Advanced Deep Learning-Based Supervised Classification of Multi-Angle Snowflake Camera Images

We present improvements over our previous approach to automatic winter hydrometeor classification by means of convolutional neural networks (CNNs), using more data and improved training techniques to achieve higher accuracy on a more complicated dataset than we had previously demonstrated. As an advancement of our previous proof-of-concept study, this work demonstrates broader usefulness of deep CNNs by using a substantially larger and more diverse dataset, which we make publicly available, from many more snow events. We describe the collection, processing, and sorting of this dataset of over 25,000 high-quality multiple-angle snowflake camera (MASC) image chips split nearly evenly between five geometric classes: aggregate, columnar crystal, planar crystal, graupel, and small particle. Raw images were collected over 32 snowfall events between November 2014 and May 2016 near Greeley, Colorado and were processed with an automated cropping and normalization algorithm to yield 224x224 pixel images containing possible hydrometeors. From the bulk set of over 8,400,000 extracted images, a smaller dataset of 14,793 images was sorted by image quality and recognizability (Q&R) using manual inspection. A presorting network trained on the Q&R dataset was applied to all 8,400,000+ images to automatically collect a subset of 283,351 good snowflake images. Roughly 5,000 representative examples were then collected from this subset manually for each of the five geometric classes. With a higher emphasis on in-class variety than our previous work, the final dataset yields trained networks that better capture the imperfect cases and diverse forms that occur within the broad categories studied to achieve an accuracy of 96.2% on a vastly more challenging dataset.

Optimization of HPGe detector response using fast and reliable method

Safeguards system with high effectiveness and efficiency must comprise a set of measurements with capabilities satisfactory for the verification of nuclear materials. In this paper, we present key parameter measurements of detector modeling in a commercial n-type low energy germanium detector of a planar crystal with a relative efficiency of nearly 15 %. The detector optimization will hold a significant function in measuring nuclear materials for safeguards application. Standard nuclear materials with diverse enrichment (depleted and low enriched) of uranium and point-like sources (137Cs, 60Co) and mixed radioactive source for Eu isotopes (152Eu, 154Eu, and 155Eu) were benefited to explore the energy resolution and detector efficiency. The energy resolution is measured over a wide range of rise time and flattop. In addition to the experimental work, the Monte Carlo simulation code is used for modeling the setup configuration to obtain the absolute efficiency at different energies. A fast and reliable method was applied in detector efficiency measurements. The data are discussed and interpreted.

A Probabilistic Radar Forward Model for Branched Planar Ice Crystals

Polarimetric radar measurements provide information about ice particle growth and offer the potential to evaluate and better constrain ice microphysical models. To achieve these goals, one must map the ice particle physical properties (e.g., those predicted by a microphysical model) to electromagnetic scattering properties using a radar forward model. Simplified methods of calculating these scattering properties using homogeneous, reduced-density spheroids produce large errors in the polarimetric radar measurements, particularly for low-aspect-ratio branched planar crystals. To overcome these errors, an empirical method is introduced to more faithfully represent branched planar crystal scattering using scattering calculations for a large number of detailed shapes. Additionally, estimates of the uncertainty in the scattering properties, owing to ambiguity in the crystal shape given a set of bulk physical properties, are also incorporated in the forward model. To demonstrate the utility of the forward model developed herein, the radar variables are simulated from microphysical model output for an Arctic cloud case. The simulated radar variables from the empirical forward model are more consistent with the observations compared to those from the homogeneous, reduced-density-spheroid model, and have relatively low uncertainty.

The Statistical Characteristics of Static Friction

Friction is one of the fundamental issues in physics, mechanics and material science with many practical applications. However, the understanding of macroscopic friction phenomena from microscopic aspect is still on the way. In this paper, molecular dynamics simulations are performed to investigate the static friction between two planar crystal surfaces. The friction force experienced by each atom is tracked and the statistical characteristics of atomic friction force are illuminated. More importantly, the influences of normal load and temperature on the statistical features are generalized. This study provides a new insight on the micro-states of friction.

Incorporating electrostatic effects into the effective stress relation — Insights from chalk experiments

Which forces are responsible for holding highly porous chalks together? We use the effective stress to quantify the electrostatic effects around particle contacts originating from the adsorption of ions onto charged mineral surfaces. The induration of chalk indicates that it is held together by contact cement, where planar crystal contacts allow the action of short-ranged adhesive Van der Waals forces. At particle distances exceeding a few nanometers, recent studies have indicated electrostatic repulsion between water-embedded adjacent particles. The magnitude of the repelling force depends, among other parameters, upon temperature and brine composition. Our premise is that by perturbing the electrostatic forces at the particle level, we can control the mechanical behavior of chalk samples tested in triaxial cells. We report the results of an experimental series, investigating how the mechanical strength and stiffness varied among samples saturated with four different brines, tested at two temperatures, and tested directly or after aging for three weeks at high temperature. We associate stiffness with bulk modulus and strength with the stress at yield. Systematic softening and weakening is observed, especially when the pore fluid is sulfate bearing, as well as for some high-temperature experiments and for aged samples. However, softening and weakening are not totally correlated, and neither brine composition, temperature, nor aging can alone dictate the mechanical behavior of the chalk — a combination is required to predict the chalk stiffness and strength. To obtain a coherent description of our experimental results, we estimated the electrostatic stress arising from ion adsorption and found it unnecessary for these experiments to postulate significant dissolution or precipitation-related changes to the rock frame.

Examining Polarimetric Radar Observations of Bulk Microphysical Structures and Their Relation to Vortex Kinematics in Hurricane Arthur (2014)

Dual-polarization radar observations were taken of Hurricane Arthur prior to and during landfall, providing needed insight into the microphysics of tropical cyclone precipitation. A total of 30 h of data were composited and analyzed by annuli capturing storm features (eyewall, inner rainbands, and outer rainbands) and by azimuth relative to the deep-layer environmental wind shear vector. Polarimetric radar variables displayed distinct signatures indicating a transition from convective to stratiform precipitation in the downshear-right to downshear-left quadrants, which is an organization consistent with the expected kinematic asymmetry of a sheared tropical cyclone. In the downshear-right quadrant, vertical profiles of differential reflectivity ZDR and copolar correlation coefficient ρHV were more vertically stretched within and above the melting layer at all annuli, which is attributed to convective processes. An analysis of specific differential phase KDP indicated that nonspherical ice particles had an increased presence in two layers: just above the melting level and near 8-km altitude. Here, convective updrafts generated ice particles in the lower layer, which were likely columnar crystals, and increased the available moisture in the upper layer, leading to increased planar crystal growth. A sharp transition in hydrometeor population occurred downwind in the downshear-left quadrant where ZDR and ρHV profiles were more peaked within the melting layer. Above the melting layer, these signatures indicated reduced ice column counts and shape diversity owing to aggregation in a predominantly stratiform regime. The rainband quadrants exhibited a sharper transition compared to the eyewall quadrants owing to weaker winds and longer distances that decreased azimuthal mixing of ice hydrometeors.

Effect of Ni Addition on In Situ WC-Cr3C2 Cermet Coating by Laser Controlled Reactive Synthesisuse

WC-Cr3C2 cermet coating on carbon steel was fabricated by laser controlled reactive synthesis and exhibited the metallurgical bonding at the interface between coating and substrate. Experimental results show that addition of Ni results in a significant change on phases and microstructure of coating. Ni addition remarkable improves resistance to thermal shock of coating. This is accounted for by increasing planar crystal zone of the coating due to Ni addition, which decreases stress concentration at the interface.