Differentiating and predicting HU values of 12 aluminum alloys via X-ray computed tomography

Abstract

X-ray computed tomography (XCT) scanning as a quantitative analysis method is hindered by the lack of literature-reported values for countless metals and their alloys. Given the large number of scanning parameters and the ever-increasing number of material alloys, this gap cannot be remedied simply by running an infinitely large number of scans. This study aims to develop a novel method of predicting the mean Hounsfield Unit (HU) value of a metal alloy from the knowledge of its elemental composition. The weighted average mixture model is utilized to predict the mean HU value for aluminum (atomic number 13, amu 26.98) based alloys. The model can generate predictions for the mean HU value and its standard deviation given the elemental composition and the HU values of each constitutive element. The model is validated using XCT scans of 12 aluminum alloys (one 1XXX, two 2XXX, two 5XXX, five 6XXX, and two 7XXX series) conducted at three thickness levels (1.5, 3, and 6 mm).The findings show that increasing sample thickness results in an increased ability to differentiate between alloy pairs with 12 pairs being differentiable at 1.5 mm, 39 at 3 mm, and 45 pairs at 6 mm thickness. The predictions of the model (average error of 5.01%) are tested for individually enclosed aluminum samples. The model is then applied to generate predictions (average error was 5.65%) of stacked aluminum samples to simulate cases where scanned objects are often composed of multiple materials. Even when HU values for some constituent elements were unavailable (as the case for AA6020 where 4 out of 11 elements could not be found), the model achieved a predictive accuracy of 76%.