The accurate numerical simulation of the eddycurrent testing (ECT) experiments usually requires large computational efforts. To avoid the time-consuming computations, the idea – which is new in the domain of ECT –of using databases appeared recently. The database consists of well-chosen pairs of input-output samples of a specified ECT experiment. Once it is built, one just has to retrieve the sought data from the database instead of recoursing to the complicated and expensive-to-run simulation methods. However, the generation of such databases is not a straightforward task. This paper presents a new, kriging-based adaptive methodology which yields to databases optimized to the given problem.