53 research outputs found
Risk-based implementation of COLREGs for autonomous surface vehicles using deep reinforcement learning
No full textAutonomous systems are becoming ubiquitous and gaining momentum within the marine sector. Since the electrification of transport is happening simultaneously, autonomous marine vessels can reduce environmental impact, lower costs, and increase efficiency. Although close monitoring is still required to ensure safety, the ultimate goal is full autonomy. One major milestone is to develop a control system that is versatile enough to handle any weather and encounter that is also robust and reliable. Additionally, the control system must adhere to the International Regulations for Preventing Collisions at Sea (COLREGs) for successful interaction with human sailors. Since the COLREGs were written for the human mind to interpret, they are written in ambiguous prose and therefore not machine-readable or verifiable. Due to these challenges and the wide variety of situations to be tackled, classical model-based approaches prove complicated to implement and computationally heavy. Within machine learning (ML), deep reinforcement learning (DRL) has shown great potential for a wide range of applications. The model-free and self-learning properties of DRL make it a promising candidate for autonomous vessels. In this work, a subset of the COLREGs is incorporated into a DRL-based path following and obstacle avoidance system using collision risk theory. The resulting autonomous agent dynamically interpolates between path following and COLREG-compliant collision avoidance in the training scenario, isolated encounter situations, and AIS-based simulations of real-world scenarios. (C) 2022 The Author(s). Published by Elsevier Ltd
Supervised Classification Of Unlabeled Acoustic Data Utilizing Cross-Referencing With Labeled Images
No full textMed det økende fokuset på menneskeskapte endringer, settes mer og mer trykk på gjennvinbar og ansvarlig innhøsting av ressurser. I et forsøk på å bevare marint liv har den norske regjeringen bestemt at det må lages økologiske kart, som beskriver posisjon og mengde av viltlivsarter i norske farvann.
For å oppnå dette er et sonarsystem blitt utplassert i oslofjorden, for innsamling av store mengder marin data. Å "lable" akustisk data er tidkrevende og dyrt, og analyse er hovedsakelig basert på ad hoc matematiske metoder som er vanskelige å verifisere. Det er av interesse å finne mer kostnadseffektive metoder for å "lable" data, samt om nye gjennombrudd innen Maskinlæring (ML) kan forbedre klassifisering, sammenlignet med klassiske matematiske metoder.
I denne oppgaven demonstrerer forfatteren teknikker for innhøsting og analyse av marin data. En prosedyre for sammenkobling av optisk og akustisk data er utviklet og dens gyldighet demonstrert empirisk. Det er vist at de to datakildene kan tilstrekkelig relateres, både spatialt og temporalt. Resultatet er et rikt datasett, som er i stand til å utnytte de individuelle styrkene til hver datakilde.
Teknikker innenfor dyp læring er benyttet og et nevralt nettverk (NN) er utviklet og trent på opti-akustiske data. Dette viser at overvåket klassifisering av "unlabeled" akustisk data kan gjennomføres ved hjelp av kryssreferering med "lablet" optisk data.
Metodene var i stand til å korrekt klassifisere tilstedeværelsen av fisk med en nøyaktighet på 64.8% og regnes som et gjennomførbarhetsbevis.With the increased focus on man made changes to our planet and wildlife, more and more emphasis is put on sustainable and responsible gathering of resources. In an effort to preserve marine wildlife, the Norwegian government has proclaimed a necessity for creating ecological maps, detailing the presence and amount of wildlife species in Norwegian fjords and oceans.
To this end, a submerged sonar system has been deployed in the Oslo Fjord, gathering vast amounts of marine data. Procuring labeled acoustic data is time consuming and expensive, and analysis is predominantly based on ad hoc mathematical methods that are difficult to verify. It is of interest to determine if a more cost effective labeling procedure can be devised, and if the recent breakthroughs within machine learning (ML) enables improvements within classification, compared to classical mathematical methods.
In this thesis the author demonstrates techniques for acquiring and analysing marine data.
A procedure for interweaving optic and acoustic data is developed and its validity demonstrated empirically. It is shown that the two data sources can be sufficiently related, spatially and temporally, yielding a rich dataset capable of harnessing the individual strengths of each data source.
Deep learning techniques are employed and a neural network (NN) is developed and trained on opti-acoustic data. The results show that supervised classification of unlabeled acoustic data can be performed, utilizing cross-referencing with labeled optic data.
The methods were able to correctly classify the presence of fish with an accuracy of 64.8%, demonstrating a proof of concept
Graph‐based spatial–spectral feature learning for hyperspectral image classification
No full textClassifying hyperspectral data within high dimensionality is a challenging task. To cope with this issue, this study implements a semi-supervised multi-kernel class consistency regulariser graph-based spatial-spectral feature learning framework. For feature learning process, establishing the neighbouring relationship between the distinct samples from the highdimensional space is the key to a favourable outcome for classification. The proposed method implements two kernels and a class consistency regulariser. The first kernel constructs simple edges where every single vertex represents one particular sample and the edge weight encodes the initial similarity between distinct samples. Later the obtained relation is fed into the second kernel to obtain the final features for classification where the semi-supervised learning is conducted to estimate the grouping relations among different samples according to their similarity, class, and spatial information. To validate the performance of proposed framework, the authors conduct several experiments on three publically available hyperspectral datasets. The proposed work equates favourably with state-of-the-art works with an overall classification accuracy of 98.54, 97.83, and 98.38% for Pavia University, Salinas-A, and Indian Pines datasets, respectively.</p
High Fidelity Computational Fluid Dynamics Assessment of Wind Tunnel Turbine Test
No full textWe present, what to our best knowledge, is the most accurate numerical investigation of the wind tunnel tests carried out over a model wind turbine (known as NTNU Blind Test) at the Norwegian University of Sciences and Technology. We show numerical benchmarking of wake measurements against experimental data and similar investigations performed previously by researchers using Computational Fluid Dynamics (CFD) simulations. We have made a full 3D model of the wind turbine and used Sliding Mesh Interface (SMI) approach to handling the rotation of the rotor. The simulations are done with the use of OpenFoam and the k — ω Shear Stress Transport model to resolve turbulence using the Reynolds Average Navier-Strokes (RANS) technique. We present the numerically simulated spatial distribution of the flow field across the wake at zero angles of yaw for horizontal lines downstream of the rotor plane as that was the focus of the NTNU Blind Test presented in [1
Sustainable Active Travel in Environmentally Challenging Cities: A Systematic Review of Barriers and Strategies
No full textActive travel modes, such as walking and cycling, are essential for fostering sustainable urban transportation. However, their adoption in environmentally challenging areas—characterised by steep slopes, extreme weather, and rugged terrain—presents significant obstacles. This study addresses these challenges by conducting a systematic literature review of studies published between 2000 and 2024 to identify strategies that promote active travel in such contexts. Using a structured five-step methodology, 62 relevant articles were selected and analysed to explore common challenges and propose tailored solutions. The findings highlight critical barriers, including topographical difficulties, harsh climatic conditions, and adverse weather, all of which hinder walking and cycling. To address these barriers, this study identifies a range of solutions, including infrastructure enhancements such as bike lifts, e-bike systems, shaded walkways, and heated pavements, as well as policy measures like financial incentives and disincentive regulations. Importantly, this study makes a deliberate effort to avoid overgeneralised solutions by emphasising the need for interventions that are context-sensitive and tailored to specific environmental challenges, urban scales, and local conditions. By providing options for actionable strategies, this research offers a comprehensive foundation for developing inclusive and sustainable policies that encourage active travel in diverse and environmentally constrained urban settings
Industrial scale turbine and associated wake development -comparison of RANS based Actuator Line Vs Sliding Mesh Interface Vs Multiple Reference Frame method
No full textThis current work compares the three methodologies (Actuator Line model (ALM), Sliding Mesh Interface (SMI) and Multiple Reference Frame (MRF)) in modeling an industrial scale reference turbine at different tip speed ratios (TSR). The comparison shows that all the 3 models qualitatively predict the expected trend of power coefficient (Cp) vs TSR curve with an optima around TSR ≈ 7.5. But the quantitative values of the predicted Cp, the wake deficits and the flow patterns differ from model to model. Between ALM and MRF, the former predicts a relatively milder variation of Cp with TSR. A deeper analysis of the flow pattern and wake deficit behind the turbine helps in understanding the behavioral characteristics of these models. MRF shows variations in flow pattern with TSRs, like the associated stall conditions at TSR = 6 and an optimum angle of attack condition at TSR = 7.5. ALM shows only a slight variation in the flow pattern near the hub region (DU40 location) at different TSRs. This is because the blades are not resolved in ALM. Perhaps, such differences in flow-pattern predictions result in the differences in Cp vs TSR trends predicted by the MRF and ALM models. SMI, on the other hand, captures the complex 3D flow structures. The wake deficit comparison shows that both ALM and MRF model captures qualitatively higher wake deficit at TSR = 7.5 in the core wake region 0.8 > z/R > 0.2 as compared to TSR = 9 and TSR = 6. This behavior too is related to the observed flow pattern as captured by these models. Future studies may involve using LES in ALM to see if it improves the RANS predictions.publishedVersio
Numerical Analysis of NREL 5MW Wind Turbine: A Study Towards a Better Understanding of Wake Characteristic and Torque Generation Mechanism
No full textWith the increased feasibility of harvesting offshore wind energy, scale of wind turbines is growing rapidly and there is a trend towards clustering together higher number of turbines in order to harvest maximum yield and to leave a smaller footprint on the environment. This causes complex flow configurations inside the farms, the study of which is essential to making offshore wind energy a success. The present study focuses on NREL 5MW wind turbine with the following objectives (a)To compare Sliding Mesh Interface and Multiple Reference Frame modeling approaches and their predictive capabilities in reproducing the characteristics of flow around the full scale wind turbine. (b)To get a better insight into wake dynamics behind the turbine in near and far wake regions operating under different tip-speed-ratio and incoming turbulence intensities.publishedVersio
Influence of Tip Speed Ratio on Wake Flow Characteristics Utilizing Fully Resolved CFD Methodology
No full textDominant flow structures in the wake region behind the turbine employed in the Blind Test campaign [1], [2] is investigated numerically. The effect on the wake configuration at variable operating conditions are studied. The importance of the introduction of turbine tower inside the numerical framework is highlighted. High-fidelity simulations are performed with Multiple Reference Frame (MRF) numerical methodology. A thorough comparison among the cases is presented, and the wake evolution is analyzed at variable stations downstream of the turbine. Streamlines of flow field traveled towards ground adjacent to turbine tower and strongly dependent on the operating tip speed ratio. Wake is composed of tower shadow superimposed by rotor wake. Shadow of the tower varies from x/R=2 until x/R=4 and breaks down into small vortices with the interaction of rotor wake. This study also shows that the wake distribution consists of two zones; inner zone composed of disturbances generated by blade root, nacelle and the tower, and an outer zone consisting of tip vortices.publishedVersio
Effect of turbulence intensity on the performance of an offshore vertical axis wind turbine
No full textOffshore wind energy is one of the most competitive renewable energy resources available to us, which until now been under- exploited. Most of the problems associated with wind farm installation like land acquisition, low wind conditions and its visual impact can be eliminated to large extent by going offshore. In fact it is expected that by the year 2020, 40GW of offshore wind power capacity will be in operation. In an offshore context the wind turbine design methodologies have to address new challenges. For optimal performance the turbine needs to be huge in size and for horizontal axis wind turbines (HAWT) the diameter has already reached a size of 200m. Till now little attention has been paid to vertical axis offshore wind turbines. However, within the NOWITECH project new concepts for vertical axis turbine have been proposed and it might not take a long time before such turbines may become an realistic alternative for use offshore. The current work characterizes variable turbulence intensity flow field around a rotating vertical axis wind turbine (VAWT) in an offshore context. Complete three dimensional numerical transient simulations are performed accounting for the variation of multiple turbulence intensity levels associated with the oncoming wind. Usually offshore winds are highly turbulent in nature partially because of the rapid changes in wind directions along with the sea-air interaction. The results from the study indicate that due to the increase in the turbulence intensity level of 5% to 25% the performance of wind turbine decreases by almost 23% to 42% compared to no turbulence in the incoming wind field.publishedVersio
A full-scale 3D Vs 2.5D Vs 2D analysis of flow pattern and forces for an industrial-scale 5MW NREL reference wind-turbine
No full textNREL 5MW reference turbine, which is a popular, realistic and standardized industrial scale offshore turbine model, is used in this work for understanding the associated flow-complexities and for testing models. For such full-scale wind-turbine, the wish list is towards a accurate real-time prediction for better control and monitoring. Here, models like 2D based strip-theory [1] can help, but the extent of its applicability can be understood through a comparative performance of 2D Vs 2.5D Vs 3D CFD models. A stand-still blade condition is chosen for this study which can arise in a wind-farm when both yaw and pitch regulations are off-line. Further, even this simple stand-still condition is expected to have complex 3D effects due to blade geometry and due to the non-optimized conditions (blades twist being optimized for rotation). For such cases, the current work compares flow-profiles and forces computed by a 3D, 2.5D and 2D CFD models along the different sections of the NREL 5 MW turbine. The 2D CFD results are compared with experimentally measured drag and lift coefficient values as reported in DOWEC report [2]. The results from this study indicates that the flow close to the hub is dominated by complex 3D structures and unsteadiness while the three dimensionality and unsteadiness diminish as one moves away from the hub and towards the tip so much so that 2D simulations are sufficient for a faithful representation of the flow behavior. However, closer to the hub 2D simulations can not be utilized without adequate corrections. The work has implications for the 2D based approaches like strip-theory.publishedVersio
- …
