Explore the vast range of titles available.

In the present work, the three-dimensional heat and fluid flows around staggered pin-fin arrays are predicted using two hybrid RANS/LES models (an improved delayed detached eddy simulation (IDDES) model and a stress-blended eddy simulation (SBES) model), and one transitional unsteady Reynolds averaged Navier-Stokes (URANS) model, called k-ω SSTLM. The periodic segment geometry with a total of nine pins is considered with a channel height of 2D and a distance of 2.5D between each pin. The corresponding Reynolds number based on the pin diameter and the maximum velocity between pins is 10,000. The two hybrid RANS/LES results show the superior prediction of the mean velocity profiles around the pins, pressure distributions on the pin wall, and Nusselt number distributions. However, the transitional model, k-ω SSTLM, shows large discrepancies except in front of the pins where the flow is not fully developed. The vortical structures are well resolved by the two hybrid RANS/LES models. The SBES model is particularly adept at capturing the 3-D vortex structures after the pins. The effects of the blending function switching between RANS and LES mode of the two hybrid RANS/LES models are also investigated.

In the present work, uncertainty quantification of a venturi tube simulation with the cavitating flow is conducted based on Bayesian inference and point-collocation nonintrusive polynomial chaos (PC-NIPC). A Zwart–Gerber–Belamri (ZGB) cavitation model and RNG k-ε turbulence model are adopted to simulate the cavitating flow in the venturi tube using ANSYS Fluent, and the simulation results, with void fractions and velocity profiles, are validated with experimental data. A grid convergence index (GCI) based on the SLS-GCI method is investigated for the cavitation area, and the uncertainty error (UG) is estimated as 1.12 × 10−5. First, for uncertainty quantification of the venturi flow simulation, the ZGB cavitation model coefficients are calibrated with an experimental void fraction as observation data, and posterior distributions of the four model coefficients are obtained using MCMC. Second, based on the calibrated model coefficients, the forward problem with two random inputs, an inlet velocity, and wall roughness, is conducted using PC-NIPC for the surrogate model. The quantities of interest are set to the cavitation area and the profile of the velocity and void fraction. It is confirmed that the wall roughness with a Sobol index of 0.72 has a more significant effect on the uncertainty of the cavitating flow simulation than the inlet velocity of 0.52.

Herein, a novel strategy to fabricate haze films employing liquid crystal (LC) technology for photovoltaic (PV) applications is reported. We fabricated a high optical haze film composed of low-molecular LCs and polymer and applied the film to improve the energy conversion efficiency of PV module. The technique utilized to fabricate our haze film is based on spontaneous polymerization-induced phase separation between LCs and polymers. With optimized fabrication conditions, the haze film exhibited an optical haze value over 95% at 550 nm. By simply attaching our haze film onto the front surface of a silicon-based PV module, an overall average enhancement of 2.8% in power conversion efficiency was achieved in comparison with a PV module without our haze film.

The success of multimodal data fusion in deep learning appears to be attributed to the use of complementary in-formation between multiple input data. Compared to their predictive performance, relatively less attention has been devoted to the robustness of multimodal fusion models. In this paper, we investigated whether the current multimodal fusion model utilizes the complementary intelligence to defend against adversarial attacks. We applied gradient based white-box attacks such as FGSM and PGD on MFNet, which is a major multispectral (RGB, Thermal) fusion deep learning model for semantic segmentation. We verified that the multimodal fusion model optimized for better prediction is still vulnerable to adversarial attack, even if only one of the sensors is attacked. Thus, it is hard to say that existing multimodal data fusion models are fully utilizing complementary relationships between multiple modalities in terms of adversarial robustness. We believe that our observations open a new horizon for adversarial attack research on multimodal data fusion.

To combat against adversarial attacks, autoencoder structure is widely used to perform denoising which is regarded as gradient masking. In this paper, we revisit the role of autoencoders in adversarial settings. Through the comprehensive experimental results and analysis, this paper presents the inherent property of adversarial robustness in the autoencoders. We also found that autoencoders may use robust features that cause inherent adversarial robustness. We believe that our discovery of the adversarial robustness of the autoencoders can provide clues to the future research and applications for adversarial defense.