Inference of the ocean environment using measured and simulated acoustic data to train deep learning algorithms

Sound generated by anthropogenic activities associated with the construction and maintenance of marine renewable energy platforms, such as piling and geoacoustic surveys, has the potential to adversely affect the health and wellbeing of marine mammals. Understanding how sound propagates in marine environments is critical to ensuring the responsible deployment of marine platforms. The use of machine learning algorithms to determine physical quantities of a complex ocean environment via acoustic data is relatively unexplored. Traditionally, research in underwater acoustics (the field of sound wave generation, propagation, scattering and reception in water) focuses on the use of sound wave navigation and ranging (sonar) systems for communication, sensing, marine wildlife monitoring, target detection, and exploration. The maritime energy sector uses sonar for environmental monitoring, to assess the impact of renewable energy platforms on marine life and the seabed, for example the FLOWBEC project at the European Marine Energy Centre in Orkney, and the Menter Môn led Marine Characterisation & Research Project (MCRP) in the Morlais Demonstration Zone off Holy Island, Anglesey.

The operational use of sonar systems is strongly dependent on an accurate acoustic description of the marine environment. Unfortunately, direct measurement of these acoustic properties is difficult and expensive, and acoustic models are limited by the extent of data and information that is available. Ocean environment information is equally important for our understanding of complex ocean processes and sustainable use of the oceans. The process of extracting information indirectly from acoustic data is inversion. Since an acoustic measurement results from an acoustic signal propagated through the environment, it contains acoustic information about the ocean environment that can be derived using appropriate models and methods. The aim of this project is to use data-driven machine learning models to improve the description of the ocean environment resulting from inversion, to represent a broader range of ocean properties relevant to underwater acoustics, and to support our understanding of the impact of maritime energy platforms on the ocean environment. The focus of the project will be the development of machine learning models which can be used for acoustic data collected from in-situ and remote sensors, historical data, and synthetic data obtained from simulation and modelling. The ultimate goal is to obtain an up-to-date and accurate representation of the acoustic environment for any sonar deployment.

Met2Adapt: Modelling of the dynamic response for offshore wind farms

We will develop novel mathematical models to predict the dynamic phenomena arising in offshore wind farms and further model the acoustic waves generated by individual, and clusters of, wind turbines to help quantify the effect on wildlife, including birds, marine mammals, and fish.

Funded by the Marie Skłodowska-Curie Actions (Horizon Europe, Grant Agreement No. 101227175), Met2Adapt aims to recruit 16 PhD candidates who will be employed by one of the 10 partner institutions across Europe. Met2Adapt puts forward an ambitious research and training plan that will foster a new generation of researchers able to design and deliver sustainable meta-materials for vibration mitigation, self-aware meta-components and eventually carbon-efficient yet safe meta-structures for the renewable energy sector. The focal point of this research will be the deployment of custom-fit solutions for infrastructure that is critical to the European energy resilience, i.e. offshore and onshore wind farms, and wave-energy converters.

BATNav: Bioinspired Acoustic Tactical Navigation

Our highly interdisciplinary research team is working on an innovative project aimed at enhancing first-person navigation in challenging environments using acoustics. The project focuses on developing an acoustic system that provides rapid identification of people, obstacles and topology (e.g. tunnels, stairways) in closed spaces, facilitating navigation for teams such as firefighters in complex, hazardous environments.

Multiscale multiphysics structured interfaces

Whilst research in metamaterials (structured materials that can control waves and energy) has undergone a renaissance, the majority of work has focused on controlling a single physical domain - usually light. In contrast, this project seeks to develop a mathematical framework to study, design, and create multiphysical metasurfaces (Maradudin, 2011) capable of manipulating waves across many physical phenomena. Metasurfaces are structured interfaces where two or more physical systems interact, such as solids and fluids. Structured interfaces arise in many real-world applications - anywhere two media meet, from antenna designs, to the interface between the ocean and sea-bed, the foundations of modern buildings, medical implants, and safety-critical power generation components.

Machine Learning for Data Driven Sound Propagation Modelling

This project will develop a series of high-fidelity digital twins capable of encapsulating a number of critical dynamic phenomena, which affect the propagation of sound waves through ocean environments, including internal waves, multi-scale structural thermal and temporal variations and fluctuations, scattering by non-smooth interfaces and boundaries (e.g. semi-submerged structures, sea bed, surface), currents, eddies, and fronts. The resulting sound propagation models, and advanced understanding embodied within these models, will enable the Royal Navy and other users of the ocean, the means to improve the effectiveness of their sonar systems and achieve the best results from sonar deployments and operations.

The developed models will be capable of intelligently adapting their approach and selecting the best solution method based on the input data, computational and operational constraints, desired outputs, and physical configuration. These models will incorporate advances in Finite Element Methods, via GPU parallel computing capability that has been largely untapped for underwater acoustics, along with hybrid semi-analytic coupling methods. Stochastic analysis of physical scattering models, incorporating real-world data, will provide a major step forward in the capability available to investigate dynamic ocean mechanisms. Analysis of simulated and measured oceanographic and acoustic data, including the use of artificial intelligence and machine learning techniques, will be supported by a parallel PhD project in the Department of Mathematical Sciences on Advances in mathematical modelling to study complex sound propagation in an inhomogeneous moving ocean. This will help to identify the properties and behaviour of different mechanisms. New ways of representing the specified mechanisms will be developed, and environment-specific modelling tools and innovative mathematical representations will be implemented.

MUSICA: Modelling UltraSonic Inspection of Challenging defects for Automated analysis

We will develop automated ultrasonic data analysis tools to improve the reliability of detection, sizing, and characterisation of defects which occur in high safety significance industrial plant. The project will develop capability for defects of critical industrial importance, targeting two species: Thermal Fatigue (a service induced species) and Hydrogen cracking (a manufacturing/welding defect). Both species are known to occur and often materially impact plant availability/longevity. This project will also enable capability for other challenging defect species going forward, such as stress corrosion cracks.

Using machine learning and artificial intelligence to improve the tracking of vessels in sonar spectrograms

This PhD project explores creating an AI model that can correctly classify quiet targets in waterfall (sonar) data. Currently, waterfall data is analysed by human operators; however, this is time-consuming and expensive; these human operators outperform traditional automated passive contact follower algorithms, such as the Kalman and Alpha-Beta filters: these filters are susceptible to the abundant underwater noise and struggle with crossing tracks and quiet contacts. In contrast, humans can use their experience to learn how to mitigate the challenging aspects of the task. An automatic detection and tracking model that is more accurate and robust than traditional methods would reduce the human operator’s workload.

Task 64: Modelling and Algorithm Development

Development of a modelling capability to simulate acoustic signal propagation through an oceanic environment, interactions of the signals with objects within the environment, and to predict the signals received by sensors.