Multi-sensor data fusion for high-resolution material characterization

In typical nondestructive evaluation (NDE) of materials, the material under test is inspected using one or more NDE techniques to evaluate its condition. However, measurement data from different inspection techniques are often complementary in nature and higher accuracy may be achieved by fusing information from these different inspection modes. This paper proposes a classifier-fusion based approach to combine multifrequency eddy current and ultrasound data for material characterization. The proposed algorithm uses a hierarchy of classifiers to determine the material state (e.g. stress, heat treatment etc.) and level of exposure to this condition, with classifier fusion achieved through a majority-voting rule. Preliminary results on applying the proposed algorithm to data from Inconel 600 samples are presented.