Massimiliano Ferrucci

4D time-space reconstruction of dynamic events or deforming objects using X-ray computed tomography (CT) is an important inverse problem in non-destructive evaluation. Conventional back-projection based reconstruction methods assume that the object remains static for the duration of several tens or hundreds of X-ray projection measurement images (reconstruction of consecutive limited-angle CT scans). However, this is an unrealistic assumption for many in-situ experiments that causes spurious artifacts and inaccurate morphological reconstructions of the object. To solve this problem, we propose to perform a 4D time-space reconstruction using a distributed implicit neural representation (DINR) network that is trained using a novel distributed stochastic training algorithm. Our DINR network learns to reconstruct the object at its output by iterative optimization of its network parameters such that the measured projection images best match the output of the CT forward measurement model. We use a forward measurement model that is a function of the DINR outputs at a sparsely sampled set of continuous valued 4D object coordinates. Unlike previous neural representation architectures that forward and back propagate through dense voxel grids that sample the object's entire time-space coordinates, we only propagate through the DINR at a small subset of object coordinates in each iteration resulting in an order-of-magnitude reduction in memory and compute for training. DINR leverages distributed computation across several compute nodes and GPUs to produce high-fidelity 4D time-space reconstructions. We use both simulated parallel-beam and experimental cone-beam X-ray CT datasets to demonstrate the superior performance of our approach.

In this work, we detail a novel application of inverse design and advanced manufacturing to rapidly develop and experimentally validate modifications to a shaped charge jet analog. The shaped charge jet analog comprises a copper liner, a high explosive (HE), and a silicone buffer. We apply a genetic algorithm to determine an optimal buffer design that can be placed between the liner and the HE that results in the largest possible change in jet velocity. The use of a genetic algorithm allows for discoveries of unintuitive, complex, yet optimal buffer designs. Experiments using the optimal design verified the effectiveness of the buffer and validated the machine learning approach to hydrodynamic design optimization.

We present a concept for propagating uncertainty in x-ray computed tomography (CT) using a Monte Carlo reconstruction (MCR) technique, comprising repeated reconstructions with varying input parameters. The proposed technique follows the framework for model-based x-ray CT uncertainty assessment per the Monte Carlo method (JCGM 101), although it provides several advantages over the conventional implementation, which relies on simulating all individual steps in the x-ray CT measurement procedure and is therefore considered to be impractical due to its high computational demand. The proposed method requires only a single set of simulated projections. For each Monte Carlo trial, the instrument geometrical parameters in a filtered back projection reconstruction algorithm are randomly sampled from specified uncertainty distributions. The output is a four-dimensional volumetric model where each voxel, defined by its three-dimensional indices, is represented by a distribution of reconstructed gray values. We reduce the four-dimensional volumetric model to three single-gray-value voxel models by calculating descriptive statistics: a voxel-wise lower gray value confidence limit, a central gray value, and an upper gray value confidence limit. Bi-directional length measurements performed on the surfaces determined from each single-gray-value model provide insight into the effect of uncertainty in the instrument geometry. The proposed approach requires significantly fewer computations and data storage per Monte Carlo trial and provides a straightforward way to relate uncertainties in reconstructed gray values to uncertainties in subsequent dimensional measurements. This, in turn, facilitates the practical application of the Monte Carlo method in x-ray CT. We implement MCR to determine uncertainty distributions in the simulated x-ray CT measurement of a simple cube and an impeller due to uncertainties in the instrument geometry. Possible extension of MCR to other sources of uncertainty in the x-ray CT measurement process is discussed.

Long synonymous with medical imaging, x-ray computed tomography (CT) is proving its worth in industrial applications such as material characterization, defect analysis, and dimensional inspection e.g. of manufactured components. Since 2005, when the first purpose-built industrial x-ray CT instrument was made commercially available, there have been significant efforts to understand just how reliable x-ray CT dimensional measurements are. In metrology, the ultimate measure of confidence is metrological traceability—a clearly-defined but often not well-understood property of a measurement result. Metrological traceability places measurements on a comparable scale and provides a quantitative indication of measurement quality in the form of an uncertainty statement. From a practical perspective, achieving traceability can be divided into two tasks: (a) ensuring traceability of the scale with which the measurements are made and (b) assessing uncertainty in the resulting quantity assigned to the measurand. Based on this breakdown of the traceability problem, we provide a high-level discussion about traceability of x-ray CT dimensional measurements, review relevant work in understanding x-ray CT measurement uncertainty, and propose a framework for model-based uncertainty assessment per the Monte Carlo Simulation method and instrument scale calibration as a unified approach to traceability.

The recent emergence of advanced manufacturing techniques such as additive manufacturing and an increased demand on the integrity of components have motivated research on the application of x-ray computed tomography (CT) for dimensional quality control. While CT has shown significant empirical potential for this purpose, there is a need for metrological research to accelerate the acceptance of CT as a measuring instrument. The accuracy in CT-based measurements is vulnerable to the instrument geometrical configuration during data acquisition, namely the relative position and orientation of x-ray source, rotation stage, and detector. Consistency between the actual instrument geometry and the corresponding parameters used in the reconstruction algorithm is critical. Currently available procedures provide users with only estimates of geometrical parameters. Quantification and propagation of uncertainty in the measured geometrical parameters must be considered to provide a complete uncertainty analysis and to establish confidence intervals for CT dimensional measurements. In this paper, we propose a computationally inexpensive model to approximate the influence of errors in CT geometrical parameters on dimensional measurement results. We use surface points extracted from a computer-aided design (CAD) model to model discrepancies in the radiographic image coordinates assigned to the projected edges between an aligned system and a system with misalignments. The efficacy of the proposed method was confirmed on simulated and experimental data in the presence of various geometrical uncertainty contributors.