CIA-CT Conference

Over the past decade, X-ray computed tomography (XCT) has been increasingly recognised as a means of performing co-ordinate and surface measurement. However, the physics that govern XCT systems are highly complex, and modelling these physics is no simple feat. Evaluation of measurement uncertainty, which requires a robust measurement model, is, therefore, equally complex and evaluation of uncertainty in a general measurement case has not yet been achieved.
In this talk, Adam Thompson will cover the state of the art in uncertainty evaluation for XCT measurement, discussing the efforts made towards the development of general methods for XCT uncertainty evaluation in co-ordinate and surface metrology, as well as covering the future steps required to achieve this goal.

X‐ray computed tomography (CT) is increasingly used for evaluating parts produced by additive manufacturing (AM) in terms of geometrical and dimensional characteristics, internal defects, and surface topography. In particular, CT enables non‐destructive measurements of both internal and difficult‐to‐access geometries, features and surfaces (including micro-scale re-entrant surface features typical in metal AM parts), overcoming the main limitations of contact and optical measuring techniques.
This presentation deals with the ongoing research at the University of Padova (Italy) aimed at developing advanced CT-based surface topography measurement procedures to improve the understanding and precision of AM processes.

Reinforcement of the polymer matrix in injection-molded plastics with fibers of various kinds can improve their mechanical properties. The resulting polymer composite material provides exceptional load-bearing capabilities with flexibility in the molding of complex-shaped components. Mechanical properties of short-fiber-reinforced polymers are highly dependent on the fiberorientation in the matrix. Lack of maturity of injection molding simulation frameworks limits prediction of the complex flow-patterns during injection molding, while the requirement for resolving individual fibers in conventional imaging methods significantly restricts the imaged field-of-view.

Here we demonstrate a new approach for full-scale fiber-reinforced composite investigations providing direct insight into the fiber orientations of full-scale injection molded parts.

Recent innovations in x-ray instruments for aviation security have led to the development of in-line fixed-gantry x-ray CT scanners. Over the last decade, new high-flux intake photon counting detectors have been developed. Implementing photon counting detectors in these fixed-gantry designs open the possibility of a new generation of aviation security scanners with improved passenger throughput and enhanced threat identification. However, this new design significantly increases the number of x-ray hardware components, such as detector modules, and the computation requirements needed to handle the energy dispersive CT data. This increase in scanner complexity poses a series of challenges from a metrological perspective. This talk will discuss the complexity of the fixed-gantry XCT design and how to handle and calibrate the scanner geometry.

CT has been a preferred reference method for determination of food quality for many years. DMRI (Danish Meat Research Institute) has been exploring the technology and applicability of the X-ray modality in the transition of meat production from manual to cobot assisted procedures. The RoBUTCHER project is exploring this transition.

Effective design procedures are available from using a CT based library of pig carcasses. We optimize the cutting trajectory of the carcass into primals by planning from CT and adaptation based on 3D vision. We also explore our CT library to synthesize image data to train AI networks for prediction of important quality features.

Major slaughterhouses, which as of today represent a very labour-intensive industry, stand on the edge of industry 4.0 and will in the future be transformed with much further automated processes by the help from advanced sensor technology and computational algorithms. In this respect, collaborative cell concepts where augmented reality tools support the operator in his/her tasks are already being developed as an alternative to the standard serial production line. The aim is to develop a new adaptive production concept where robot and operator collaborate in a cellular layout. This is a collaboration between Danish research institutions and industry with the aim of improving meat production from being conveyor belt based to cell based. Part of this effort is to use on-line computer prediction of 3D configurations of carcasses based on 2D CT scans and deformation calculations with finite element models.

In the course of the increasing use of renewable energies and the conversion of the automotive industry to e-mobility, large investments are being made in the production of energy storage systems and batteries for e-cars.

For safety aspects, as well as for reasons of sustainability in production, the inline inspection of modern Li-based batteries and storage systems in production is gaining immense importance. Naturally, a large part of the inspection takes place by means of 3D CT in order to check the interior of the battery cells for defects as early as possible in production. The high requirements, such as short cycle times, multi-material inspection, as well as small defect sizes, pose great challenges for an inspection by CT and require state-of-the-art analysis methods.

In this presentation, the state of the art and the latest developments in the field of CT inspection and analysis will be presented based on this use case.

In X-ray computed tomography (CT) we reconstruct the interior structure of an object of interest based on X-ray projections. But can we be sure the reconstruction is reliable? The reconstruction is heavily dependent on the data quality and the mathematical model that represents the physics of the measurements. Uncertainty in either will propagate into the reconstruction and could cause artifacts and negatively affect decisions to be made from the image.

In Bayesian X-ray CT we produce reconstructions with associated uncertainty quantification (UQ), which helps us understand the reliability.

We present the general framework of Bayesian X-ray CT and give UQ results for an application of defect detection in subsea pipe inspection. We further introduce two open-source software tools: CUQIpy for UQ of a broad class of inverse problems, including CT (https://sites.dtu.dk/cuqi#Software), and the Core Imaging Library (CIL) – for processing and reconstruction of many types of CT data (https://ccpi.ac.uk/cil/).

Using computed tomography (CT) from a medical CT scanner to produce detailed images of the inside of food items (pigs, fish, chicken). AI can be used to optimize the analysis of CT images of food material in a number of ways.

One way that AI can be used to optimize the use of the raw material by automating the process of image interpretation. This can involve using machine learning algorithms to analyze large amounts of CT images and identify common features that are indicative of specific processes or quality traits.

Overall, AI has the potential to greatly improve the efficiency and accuracy of digitizing the material, which can ultimately lead to a better understanding and improved the food production.