Hasil untuk "Fuel"

Maja Bøg Toftegaard, J. Brix, P. Jensen et al.

Chunwen Sun, R. Hui, J. Roller

Y. Bing, Hansan Liu, Lei Zhang et al.

E. Antolini

S. Mekhilef, R. Saidur, Azadeh Safari

P. Edwards, V. Kuznetsov, W. David et al.

S. Bose, Tapas Kuila, Thi Xuan Quynh Nguyen et al.

A. Demirbaş

C. Cardona, Ó. Sánchez

Manunya Phanphanich, S. Mani

S. Haile

D. Bernardi, M. Verbrugge

B. Smitha, S. Sridhar, A. Khan

K. Yan, Guosheng Wu, T. Lafleur et al.

Deuk-Ju Kim, M. Jo, S. Nam

Muhammad Imran Khan, T. Yasmin, A. Shakoor

Yuedong Yang, Xiwen Wei, Mustafa Munir et al.

Reasoning Large Multi-modality Models (LMMs) have become the de facto choice for many applications. However, these models rely on a Chain-of-Thought (CoT) process that is lengthy and unpredictable at runtime, often resulting in inefficient use of computational resources (due to memory fragmentation) and sub-optimal accuracy (due to under- and over-thinking). We observe empirically that the CoT process follows a very simple form, whose behavior is independent of the specific generated samples. This suggests that the CoT length can be estimated ahead of time based on a hidden parameter representing the amount of "fuel" available to support the reasoning process. Based on this insight, we propose Fuel Gauge, the first method which extracts this hidden signal and predicts CoT length ahead of time. We demonstrate the utility on the Fuel Gauge on two downstream tasks: predictive KV cache allocation, which addresses memory fragmentation in LMM serving systems, and CoT length modulation, which mitigates under-thinking and over-thinking. Extensive experiments on LMMs across text-only, image-text, and video-text question answering benchmarks demonstrate the effectiveness, generalizability, and practical value of our Fuel Gauge. For example, on the GPQA-Diamond benchmark, our Fuel Gauge achieves less than half the CoT length prediction error compared to the baseline; this translates into a 13.37x reduction in the memory allocation frequency.

Davannendran Chandran, Revathi Raviadaran, Taib Iskandar Mohamad et al.

This paper aims to determine the effect of black soldier fly larvae (BSFL) oil-based biodiesel (B100) towards metal corrosion and elastomer degradation. Copper (Cu) and nitrile butadiene rubber (NBR) were exposed to BSFL-B100, industrial diesel (D2) and palm oil biodiesel (P-B100) for 1200 h at 25 °C. Corrosion rate, elastomer volume, tensile and hardness change, as well as surface morphology and total acid number (TAN) were determined. Cu had highest corrosion rates in BSFL-B100 at 0.00195 mm/yr, followed by in P-B100 at 0.00163 mm/yr and D2 at 0.00096 mm/yr. NBR exposed to BSFL-B100 had highest volume change by 31.4 % followed with in P-B100 at 29.0 % and finally in D2 at 19.4 %. BSFL-B100 exhibited significantly higher TAN increases than P-B100 and D2 after exposure to both Cu (188 %, 118 % and 84 %) and NBR (233 %, 139 % and 95 %) indicating greater fuel degradation in BSFL-B100, thus adversely affecting Cu corrosion and NBR degradation. Despite key fuel properties of prepared BSFL-B100 were within the limits specified in American Society for Testing and Materials (ASTM) D6751–23 and exhibited materials degradation comparable to P-B100, it however experienced higher fuel degradation measured in terms of TAN than P-B100 under similar experimental conditions. This could be associated to higher polyunsaturated fatty acids present in BSFL-B100 than P-B100 which is susceptible to oxidation which could adversely affect materials degradation.

Gianfranco Huaccho Zavala, Thomas Gheeraert, Thomas Gheeraert et al.

This work presents the further development and application of the multi-physics coupled code Serpent/subchanflow for analyzing cores loaded with fuel assembly designs characterized by complex geometries, such as the VVR-KN fuel assembly. A high-detail steady-state analysis of one VVR-KN fuel assembly is presented and discussed. The VVR-KN is a plate-type fuel assembly, arranged coaxially with hexagonal fuel-plate tubes. Its particular geometry layout configuration challenges both their neutronic and thermal-hydraulic modeling. In this work, the versatility of Serpent’s multi-physics interface is exploited by using the unstructured mesh-based interface to update the properties of the fuel and coolant materials in a coupled neutronic/thermal-hydraulic simulation; these properties are solved and provided by the thermal-hydraulic code Subchanflow. Both neutronic and thermal-hydraulic models are developed for a single fuel assembly of 6.83 cm distance pitch and 60 cm active height, and state conditions for the simulations are defined. Typical material composition and main thermal properties for the fuel-meat (UO2-Al) and aluminum cladding (SAV-1) materials are extracted from references. This work paves the way for multi-physics analysis of research reactors with non-regular plates or subchannel geometries.

Radouan Boukharfane

The injection of transverse jets into supersonic compressible crossflows represents a fundamental configuration relevant to a spectrum of high-speed applications. The intricate interactions arising between the crossflow and the injected jet induce complex flow phenomena, including shock waves and vortical structures, the characteristics of which are significantly contingent upon the thermophysical properties of the injected fuel. While prior investigations have addressed the influence of various fuels, a knowledge gap persists concerning the behaviour of alternative and synthetic multicomponent fuels within this flow regime. The present work employs high-fidelity large-eddy simulations (LES) to examine the impact of ten distinct fuels-hydrogen, methane, ethylene, ammonia, syngas mixture, and a synthetic blend, alongside several NH3/H2/N2 mixtures-on the macroscopic flow structures and mixing attributes within a transverse sonic jet immersed in a Mach 2 crossflow. By maintaining a uniform momentum flux ratio across the investigated cases, the study aims to isolate the influence of the unique thermophysical properties of each fuel on the windward mixing layer, a region critically important for initial entrainment processes. The investigation quantifies the effects of molecular weight, heat capacity ratio, and density on the development and evolution of coherent structures through a detailed examination of instantaneous flow fields, vortex dynamics, scalar distributions, and turbulence statistics. The results are expected to provide pertinent insights into fuel-dependent mixing mechanisms in supersonic flows, thereby contributing to the advancement of more efficient and versatile propulsion systems.