10784 research outputs found
Sort by
VALIDITY OF TOUGHNESS MEASUREMENTS FROM MINIATURE SPECIMENS FAILING IN DIFFERENT FRACTURE MODES
No full textPublisher Copyright: Copyright © 2024 by a non-US government agency.Using 0.16T (4 mm thick) compact tension specimens to determine toughness in reactor pressure vessel steels permits the ductile to brittle transition temperature to be derived from small amounts of material and allows more effective use of surveillance specimens. Achieving validity in the toughness measurements requires more care in specimen manufacture, load train alignment and test temperature selection than when larger specimens are used. This is particularly evident in the shapes of precrack fronts, which appear more sharply asymmetric on smaller specimens. Questions have also been raised as to whether the failure mechanisms are the same in miniature and large specimens. This work, performed within the FRACTESUS project, presents toughness measurements and detailed fractography on both a homogeneously brittle base metal and a relatively ductile, inhomogeneous weld. The fractography shows that, even in a ductile weld metal exhibiting clear shear lip, brittle fracture initiates within the part of specimen experiencing small scale yielding, so long as the toughness measurement is valid. Similarly, although the precrack front asymmetry appears more marked in smaller specimens, as long as the deviation from planarity is within the ASTM limits, the asymmetry does not affect the location of the initiation site. For materials showing a variety of fracture modes, the fracture modes observed at the initiation sites are consistent with those observed in larger specimens. Where data are available, the stress and strain conditions at the initiation sites are also found to be consistent in 0.16T and larger specimens. These fractographic observations support the thesis that toughness measurements made on miniature compact tension specimens reflect the same material characteristics and failure mechanisms as those on larger specimens. The limitations on specimen geometry and loading identified by the ASTM Standard are sufficient to ensure that transition temperature values derived from miniature compact specimens are valid for materials failing via a variety of fracture modes.Peer reviewe
The Network Orchestration Role of Accelerators for Value Creation
No full textPublisher Copyright: © 1988-2012 IEEE.Corporate accelerators are becoming increasingly important in introducing innovative methods for companies to generate value. Despite their growing importance, the specifics of their activities in value creation, especially in orchestrating and managing network collaborations for companies and startups, are still poorly understood. This study attempts to fill this knowledge gap by describing a set of orchestration activities that shape the synergy between companies and startups in an accelerator program and correlating these activities with value creation. Empirical evidence was collected through a qualitative case study of a corporate accelerator program. The results show that orchestration activities inherent in the accelerator's network are critical to driving value creation. We identified five such orchestration activities: 1) knowledge mobility and innovation sustainability, which leads to novelty; 2) innovation appropriability and 3) network stability for lock-in, keeping members in the network and increasing relationship quality; 4) innovation coherence as a source of complementarity; and 5) leveraging resources for efficiency. Additionally, our research shows that value for startups begins with expanding their network reach, which encourages the development of new skills, the introduction of novel concepts, and the expansion of existing knowledge. The findings from this study contribute to the theoretical discourse on the importance of network orchestration in the context of corporate accelerators. By integrating orchestration activities into the broader network narrative, we gain insights into the central connecting role of accelerators and their practical relevance.Peer reviewe
BESPOKE SOLVERS FOR GENERATIVE FLOW MODELS
No full textPublisher Copyright: © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.Diffusion or flow-based models are powerful generative paradigms that are notoriously hard to sample as samples are defined as solutions to high-dimensional Ordinary or Stochastic Differential Equations (ODEs/SDEs) which require a large Number of Function Evaluations (NFE) to approximate well. Existing methods to alleviate the costly sampling process include model distillation and designing dedicated ODE solvers. However, distillation is costly to train and sometimes can deteriorate quality, while dedicated solvers still require relatively large NFE to produce high quality samples. In this paper we introduce “Bespoke solvers”, a novel framework for constructing custom ODE solvers tailored to the ODE of a given pre-trained flow model. Our approach optimizes an order consistent and parameter-efficient solver (e.g., with 80 learnable parameters), is trained for roughly 1% of the GPU time required for training the pre-trained model, and significantly improves approximation and generation quality compared to dedicated solvers. For example, a Bespoke solver for a CIFAR10 model produces samples with Fréchet Inception Distance (FID) of 2.73 with 10 NFE, and gets to 1% of the Ground Truth (GT) FID (2.59) for this model with only 20 NFE. On the more challenging ImageNet-64×64, Bespoke samples at 2.2 FID with 10 NFE, and gets within 2% of GT FID (1.71) with 20 NFE.Peer reviewe
Assessing generalisability of deep learning-based polyp detection and segmentation methods through a computer vision challenge
No full textPublisher Copyright: © 2024, The Author(s).Polyps are well-known cancer precursors identified by colonoscopy. However, variability in their size, appearance, and location makes the detection of polyps challenging. Moreover, colonoscopy surveillance and removal of polyps are highly operator-dependent procedures and occur in a highly complex organ topology. There exists a high missed detection rate and incomplete removal of colonic polyps. To assist in clinical procedures and reduce missed rates, automated methods for detecting and segmenting polyps using machine learning have been achieved in past years. However, the major drawback in most of these methods is their ability to generalise to out-of-sample unseen datasets from different centres, populations, modalities, and acquisition systems. To test this hypothesis rigorously, we, together with expert gastroenterologists, curated a multi-centre and multi-population dataset acquired from six different colonoscopy systems and challenged the computational expert teams to develop robust automated detection and segmentation methods in a crowd-sourcing Endoscopic computer vision challenge. This work put forward rigorous generalisability tests and assesses the usability of devised deep learning methods in dynamic and actual clinical colonoscopy procedures. We analyse the results of four top performing teams for the detection task and five top performing teams for the segmentation task. Our analyses demonstrate that the top-ranking teams concentrated mainly on accuracy over the real-time performance required for clinical applicability. We further dissect the devised methods and provide an experiment-based hypothesis that reveals the need for improved generalisability to tackle diversity present in multi-centre datasets and routine clinical procedures.Peer reviewe
Do we really need dice? The hidden region-size biases of segmentation losses
No full textPublisher Copyright: © 2023 Elsevier B.V.Most segmentation losses are arguably variants of the Cross-Entropy (CE) or Dice losses. On the surface, these two categories of losses (i.e., distribution based vs. geometry based) seem unrelated, and there is no clear consensus as to which category is a better choice, with varying performances for each across different benchmarks and applications. Furthermore, it is widely argued within the medical-imaging community that Dice and CE are complementary, which has motivated the use of compound CE-Dice losses. In this work, we provide a theoretical analysis, which shows that CE and Dice share a much deeper connection than previously thought. First, we show that, from a constrained-optimization perspective, they both decompose into two components, i.e., a similar ground-truth matching term, which pushes the predicted foreground regions towards the ground-truth, and a region-size penalty term imposing different biases on the size (or proportion) of the predicted regions. Then, we provide bound relationships and an information-theoretic analysis, which uncover hidden region-size biases: Dice has an intrinsic bias towards specific extremely imbalanced solutions, whereas CE implicitly encourages the ground-truth region proportions. Our theoretical results explain the wide experimental evidence in the medical-imaging literature, whereby Dice losses bring improvements for imbalanced segmentation. It also explains why CE dominates natural-image problems with diverse class proportions, in which case Dice might have difficulty adapting to different region-size distributions. Based on our theoretical analysis, we propose a principled and simple solution, which enables to control explicitly the region-size bias. The proposed method integrates CE with explicit terms based on L1 or the KL divergence, which encourage segmenting region proportions to match target class proportions, thereby mitigating class imbalance but without losing generality. Comprehensive experiments and ablation studies over different losses and applications validate our theoretical analysis, as well as the effectiveness of explicit and simple region-size terms. The code is available at https://github.com/by-liu/SegLossBias.This work is supported by the National Science and Engineering Research Council of Canada (NSERC) via its Discovery Grant program, and the Fonds de recherche du Québec - Nature et technologies (FRQNT) through the Research Support for New Academics program. It is also supported by Prompt Quebec , and we thank Calcul Quebec and Compute Canada for providing computing resources. Adrian Galdran was funded by a Marie Sklodowska-Curie Fellowship ( No 892297 ).Peer reviewe
A Material Model Optimization Approach for the Sheet Metal Forming Process Using the Hole Expansion Test
No full textPublisher Copyright: © 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.Sheet metal forming is an important manufacturing process widely used to produce complex stamped parts from flat sheet stock in industries such as automotive and packaging. Due to the global economic climate, these industries need to be highly competitive by reducing production costs and increasing process efficiency. Numerical simulation combined with sheet metal forming expertise is one of the technological innovations adopted to meet these requirements by reducing the traditional time-consuming and costly testing steps. With the progress of finite element simulation, questions about the accuracy or limitations of the type of material description adopted have become particularly important. The influence of the plasticity model is examined in this work by a numerical study using the hole expansion test. This work first presents the yield locus criterion adopted and developed by Tata Steel, which is hereafter referred to as the Tata Steel material model. Hole expansion tests are performed at different hole diameters and the results are compared with FE simulation. The simulations are performed with the finite element software Autoform R11 in which the yield criterion proposed by Abspoel & Scholting [1] has been implemented. The discussion therefore focuses on the influence of the material model on the numerical predictions and its accuracy based on the optimization of the different material parameters measured.The authors would like to acknowledge experimental work carried out at Tata Steel R&D by Frank Schouten and Tushar Khandeparkar.Peer reviewe
Transductores Piezoeléctricos Impresos para la monitorización de placas mediante ondas ultrasónicas guiadas
No full textLas ondas ultrasónicas guiadas hacen posible la monitorización no destructiva de la salud estructural de estructuras planares, habitualmente mediante transductores piezoeléctricos PZT comerciales pegados a posteriori a la estructura. Los últimos avances en electrónica integrada y polímeros electroactivos están siendo utilizados en recientes estudios para imprimir transductores ultrasónicos directamente sobre piezas. Esto hace que sean viables topologías de sensorización personalizadas, acelera y simplifica de manera rotunda la instalación, es más compatible con grandes flexiones de la estructura y además permite una reducción de peso del sistema. En este trabajo, en curso, se están desarrollando transductores piezoeléctricos impresos sobre la propia estructura a monitorizar, con el fin último de conformar, de manera única o mixta junto con PZTs, sistemas de monitorización por ondas guiadas en estructuras planares, para hacer posibles futuras estructuras auto-sensorizadas desde la misma fabricación. En la fase inicial, como paso previo, se han desarrollado transductores impresos por serigrafía sobre films flexibles de PET, a partir de un copolímero líquido piezoeléctrico P(VDF-TrFE), así como polímeros conductores y tintas de plata nanoparticuladas para los electrodos. En una segunda fase, se han imprimido los traductores directamente sobre placas de aluminio 6082 T6 para uso aeronáutico de 1mm de espesor. Los transductores han sido concebidos para favorecer primeramente el modo guiado de Lamb A0, y se han detectado mediante ellos ondas guiadas transmitidas en placas de frecuencia principal entre 500-800 kHz, alcanzando una relación señal ruido superior a 20dB con la señal promediada a 16 muestras, que se está trabajando para mejorar. Se ha confirmado que la impresión de transductores puede protagonizar la siguiente generación de sensorización para la monitorización y evaluación no destructiva de estructuras planares críticas en el sector aeroespacial, automoción entre otros.Peer reviewe
Corrigendum to “Constructing transferable and interpretable machine learning models for black carbon concentrations” [Environ. Int. 184 (2024) 108449] (Environment International (2024) 184, (S0160412024000357), (10.1016/j.envint.2024.108449))
No full textPublisher Copyright: © 2024 The Author(s)The authors regret that important information of ‘Tareq Hussein’ in affiliation (h) was not complete in the original article. The correct affiliation (h) should be ‘Environmental and Atmospheric Research Laboratory (EARL), Department of Physics, School of Science, University of Jordan, Amman 11942, Jordan’, and it is updated as above. The authors would like to apologise for any inconvenience caused
AIROGS: Artificial Intelligence for Robust Glaucoma Screening Challenge
No full textPublisher Copyright: © 1982-2012 IEEE.The early detection of glaucoma is essential in preventing visual impairment. Artificial intelligence (AI) can be used to analyze color fundus photographs (CFPs) in a cost-effective manner, making glaucoma screening more accessible. While AI models for glaucoma screening from CFPs have shown promising results in laboratory settings, their performance decreases significantly in real-world scenarios due to the presence of out-of-distribution and low-quality images. To address this issue, we propose the Artificial Intelligence for Robust Glaucoma Screening (AIROGS) challenge. This challenge includes a large dataset of around 113,000 images from about 60,000 patients and 500 different screening centers, and encourages the development of algorithms that are robust to ungradable and unexpected input data. We evaluated solutions from 14 teams in this paper and found that the best teams performed similarly to a set of 20 expert ophthalmologists and optometrists. The highest-scoring team achieved an area under the receiver operating characteristic curve of 0.99 (95% CI: 0.98-0.99) for detecting ungradable images on-the-fly. Additionally, many of the algorithms showed robust performance when tested on three other publicly available datasets. These results demonstrate the feasibility of robust AI-enabled glaucoma screening.Peer reviewe
Extremely low coercivity of powder Co-rich amorphous alloy obtained by gas atomisation
No full textPublisher Copyright: © 2024 Elsevier B.V.A novel Co-base soft magnetic powder with ultra-low coercivity was produced by inert gas atomisation. The as-atomised powder is fully amorphous owing to its high glass forming ability and the proper selection of atomisation conditions; it exhibits a very low coercivity (0.130 Oe) mainly due to the low magnetostriction coefficient of Co-based alloys. After production, the powder was annealed at different temperatures between 300 and 600 ºC for 30 minutes. Crystallisation started at around 575 ºC, which is the temperature of the first exothermic peak detected by differential scanning calorimetry. Annealing the powder at temperatures below 550 ºC develops only structural relaxation of the amorphous structure. The smallest coercivity (0.056 Oe) was found in the sample annealed at 400 ºC. On the other hand, an increment of the coercivity of four orders of magnitude occurred after full crystallisation at 600 ºC (394.320 Oe). From the analysis of the curves of anisotropy field distribution, it can be concluded that this Co-base alloy shows lower average anisotropy field, a more gaussian shape and a wider distribution than Fe-base alloys. This is explained by a lower magnetoelastic anisotropy and a more homogeneous distribution of the magnetic anisotropy. The powder exhibits a spherical shape that makes it very suitable as ferromagnetic phase to manufacture powder cores with extremely low hysteresis loss for medium-high frequency applications.Peer reviewe