Search for: All records

The direct selective laser sintering (SLS) process was successfully demonstrated for additive manufacturing of high-entropy carbide ceramics (HECC), in which a Yb fiber laser was employed for ultrafast (in seconds) reactive sintering of HECC specimens from a powder mixture of constitute monocarbides. A single-phase non-equiatomic HECC was successfully formed in the 4-HECC specimen with a uniform distribution of Zr, Nb, Hf, Ta, and C. In contrast, a three-layer microstructure was formed in the 5-HECC specimen with five metal elements (Zr, Nb, Hf, Ta and Ti), consisting of a TiC-rich top layer, a Zr–Hf–C enriched intermediate layer, and a non-equiatomic Zr–Ta–Nb–Hf–C HECC layer. Vickers hardness of 4- and 5-HECC specimens were 22.2 and 21.8 GPa, respectively, on the surface. These findings have important implications on the fundamental mechanisms governing interactions between laser and monocarbide powders to form a solid solution of HECCs during SLS.

Graphical abstract

Neural architecture search (NAS) and network pruning are widely studied efficient AI techniques, but not yet perfect.NAS performs exhaustive candidate architecture search, incurring tremendous search cost.Though (structured) pruning can simply shrink model dimension, it remains unclear how to decide the per-layer sparsity automatically and optimally.In this work, we revisit the problem of layer-width optimization and propose Pruning-as-Search (PaS), an end-to-end channel pruning method to search out desired sub-network automatically and efficiently.Specifically, we add a depth-wise binary convolution to learn pruning policies directly through gradient descent.By combining the structural reparameterization and PaS, we successfully searched out a new family of VGG-like and lightweight networks, which enable the flexibility of arbitrary width with respect to each layer instead of each stage.Experimental results show that our proposed architecture outperforms prior arts by around 1.0% top-1 accuracy under similar inference speed on ImageNet-1000 classification task.Furthermore, we demonstrate the effectiveness of our width search on complex tasks including instance segmentation and image translation.Code and models are released.

Stable potassium (K) isotopes (41K/39K) have shown great promise as novel chemical tracers for a wide range of bio-, geo-, and cosmo-chemical processes, but high precision stable K isotope analysis remains a challenge for plasma source mass spectrometry due to intense argon-related interferences produced directly from argon plasma. Here we provide an assessment on the analytical figures of merit of a new generation collision-cell equipped multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS), Sapphire from Nu Instruments, for K isotope analysis based on our extensive tests over a duration of ~8 months. Because use of helium and hydrogen as collision/reaction gases can reduce argon-related interferences to negligible levels at optimal flow rates, the collision-cell mode can operate at low mass resolution during K isotope analysis, providing >2 orders of magnitude higher K sensitivity (>1000 V per μg mL-1 K), as compared to the widely used “cold plasma” method, and the capability for direct 40K measurement. One challenge of the collision/reaction cell analysis on Sapphire is its higher susceptibility to matrix effects, requiring effective sample purification prior to analysis. Also, the collision-cell mode on Sapphire shows a pronounced effect associated with concentration (or ion intensity) mismatch between the sample and the bracketingmore »standard during analysis, and this effect may not be fully eliminated through conventional concentration matching practice. Instead, we developed a correction method for this concentration/ion intensity mismatch effect. Our method reduces the burden to the operator and increases sample throughput. This method allows for accurate K isotope analysis with an intermediate precision of ≤0.05 ‰ (2SD) to be routinely achieved using the collision cell on Sapphire, representing a major advance to stable K isotope analysis.« less

Reachability analysis is a fundamental problem in verification that checks for a given model and set of initial states if the system will reach a given set of unsafe states. Its importance lies in the ability to exhaustively explore the behaviors of a model over a finite or infinite time horizon. The problem of reachability analysis for Cyber-Physical Systems (CPS) is especially challenging because it involves reasoning about the continuous states of the system as well as its switching behavior. Each of these two aspects can by itself cause the reachability analysis problem to be undecidable. In this paper, we survey recent progress in this field beginning with the success of hybrid systems with affine dynamics. We then examine the current state-of-the-art for CPS with nonlinear dynamics and those driven by ``learning-enabled'' components such as neural networks. We conclude with an examination of some promising directions and open challenges.