Archivio della ricerca - Fondazione Bruno Kessler
Not a member yet
21227 research outputs found
Sort by
Green hydrogen in the Alps: Mapping local stakeholders perspectives and identifying opportunities for decarbonization
Get PDFThe effects of climate change and reliance on fossil fuels in the Alps highlight the need for energy sufficiency, improved efficiency, and renewable energy deployment to support decarbonization goals. Hydrogen has gained attention as a versatile, zero-emission energy carrier with the potential to drive cleaner energy solutions and sustainable tourism in Alpine regions. This study shares findings from a hydrogen survey conducted within the Interreg Alpine Space AMETHyST project, which included questionnaires and roundtable discussions across Alpine territories. The survey explored hydrogen’s role in decarbonizing the Alps, gathering insights from local stakeholders about their knowledge, expertise, needs, and targets for hydrogen solutions. It also mapped existing hydrogen initiatives. Results revealed strong interest in hydrogen implementation, with many territories eager to launch projects. However, high investment and operational costs, along with associated risks, are key barriers. The absence of clear local hydrogen strategies and of a comprehensive regulatory framework also poses significant challenges. Incentivization schemes could facilitate initiatives and foster local hydrogen economies. The most promising application areas for hydrogen in the Alps are private and public mobility sectors. The residential sector, particularly in tourist accommodations, also presents potential. Regardless of specific uses, developing renewable energy capacity and infrastructure is essential to create green hydrogen ecosystems that can store excess renewable energy from intermittent sources for later use
Can AI be enabled to perform dynamical downscaling? A latent diffusion model to mimic kilometer-scale COSMO5.0_CLM9 simulations
No full textDownscaling based on deep learning (DL) is a key application in Earth system modeling, enabling the generation of high-resolution fields from coarse numerical simulations at reduced computational costs compared to traditional regional models. Additionally, generative DL models can potentially provide uncertainty quantification through ensemble-like scenario generation, a task prohibitive for conventional numerical approaches. In this study, we apply a latent diffusion model (LDM) to demonstrate that recent advancements in generative modeling enable DL to deliver results comparable to those of numerical dynamical models, given the same input data, preserving the realism of fine-scale features and flow characteristics at reduced computational costs. We apply our LDM to downscale ERA5 data over Italy up to a resolution of 2 km. The high-resolution target data consist of 2 m temperature and 10 m horizontal wind components from a dynamical downscaling performed with COSMO-CLM. A selection of predictors from ERA5 is used as input, and a residual approach against a reference U-Net is leveraged in applying the LDM. The performance of the generative LDM is compared with reference baselines of increasing complexity: a quadratic interpolation of ERA5, a U-Net, and a generative adversarial network (GAN) built on the same reference U-Net. Results highlight the improvements introduced by the LDM architecture combined with the residual approach, outperforming all the baselines in terms of spatial error, frequency distributions, and power spectra. These findings point out the potential of LDMs as cost-effective, robust alternatives for downscaling applications (e.g., downscaling of climate projections), where computational resources are limited but high-resolution data are critical
3D-printed swirler for enhanced heat transfer in evacuated tube solar collectors
No full textSolar thermal energy is a promising sustainable resource for decarbonizing the energy sector. The optimization of existing solar systems can be a valuable and sustainable approach to promoting a transition to sustainable energy and as well as reducing the land occupation of solar plants. For that reason, in this work, we studied and realized a “swirl generator” device that controls the turbulence of water in the pipe of a solar collector to improve heat transfer. The swirler device is a small insert (2.5 cm in length) with 3 winglets with a helicoidal shape; the customized devices by 3D printing has been used for the retrofitting existing collectors. The swirler has been realized using 3D printing using PA polymer and mounted on the inlet of the evacuated tube collector. We support our experimental work with a numerical thermos fluid-dynamics model based on the Finite Element Method (FEM). The numerical simulations have been effectively performed using the Comsol Multiphysics® software to study in a predictive manner the effect of the swirling device on the fluid dynamics of the collector pipe. The swirler has been tested on a solar collector based on an evacuated tube collector. The experimental results are in good agreement with simulation results, moreover, an improvement in the heat transfer of 25 % and an increase in the homogeneity of the temperature has been observed using the swirler device
Development and wafer-level characterization of the first production of DC-RSD sensors at FBK
No full textDC-RSDs are silicon sensors that aim to provide a time resolution for minimum ionizing particles in the order of 30 ps and a spatial resolution of a few percent of the pixel pitch. This performance is enabled by internal charge multiplication and resistive charge division between the electrodes. The time resolution is expected to be the same as Low Gain Avalanche Diodes (LGADs). The performance of the resistive charge division mechanism was demonstrated in AC-LGADs or Resistive Silicon Detectors (RSD) where a capacitive coupling between readout electrodes and resistive layer was employed. DC-RSDs use a direct coupling between the electrodes and the resistive layer avoiding the bipolar signal of the AC-coupled designs and providing a better bias distribution to the resistive layer. The channel segmentation does not rely on interrupting the gain layer, aiming to maintain a fill factor close to 100%. The first DC-RSD sensor batch was fabricated at FBK with the aim to demonstrate the soundness of the sensor concept. This batch contains design variations of the sensors, trench isolation between channels to provide signal confinement on the resistive layer, and techniques to lower the contact resistance between readout electrodes and resistive layer. This paper summarizes the characterization of the sensors performed at wafer level
Design and optimisation of radiation resistant AC- and DC-coupled resistive LGADs
No full textFuture high-energy physics experiments require a paradigm shift in radiation detector design. In response to this challenge, resistive LGADs that combine Low Gain Avalanche Diode technology with resistive readout have been developed. The prototypes created so far, employing AC-coupled contacts, have demonstrated impressive performance, achieving a temporal resolution of 38 ps and a spatial resolution of 15 μm with a pixel pitch of 450 μm.
To tackle some of the issues encountered up to this point, particularly the non-uniform response across the entire surface of the detector, a new version with DC-coupled contacts has recently been developed. The Synopsys® Sentaurus TCAD simulations that have guided the design of their first production, released by the Fondazione Bruno Kessler in November 2024, will be presented below along with a concise summary of the history of the prototypes with AC-coupled contacts
Dry Plasma Synthesis of Nanohybrids and Nanofluids for Water Electrolysis: Toward CRM-Free, Green and Efficient Catalytic Materials
No full textThis work presents a dry and green synthesis approach using RF magnetron sputtering, a physical vapor deposition (PVD) technique, to develop nanohybrids (catalyst-coated powders) and nanofluids (nanoparticles suspended in liquids) catalysts for water electrolysis.1,2 The method eliminates hazardous chemicals and multi-step processes, enabling sustainable and scalable catalyst production.2 For nanohybrids, copper-coated multi-walled carbon nanotubes (Cu/CNTs) are synthesized by O2 plasma treatment of CNTs followed by RF sputtering of Cu. A vibrating deposition stage ensured uniform nanoparticle coating. XPS confirmed enhanced surface functionality, while dynamic light scattering (DLS) and thermogravimetric analysis (TGA) revealed improved dispersion and material loading. Finally, the produced nanohybrids exhibited hydrogen evolution reaction (HER) activity. For nanofluids, gold nanoparticles (AuNPs) were sputtered directly into polyethylene glycol (PEG) and transferred into a Nafion ionomer to fabricate a catalyst-coated membrane (CCM). Characterization using UV-VIS, TEM, SEM, XPS, and AFM confirmed the preservation of nanoscale morphology and uniform dispersion. The resulting CCM exhibited excellent HER activity, with an onset potential of 50 mV and a Tafel slope of 38 mV/dec. This work demonstrates RF magnetron sputtering as a clean, scalable, dry synthesis technique for producing CRM-free, high-efficiency catalysts for sustainable hydrogen production through water electrolysis
Prepending or Cross-Attention for Speech-to-Text? An Empirical Comparison
No full textFollowing the remarkable success of Large Language Models (LLMs) in NLP tasks, there is increasing interest in extending their capabilities to speech—the most common form of communication. The most widespread approach to integrating speech into LLMs is dense feature prepending (DFP), which prepends the projected speech representations to the textual representations, allowing end-to-end training with a speech encoder. This raises questions about the need for a sophisticated speech encoder for DFP and how its performance compares with a standard encoder-decoder (i.e., cross-attention) architecture. We compare DFP and cross-attention under a variety of configurations, such as CTC compression, sequence-level knowledge distillation, on monolingual, bilingual, and multilingual models. To perform a controlled architectural comparison, we train all models from scratch rather than using large pretrained models and use comparable data and parameter settings, testing speech-to-text recognition (ASR) and translation (ST) on MuST-C v1.0 and CoVoST2 datasets. Despite the wide adoption of DFP, our results do not indicate a clear advantage of DFP over cross-attention
RF-MEMS networks in future telecommunications: a reconfigurable module with amplitude, phase and switching control for B5G/6G applications
No full textThe challenging scenario imposed by 5G and future telecommunication standards in terms of coverage and data rates is effectively addressed at system level by access points utilizing millimeter-wave (mmWave) frequency bands and antenna arrays. The inevitable densification of the access points and the presence of multiple power-hungry components in their radio frequency (RF) front ends hold the attention towards passive solutions aimed at their beamforming (BF) architectures. Among the different passive alternatives, Radio Frequency Micro Electro-Mechanical Systems (RF-MEMS) technology constitutes a valid candidate in terms of broadband behaviour, high-performance, low power consumption and miniaturization of its implementations. The potential of such technology in the field of reconfigurable BF networks is outlined in this article by the proposed module, which combines three phase-shifting cells, an attenuator and a switch into a single monolithic layout with an area of 3.36 × 9.51 mm2. The simulated performance of the device highlighted an amplitude control spanning from − 5.39 dB to − 13.51 dB, achievable phase shifts from 14.03° to 158.46°, still providing an isolation better than − 13 dB up to 27.5 GHz, that is the upper bound of the addressed N258 5G band (24.25–27.5 GHz). A low power consumption is guaranteed by the electrostatic actuation of its membranes, whose actuation voltage corresponds to 7 V
LVG-SfM: Learning-Based View-Graph Generation for Robust on-the-Fly SfM
No full textStructure from Motion (SfM) has been widely studied in many fields, such as computer vision, photogrammetry, robotics, etc. Recent advancements focus on improving the real-time performance of SfM, which is crucial for applications in augmented reality, mixed reality, robotics, etc. However, the robustness of real-time processing is still limited by outliers in the feature extraction and matching process, stemming from challenging scenes depicting objects with poor texture, repetitive structures, and symmetric objects, which can cause blunders in the view-graph. Focusing on these scenes, a Learning-based View-Graph generation method (LVG-SfM) is investigated and integrated into the on-the-fly SfM pipeline [43]. First, to provide a higher number of reliable matches and generate a more robust view-graph, a set of SoTA learning-based feature extraction and matching methods [19] are tested. Then, the spuriously incorrect two-view geometries generated from repetitive structures are removed from the view-graph with the help of SoTA learning-based disambiguation network - Doppelgangers [3]. Experimental results demonstrate that our LVG-SfM can successfully work on-the-fly on challenging ambiguous scenes with poor textures and repetitive structures, achieving correct scene reconstructions and robustifying SfM. Project website at: https://sygant.github.io/lvgsfm
Night and Day Aerial Photogrammetry
No full textRecent advancements in aerial imaging, including high-resolution sensors and integrated GNSS/IMU systems, have significantly enhanced photogrammetric methods for geospatial data acquisition. While most aerial data is captured during daylight, night-time imaging is increasingly being used in applications such as urban analysis and disaster assessment. However, automatic co-registration of day and night imagery remains challenging due to substantial radiometric differences. This study investigates the use of deep learning-based feature matching techniques for the alignment of multi-temporal, day-night aerial datasets. Experimental results show that feature extraction is highly sensitive to scale, with only a limited subset of deep learning (DL) methods—particularly ALIKED with LightGlue and SuperPoint with SuperGlue—proving robust under low-illumination conditions. Additionally, a U-Net-like model was trained to pre-process night-time images by approximating their radiometric characteristics to those of daytime images, enabling consistent feature matching across all tested methods. Among them, ALIKED with LightGlue offered the best balance between match quantity and computational efficiency. Object-space evaluations confirmed that the proposed pre-processing step significantly improves co-registration accuracy. The methodology offers a promising foundation for future multi-sensor and multi-modal image alignment tasks, including RGB-thermal and 2D-3D matching