Virtual flaws for NDE training and qualification

摘要

A typical qualification mock-up contains limited number of realistic flaws due to time and cost limitations. Increasing automated inspections offer new possibilities to overcome these limitations. In automated inspections, the data gathering and analysis phases are separated. This allows the introduction of flaw signal into the data, that were not present in the original data. On the other hand, the representativeness of these "virtual flaws" need to be confirmed. This paper introduces newly developed technique that allows introduction of realistic virtual flaws to data-sets of automated ultrasonic inspection data. The sample is first scanned before flaw introduction to establish baseline signal. Then, flaws are manufacture to the sample, and it is re-scanned. The difference in the datasets reveals the crack signal, which can be extracted. After the crack signal is separated, the it can be reapplied to an another location or locations. The new virtual crack data differs from the actual crack data, as the separated crack signal is superimposed to the (possibly different) background noise. At the same time, the signal is still realistic, since it is acquired from realistic crack and transferred unaltered. The data manipulation is undetectable from the manipulated.