Precision Engineering 66 (2020) 382–391 Available online 18 August 2020 0141-6359/© 2020 Elsevier Inc. All rights reserved. Review of the infuence of noise in X-ray computed tomography measurement uncertainty ´ Angela Rodríguez-S´ anchez a, * , Adam Thompson a , Lars K¨ orner a , Nick Brierley b , Richard Leach a a Manufacturing Metrology Team, Faculty of Engineering, University of Nottingham, UK b The Manufacturing Technology Centre, Antsy Park, Coventry, CV7 9JU, UK ABSTRACT Different aspects of noise in X-ray computed tomography (XCT) for industrial purposes are examined. An overview of the most common noise metrics is given, together with a description of XCT noise infuence quantities. We address the current state of the art in understanding the contribution of noise to XCT measurement uncertainty, giving a chronological view of the different attempts that have been made to account for the contribution from noise to XCT measurement uncertainty. We conclude that approaches to estimating the contribution of noise to XCT measurement uncertainty that account for not only noise, but also other factors that affect image quality (e.g., scattering, beam hardening and blurring) are preferable to approaches that only account for noise. 1. Introduction X-ray computed tomography (XCT) is an image acquisition technique based on the interaction of radiation and matter [1]. XCT uses the in- formation provided by the attenuation of X-rays through an object to reconstruct a two- or three-dimensional (2D or 3D, depending on the specifc XCT system) representation of the object being measured [2–4]. XCT was introduced for medical imaging purposes in the early 1970s [5] and, although some attempts to apply this technology for non-destructive-testing [6] and dimensional measurements [7,8] were performed at the end of the last century, XCT has only been considered a viable tool for these purposes in the last ten years [2,9]. The introduction of XCT into the feld of metrology is justifed by the advantages it brings compared to other types of coordinate measuring systems (CMSs), such as tactile or optical CMSs. XCT offers a high measurement point density and is the only CMS capable of non-destructively determining the external and internal geometries of an object, as well as distinguishing between different materials. As such, XCT is now considered a multi-purpose non-destructive testing and measurement technique that allows for simultaneous material testing and dimensional quality control. One of the main issues with using XCT for dimensional measure- ments in industry and research is the evaluation of measurement un- certainty [2,10–12]. The diffculties of XCT measurement uncertainty evaluation arise from its plethora of error sources, relating not only to the system itself, but to the operator settings, the object, the data pro- cessing and the environmental conditions during the scan [13]. To apply XCT in dimensional quality control scenarios [14], with applications in aerospace [15], automotive [16] and medical [17] industries, trace- ability of measurements to the SI unit of length (i.e. the metre) is required [18]; thus, evaluation of the uncertainty of XCT measurements is required. Although many attempts have been made to fnd a method to evaluate XCT measurement uncertainty [19–30], a defnitive, stand- ardised method, applicable in measurement laboratories and industrial environments, has not yet been established. The closest a document has achieved so far is VDI/VDE 2630–2.1 [25] – while many consider this document to be a standard, the VDI/VDE documents are intended to act as guidelines, as opposed to explicit standards. The uncertainty evalu- ation methods used most widely in the literature are the analytical approach stated in the Guide to the Expression of Uncertainty in Mea- surement (GUM) [31], the GUM simulation approach using Monte Carlo methods [32] and the so-called substitution method. The substitution method is used to estimate the bias contribution of XCT (or indeed any CMS) measurement uncertainty using reference objects that are geometrically and materially similar to the object being measured and have previously been calibrated e.g. by a tactile CMS [22]. Calibration uncertainty is evaluated, and then combined with other infuence quantities (e.g. repeatability, thermal effects) to provide the measure- ment uncertainty. Generic details of the substitution method are addressed in ISO 15530-3 [33] and guidance for its implementation to XCT measurements is given in VDI/VDE 2630–2.1 [25]. However, the substitution method is not generalizable to dissimilar workpieces due to the stringent similarity conditions with respect to the calibrated reference. * Corresponding author. E-mail address: angela.rodriguezsanchez@nottingham.ac.uk ( ´ A. Rodríguez-S´ anchez). Contents lists available at ScienceDirect Precision Engineering journal homepage: http://www.elsevier.com/locate/precision https://doi.org/10.1016/j.precisioneng.2020.08.004 Received 22 July 2020; Received in revised form 4 August 2020; Accepted 7 August 2020