Improved reliability of automated non-destructive evaluation

摘要

In recent years, Non-Destructive Evaluation (NDE) has trended toward increased automation in data acquisition. Automated scanning has the potential to greatly increase reliability of the NDE results, but it also tends to increase the volume of data that must be inspected manually by a skilled technician. This is a time-consuming task made tedious by the fact that most of the data contains no indication of a defect. There is a great need for software that can partially automate the data analysis by prioritizing regions of interest to the inspector. This thesis describes an approach to that end, laying out a framework that is general enough to fit a wide array of NDE applications. It also describes practical considerations for the specific application of ultrasound inspection of titanium turbine discs.