Nondestructive Evaluation (NDE) applies noninvasive techniques to determine the integrity of a material, component or structure. In latest centuries, NDE methods have been widely used in quality assurance and control, inspection personnel qualification, component characterization identification, tolerance design parameter determination etc. Probability of detection (POD) is one of the most important metrics to quantify the capability of nondestructive evaluation techniques. The current studies are based on empirical testing with well fabricated artificial defects. However, the preparation of the specimens with seeded defects is costly and time consuming, which increase the difficulties and cost of POD studies by traditional empirical experiments. And it is not able to evaluate the field inspection capability because lacking the complete knowledge of defects. Furthermore, because of the characteristics of the NDE methods, the POD is sensitive to many controllable and uncontrollable factors. It is very important to have a robust inspection system to the noises.
In this research, a novel method by establishing a composite NDE system is firstly discussed to evaluate the field inspection capability with in-service natural occurred defect samples. The simulation and verification results show the effectiveness of the methodology. Then a truncated logistic regression model is developed based on the composite NDE methods to enhance the estimation accuracy considering the undetected existing defects. The simulation results illustrated that the truncated logistic regression reached a better results of POD estimation than the standard logistic regression model. Finally, we developed an application of robust design to optimize the robustness of the probability of detection of an innovative acoustic thermographic nondestructive evaluation system. It is shows that Taguchi method provides a simple and efficient methodology for the robust design and feasible to be applied in optimizing nondestructive inspection systems. At the end of this dissertation, the work of this dissertation and the future work are discussed and concluded.
