Norwegian University of Science and Technology

Ambitious targets on reduction of greenhouse-gas emissions have motivated further studies to improve energy efficiency in offshore gas and oil production. Identifying the causes of inefficiencies and the improvement potentials within these processes is crucial. The oil and gas processing plant on a North Sea platform is evaluated by advanced exergy analysis for a real production day. The study focuses on components and sub-systems with high exergy destruction through conventional exergy analysis in previous research. Splitting the exergy destruction into endogenous and exogenous parts provides information about mutual interdependencies among the system components. The results show that the inefficiencies of compressors are attributed to their inherent irreversibility, while the exergy destruction within the coolers could particularly be reduced by improving the remaining system components. Further, the total exergy destruction avoidable by improving each single component determines the importance of the components. The results indicate that the compressors have relatively large exergy saving potential (14% of total power consumption), while it is relatively low for coolers. Advanced exergy analysis suggests an optimization sequence different from the conventional exergy analysis. The findings indicate that the improvement efforts should be focused essentially on the compressors, especially for the recompression compressors with anti-surge operations.

Convolutional neural networks (CNNs) have demonstrated encouraging results in image classification tasks. However, the prohibitive computational cost of CNNs hinders the deployment of CNNs onto resource-constrained embedded devices. To address this issue, we propose, a comprehensive compression framework to reduce the computational overhead of CNNs. In, we first introduce dynamic image cropping, where we design a lightweight foreground predictor to accurately crop the most informative foreground object of input images for inference, which avoids redundant computation on background regions. Subsequently, we present compound shrinking to collaboratively compress the three dimensions (depth, width, and resolution) of CNNs according to their contribution to accuracy and model computation. Dynamic image cropping and compound shrinking together constitute a multi-dimensional CNN compression framework, which is able to comprehensively reduce the computational redundancy in both input images and neural network architectures, thereby improving the inference efficiency of CNNs. Further, we present a dynamic inference framework to efficiently process input images with different recognition difficulties, where we cascade multiple models with different complexities from our compression framework and dynamically adopt different models for different input images, which further compresses the computational redundancy and improves the inference efficiency of CNNs, facilitating the deployment of advanced CNNs onto embedded hardware. Experiments on ImageNet-1K demonstrate that reduces the computation of ResNet-50 by 48.8% while improving the top-1 accuracy by 0.8%. Meanwhile, we improve the accuracy by 4.1% with similar computation compared to HRank. the state-of-the-art compression framework. The source code and models are available at

Sparse matrix-vector multiplication (SpMV) on FPGAs has gained much attention. The performance of SpMV is mainly determined by the number of multiplications between non-zero matrix elements and the corresponding vector values per cycle. On the one side, the off-chip memory bandwidth limits the number of non-zero matrix elements transferred from the off-chip DDR to the FPGA chip per cycle. On the other side, the irregular vector access pattern poses challenges to fetch the corresponding vector values. Besides, the read-after-write (RAW) dependency in the accumulation process shall be solved to enable a fully pipelined design. In this work, we propose an efficient FPGA-based sparse matrix-vector multiplication accelerator with data reuse-aware compression. The key observation is that repeated accesses to a vector value can be omitted by reusing the fetched data. Based on the observation, we propose a reordering algorithm to manually exploit the data reuse of fetched vector values. Further, we propose a novel compressed format called data reuse-aware compressed (DRC) to take full advantage of the data reuse and a fast format conversion algorithm to shorten the preprocessing time. Meanwhile, we propose an HLSfriendly accumulator to solve the RAW dependency. Finally, we implement and evaluate our proposed design on the Xilinx Zynq-UltraScale ZCU106 platform with a set of sparse matrices from the SuiteSparse matrix collection. Our proposed design achieves an average 1.18x performance speedup without the DRC format and an average 1.57x performance speedup with the DRC format w.r.t. the state-of-the-art work respectively.

Edge intelligence systems, the intersection of edge computing and artificial intelligence (AI), are pushing the frontier of AI applications. However, the complexity of deep learning models and heterogeneity of edge devices make the design of edge intelligence systems a challenging task. Hardware-agnostic methods face some limitations when implementing edge systems. Thus, hardware-aware methods are attracting more attention recently. In this paper, we present our recent endeavors in hardware-aware design and optimization for edge intelligence. We delve into techniques such as model compression and neural architecture search to achieve efficient and effective system designs. We also discuss some challenges in hardware-aware paradigm.

Online federated learning (FL) enables geographically distributed devices to learn a global shared model from locally available streaming data. Most online FL literature considers a best-case scenario regarding the participating clients and the communication channels. However, these assumptions are often not met in real-world applications. Asynchronous settings can reflect a more realistic environment, such as heterogeneous client participation due to available computational power and battery constraints, as well as delays caused by communication channels or straggler devices. Further, in most applications, energy efficiency must be taken into consideration. Using the principles of partial-sharing-based communications, we propose a communication-efficient asynchronous online federated learning (PAO-Fed) strategy. By reducing the communication load of the participants, the proposed method renders participation more accessible and efficient. In addition, the proposed aggregation mechanism accounts for random participation, handles delayed updates and mitigates their effect on accuracy. We study the first and second-order convergence of the proposed PAO-Fed method and obtain an expression for its steady-state mean square deviation. Finally, we conduct comprehensive simulations to study the performance of the proposed method on both synthetic and real-life datasets. The simulations reveal that in asynchronous settings, the proposed PAO-Fed is able to achieve the same convergence properties as that of the online federated stochastic gradient while reducing the communication by 98 percent.

We expand the security model for group key exchange of Poettering et al. (CT-RSA 2021) to allow for more fine-tuned reveal of both state and keying material. The expanded model is used to analyse the security of hybrid group key exchange schemes, compositions of distinct group key exchange schemes where either subprotocol may be separately compromised. We then construct a hybrid group key exchange protocol that we show to be as secure as its sub-protocols. Furthermore, we use the notion of a secure element to develop a lightweight, low transmission group key exchange protocol. This protocol is used to develop a hybrid scheme that offers dynamic group membership and is suitable for use in constrained networks.

Following the European Union’s implementation of Energy Performance Certificates (EPCs) for buildings, the capitalization of energy efficiency in transaction prices and rents has been subject to much research. This paper uses different identification strategies for the Norwegian residential sales (N = 750,000) and rental (N = 670,000) markets to highlight the endogeneity and methodological limitations associated with assessing the price effects of energy efficiency and the signaling effect of label adoption. We find that the valuation of energy efficiency is subject to unobserved location and quality bias, that labeling has immediate, short-run, and long-run price effects and that different effects are observed in different submarkets. We provide evidence that sample selection issues related to location and time, with methodological and data limitations, are essential factors that must be considered when assessing the effects of the EPC implementation.

Blood volume (BV) is an important clinical parameter and is usually reported per kg of body mass (BM). When fat mass is elevated, this underestimates BV/BM. One aim was to study if differences in BV/BM related to sex, age, and fitness would decrease if normalized to lean body mass (LBM). The analysis included 263 women and 319 men (age: 10–93 years, body mass index: 14–41 kg/m²) and 107 athletes who underwent assessment of BV and hemoglobin mass (Hbmass), body composition, and cardiorespiratory fitness. BV/BM was 25% lower (70.3 ± 11.3 and 80.3 ± 10.8 mL/kgBM) in women than men, respectively, whereas BV/LBM was 6% higher in women (110.9 ± 12.5 and 105.3 ± 11.2 mL/kgLBM). Hbmass/BM was 34% lower (8.9 ± 1.4 and 11.5 ± 11.2 g/kgBM) in women than in men, respectively, but only 6% lower (14.0 ± 1.5 and 14.9 ± 1.5 g/kgLBM)/LBM. Age did not affect BV. Athlete's BV/BM was 17.2% higher than non‐athletes, but decreased to only 2.5% when normalized to LBM. Of the variables analyzed, LBM was the strongest predictor for BV (R² = .72, p < .001) and Hbmass (R² = .81, p < .001). These data may only be valid for BV/Hbmass when assessed by CO re‐breathing. Hbmass/LBM could be considered a valuable clinical matrix in medical care aiming to normalize blood homeostasis.

Background Loneliness has become a significant public health problem and should be addressed with more research over a broader period. This study investigates the variations in the prevalence of loneliness among a nationally representative study population of Norwegian adolescents over the last three decades and whether age, gender, self-rated health, and mental distress are associated with these changes. Methods Adolescents aged 13–19 years completed the structured and validated questionnaires from the three waves of the Young-HUNT Study: 1995–1997, 2006–2008, and 2017–2019. Loneliness was measured with one item asking, ‘Are you lonely?’. Hopkins Symptom Checklist-5 was used to measure mental distress (cut-off ≥ 2). Self-rated health was assessed by a single question ‘How is your health at the moment?’ Measures were provided by self-report. Descriptive analyses were stratified by age, gender, self-rated health, and mental distress. Linear-by-Linear association test across survey years was performed to test time trends of loneliness. Logistic regression was used to analyze the cross-sectional associations of self-rated health and mental distress with loneliness, adjusting for sociodemographic factors in all three waves of Young-HUNT. Results Loneliness prevalence doubled from 5.9% in 1995/97 to 10.2% in 2017/19 in the total population sample. The highest loneliness prevalence and an increase from 8.9% in 1995/97 to 16.7% in 2017/19 was observed in girls of 16–19 years. Among mentally distressed adolescents, loneliness increased from 22.3% in 1995/97 to 32.8% in 2006/08 and lowered to 27% in 2017/19. Increasing loneliness prevalence was seen in those with poor self-rated health, i.e., 14.6% in 1995-97 and 26.6% in 2017-19. Mental distress and poor self-rated health were associated with higher odds of loneliness in each wave (p < 0.001). Conclusion The results highlight the increasing burden of loneliness in the Norwegian adolescent population, especially girls. Those with mental distress and poor self-rated health have a higher risk of experiencing loneliness. Thus, health-promoting upbringing environments for children and adolescents that support mutual affinity, social support, integration, and belongingness in adolescents’ daily arenas are essential.

Conversational remembering entails that people engage in recalling past experiences, which may themselves have been shared. Conversational remembering comes with social benefits for the person telling the narrative and the one receiving it (e.g., developing and strengthening friendships, fostering entertainment, and consolidating group identity). COVID-19 lockdowns have significantly affected social interaction, including face-to-face interactions where conversational remembering occurs. The aim of this study was to explore how WhatsApp group messages supported conversational remembering in a large group of friends living in Buenos Aires where a complete lockdown was established between 19 March 2020 and 6 November 2020. To accomplish such aim, we conducted a mixed-methods longitudinal study. The data consisted of 32,810 WhatsApp group messages collected over a period of 700 consecutive days, from 13 April 2019 to 13 March 2021. Our study shows that WhatsApp group messages enabled group members to keep connected during the COVID-19 lockdown period. This occurred by remembering together situations, events, and actions associated with the group's identity. The use of WhatsApp group messages may have represented an adaptive collective behaviour in response to changes in global social norms.

Objectives Comparative longitudinal analyses of cellular composition and peripheral blood gene expression in Rheumatoid arthritis (RA), systemic lupus erythematosus (SLE) and healthy pregnancies. Methods In total, 335 whole blood samples from 84 RA, SLE and healthy controls before pregnancy, at each trimester, 6 weeks, 6 months and 12 months post partum were analysed. We combined bulk and single cell RNA analyses for cell-type estimation, validated by flow cytometry, before combining this in a cell-type adjusted analysis for an improved resolution of unrecognised gene expression changes associated with RA and SLE pregnancies. Results Patients were well regulated throughout pregnancy, and few had pregnancy complications. In SLE, the interferon signature was augmented during pregnancy, and the pregnancy signature was continued post partum. An altered cell type composition strongly influences the profile. In the pregnancy signature, transcripts involved in galactosylation potentially altering the effector functions of autoantibodies became more evident. Several genes in the adjusted RA signature are expressed in mucosal associated invariant T cells. Conclusion We found distinct RA, SLE and pregnancy signatures, and no expression patterns could be attributed to medication or disease activity. Our results support the need for close postpartum follow-up of patients with SLE. Gene expression patterns in RA were closer to healthy controls than to SLE, and primarily became evident after cell-type adjustment. Adjusting for cell abundance unravelled gene expression signatures less associated with variation in cell-composition and highlighted genes with expression profiles associated with changes in specialised cell populations.

This is a case of a 60-year-old woman who presented in hypovolemic shock due to an upper gastrointestinal bleeding. She had a history of chronic alcohol abuse. Standard resuscitation measures with intravenous access and blood transfusions were started. She underwent an upper endoscopy in the operating room under general anesthesia. A bleeding fibrotic lesion in the distal esophagus was found and treated with adrenaline injections and placement of a temporary esophageal stent. After an initial stabilization, she deteriorated in the operating room and developed severe circulatory failure. A bedside echocardiogram was performed and demonstrated severely depressed left ventricular systolic function and mid-ventricular hypokinesia consistent with cardiogenic shock secondary to Takotsubo cardiomyopathy. Instead of further volume resuscitation, treatment efforts pivoted toward inotropic support and afterload reduction with dobutamine and nitroglycerine, respectively. She remained sedated, intubated, and ventilated in the intensive care unit for 4 days and gradually regained left ventricular function. On day 10 left ventricular hypokinesia had subsided and ejection fraction had improved from around 15 to 58%. The case demonstrates the usefulness of echocardiography in phenotyping shock and how it can dramatically change management with improved patient outcome.

Trackwork planning and scheduling are demanding because they require strategic foresight and must be completed well in advance. In Sweden, trackwork is performed by maintenance contracting companies during an operation period free from trains. In the contractors’ practice, once the maintenance plan is authorised, some unexpected events might interrupt the plan’s execution, leading to uncertainties. The purpose of this study is to identify and classify uncertainties and strategies applied to manage uncertainties in the contractors’ everyday planning and scheduling of trackwork. This work presents semi-structured interviews with foremen and planners at railway maintenance contracting companies in Sweden. The main findings show that in trackwork planning and scheduling, contractors deal with two types of uncertainties: internal and external. We categorised uncertainties and strategies to deal with uncertainties and described them on tactical and operational levels. The majority of the revealed uncertainties led to trackwork rescheduling. Furthermore, we suggest that current strategies to manage uncertainties applied at contracting companies can be improved by revising organisational design strategies for maintenance projects. This work increases the understanding and supports the management of uncertainties in trackwork planning and scheduling.

In recent years, reinforcement learning (RL) systems have shown impressive performance and remarkable achievements. Many achievements can be attributed to combining RL with deep learning. However, those systems lack explainability, which refers to our understanding of the system’s decision-making process. In response to this challenge, the new explainable RL (XRL) field has emerged and grown rapidly to help us understand RL systems. This systematic literature review aims to give a unified view of the field by reviewing ten existing XRL literature reviews and 189 XRL studies from the past five years. Furthermore, we seek to organize these studies into a new taxonomy, discuss each area in detail, and draw connections between methods and stakeholder questions (e.g., “how can I get the agent to do _?”). Finally, we look at the research trends in XRL, recommend XRL methods, and present some exciting research directions for future research. We hope stakeholders, such as RL researchers and practitioners, will utilize this literature review as a comprehensive resource to overview existing state-of-the-art XRL methods. Additionally, we strive to help find research gaps and quickly identify methods that answer stakeholder questions.

Phage treatment is suggested as an alternative to antibiotics; however, there is limited knowledge of how phage treatment impacts resident bacterial community structure. When phages induce bacterial lysis, resources become available to the resident community. Therefore, the density of the target bacterium is essential to consider when investigating the effect of phage treatment. This has never been studied. Thus, we invaded microcosms containing a lake-derived community with Flavobacterium columnare strain Fc7 at no, low or high densities, and treated them with either the bacteriophage FCL-2, the antibiotic Penicillin or kept them untreated (3 × 3 factorial design). The communities were sampled over the course of one week, and bacterial community composition and density were examined by 16S rDNA amplicon sequencing and flow cytometry. We show that phage treatment had minor impacts on the resident community when the host F. columnare Fc7 of the phage was present, as it caused no significant differences in bacterial density α- and β-diversity, successional patterns, and community assembly. However, a significant change was observed in community composition when the phage host was absent, mainly driven by a substantial increase in Aquirufa. In contrast, antibiotics induced significant changes in all community characteristics investigated. The most crucial finding was a bloom of γ-proteobacteria and a shift from selection to ecological drift dominating community assembly. This study investigated whether the amount of a bacterial host impacted the effect of phage treatment on community structure. We conclude that phage treatment did not significantly affect the diversity or composition of the bacterial communities when the phage host was present, but introduced changes when the host was absent. In contrast, antibiotic treatment was highly disturbing to community structure. Moreover, higher amounts of the bacterial host of the phage increased the contribution of stochastic community assembly and resulted in a feast-famine like response in bacterial density in all treatment groups. This finding emphasises that the invader density used in bacterial invasion studies impacts the experimental reproducibility. Overall, this study supports that phage treatment is substantially less disturbing to bacterial communities than antibiotic treatments.

This study presents an innovative framework for classifying and predicting odor intensity in perfumery, combining scientific machine learning with mechanistic modeling to enhance fragrance design precision. A probabilistic weight assignment is introduced, utilizing scent classifier outputs to determine the contribution of each fragrance component, thereby recognizing the subjective nature of scent classification and variability in olfactory perception. Additionally, an uncertainty analysis framework is integrated, quantifying uncertainties within perfume diffusion and human sensory perception models, thus improving model adaptability and reliability. The methodology comprises three parts: a perfume diffusion model that simulates fragrance molecule evaporation and dispersion, an odor perception model using Odor Value for scent intensity quantification, and an uncertainty quantification that rigorously analyzes model parameters and predictions. This approach aims to scientifically advance the art of perfumery, allowing for the creation of sophisticated fragrances with enhanced predictive accuracy.

Ice nucleation and formation play pivotal roles across various domains, from environmental science to food engineering. However, the exact ice formation mechanisms remain incompletely understood. This study introduces a novel...