Machine learning for detection of 3D features using sparse x-ray tomographic reconstruction.

In many inertial confinement fusion (ICF) experiments, the neutron yield and other parameters cannot be completely accounted for with one and two dimensional models. This discrepancy suggests that there are three dimensional effects that may be significant. Sources of these effects include defects in the shells and defects in shell interfaces, the fill tube of the capsule, and the joint feature in double shell targets. Due to their ability to penetrate materials, x rays are used to capture the internal structure of objects. Methods such as computational tomography use x-ray radiographs from hundreds of projections, in order to reconstruct a three dimensional model of the object. In experimental environments, such as the National Ignition Facility and Omega-60, the availability of these views is scarce, and in many cases only consists of a single line of sight. Mathematical reconstruction of a 3D object from sparse views is an ill-posed inverse problem. These types of problems are typically solved by utilizing prior information. Neural networks have been used for the task of 3D reconstruction as they are capable of encoding and leveraging this prior information. We utilize half a dozen, different convolutional neural networks to produce different 3D representations of ICF implosions from the experimental data. Deep supervision is utilized to train a neural network to produce high-resolution reconstructions. These representations are used to track 3D features of the capsules, such as the ablator, inner shell, and the joint between shell hemispheres. Machine learning, supplemented by different priors, is a promising method for 3D reconstructions in ICF and x-ray radiography, in general.

Neural network for 3D inertial confinement fusion shell reconstruction from single radiographs.

In inertial confinement fusion (ICF), x-ray radiography is a critical diagnostic for measuring implosion dynamics, which contain rich three-dimensional (3D) information. Traditional methods for reconstructing 3D volumes from 2D radiographs, such as filtered backprojection, require radiographs from at least two different angles or lines of sight (LOS). In ICF experiments, the space for diagnostics is limited, and cameras that can operate on fast timescales are expensive to implement, limiting the number of projections that can be acquired. To improve the imaging quality as a result of this limitation, convolutional neural networks (CNNs) have recently been shown to be capable of producing 3D models from visible light images or medical x-ray images rendered by volumetric computed tomography. We propose a CNN to reconstruct 3D ICF spherical shells from single radiographs. We also examine the sensitivity of the 3D reconstruction to different illumination models using preprocessing techniques such as pseudo-flatfielding. To resolve the issue of the lack of 3D supervision, we show that training the CNN utilizing synthetic radiographs produced by known simulation methods allows for reconstruction of experimental data as long as the experimental data are similar to the synthetic data. We also show that the CNN allows for 3D reconstruction of shells that possess low mode asymmetries. Further comparisons of the 3D reconstructions with direct multiple LOS measurements are justified.

Unsupervised learning about 4D features of microparticle motion.

Material clusters of different sizes are known to exist in high-temperature plasmas due to plasma-wall interactions. The facts that these clusters, ranging from sub-microns to above mm in size, can move from one location to another quickly and that there are a lot of them make high-speed imaging and tracking one of the best, effective, and sometimes only diagnostic. An unsupervised machine learning technique based on deconvolutional neural networks is developed to analyze two-camera videos of high-temperature microparticles generated from exploding wires. The neural network utilizes a locally competitive algorithm to infer representations and optimize a dictionary composed of kernels, or basis vectors, for image analysis. Our primary goal is to use this method for feature recognition and prediction of the time-dependent three-dimensional (or "4D") microparticle motion. Features equivalent to local velocity vectors have been identified as the dictionary kernels or "building blocks" of the scene. The dictionary elements from the left and right camera views are found to be strongly correlated and satisfy the projection geometrical constraints. The results show that unsupervised machine learning techniques are promising approaches to process large sets of images for high-temperature plasmas and other scientific experiments. Machine learning techniques can be useful to handle the large amount of data and therefore aid the understanding of plasma-wall interaction.

Measurement of incandescent microparticle acceleration using stereoscopic imaging.

Microparticles ranging from sub-microns to millimeter in size are a common form of matter in magnetic fusion environment, and they are highly mobile due to their small mass. Different forces in addition to gravity can affect their motion both inside and outside the plasmas. Several recent advances open up new diagnostic possibilities to characterize the particles' motion and their forces: high-speed imaging camera technology, microparticle injection techniques developed for fusion, and image processing software. Extending our earlier work on high-temperature 4D microparticle tracking using exploding wires [Z. Wang et al. Rev. Sci. Instrum. 87, 11D601 (2016)], we report here the latest results on time-resolved microparticle acceleration measurement. New particle tracking algorithm is found to be effective in particle tracking even when there are a large number of particles close to each other. Epipolar constraint is used for track-pairing from two-camera views. The error field based on an epi-geometry model is characterized on the basis of a large set of 2D track data and 3D track reconstructions. Accelerations based on individual reconstructed 3D tracks are obtained. Force sensitivity in the order of ten gravitational acceleration has been achieved. High-speed imaging is a useful diagnostic tool for microparticle physics, computer model validation, and mass injection technology development for magnetic fusion.