations that are considered static based monitoring and require the sensor to be attached to the test specimen surface. This technique is not suitable as the damage caused by the impact normally occurred by accident at random location. Thus, detection and classification of IID using artificial neural network from ultrasonic signal has great potential to be applied, but no attempt has been made to detect and classify this failure mode in FGLC material. The classification of delamination against impact not only applicable as prediction tool to characterise the delamination, it also can be used as reference during inspecting the FGLC under specific conditions. In this study, the potential of using ultrasonic immersion testing for detecting the IID in FGLC type 7781 E-Glass fabric is studied. Several findings and development have been achieved in this study such as the relationship between delamination area and the increasing of an impact energy, where the rate is between 23 to 45 percent. Besides, it was found that the diameter of the impact damage is directly increase with the increasing of the impact energy in the range of 21 until 46 percent while for the impact damage area is between 24 until 42 percent. In addition, the dynamic segmentation algorithm has been successfully developed in this study to automatically segment the A-scan signal with regardless the variation of gap distance between transducer and specimen surface. Based on the ultrasonic inspection result, it was found that the delamination is extend internally up to 35.90 percent and the average percentage different of the measurement result which is taken from DT and NDT is just 4.72 percent and acceptable. Since the achieved classification result is highly accurate, which is exceeded 99.29 percent, it can be concluded that the selected features for the classification input is successful and the use of artificial neural network from ultrasonic A-scan signal has shown its applicability to classify the different type of the impact-induced delamination in FGLC plates.