| Abstract [eng] |
This project investigates probability of defect detection in the adhesive layer and identifies and investigates the parameters that affect the probability of detecting a defect. The CIVA 2021 software is used to create a model and study a bonded joint made from 2024 aluminium alloy plates. These two 1.8 mm thick plates are connected by an adhesive layer with a thickness of 0.16 mm. Importantly, due to the accuracy and speed of the software calculations, only the bonded joint, which is 25 mm wide and 280 mm long, is modelled. In the bonded joint, 9 square defects of three different sizes are modelled, with side lengths of: 5 mm, 10 mm, 15 mm. One defect of each size is positioned at different depths: 1,8 mm, 1,88 mm, 1,96 mm. The study is performed using a non-contact ultrasonic method. This means that the sample and transducer are immersed in a water tank. A computer-based comparison study of focused and unfocused transducers with a frequency of 15 MHz and diameter of 10 mm is performed, during which it is determined that the use of a focused transducer ensures that the signal will be reflected from the defect, and not from neighbour surfaces. The results of create computer model are verified by the experimental test. The experimental test is performed using a focused Olympus V328-SU-F transducer. A computer metamodel has been created, which is required for calculations, and consists of 1000 different combinations. Next, a computer-based sensitivity study of the focused transducer is performed, when various parameters such as the thickness of the sample layers, the size of the defect, and its depth position are changed, and it is estimated that the defect size and the position of the transducer have the greatest influence on the defect detectability. Sensitivity study is performed using Sobol indices and Kriging interpolation methods. It was found that both tests are most sensitive to the defect size and the position of the transducer relative to the x-axis of the sample. However, it has been observed that using an unfocused signal increases the sensitivity of the study to the angle of the transducer placement. An analysis of the probability of defect detection is also performed using the Monte Carlo method. The probability of detecting a defect with a 90% probability of 95% confidence limit using different types of transducers is analysed. It was observed that when the study performed with an unfocused transducer, a 90% probability of detecting a defect is not achieved. When using a focused transducer, this probability is achieved when the defect is not smaller than 3,57 mm, and with a 95% confidence limit defects are detected that are not less than 4,28 mm. |
