Dynamic bias error correction in gamma-ray computed tomography

Abstract The dynamic bias error is a well-known effect in transmission radiometry. It appears when an object distribution, e.g. a multiphase flow, changes its constitution during the scanning process and reconstructed data, e.g. a cross-sectional image, is assumed to represent a time-average of the object distribution. In gamma-ray tomography long sampling intervals are necessary in order to obtain sufficient photon count statistics. Therefore, the measured photon count projection data is inherently time-averaged. The attenuation law gives a non-linear relation between attenuation and photon counts. Therefore, the calculation of the time-averaged attenuation from the time-averaged projection data may lead to non-negligible and systematic errors, commonly an underestimation of the real attenuation, which e.g. means an overestimation of the gas holdup in tomography images of two-phase flows. In this work the application of a recently presented dynamic bias error correction method on time-averaged gamma-ray tomography is demonstrated. As an exemplary object we scanned a mock-up of a centrifugal pump. The suitability of this method was investigated for a generic highly turbulent two-phase flow scenario with both a virtual tomography data set as well as real measured data. These investigations allowed the quantification of both the tomography image artefacts caused by the dynamic bias error and the grade of potential correction by the presented method.