About the Project
This project is part of the Centre for Doctoral Training in Geospatial Systems. The successful candidate will be co-supervised by academics from within the School of Engineering at Newcastle University and work alongside our external partner Northumbrian Water Group. The student will form part of Cohort 5 and commence in September 2023. Please visit our website for details on how to apply.
Critical infrastructure systems are highly spatially distributed interdependent temporally complex systems. Increasingly infrastructure operators require not only the ability to understand the long term (decadal) spatial risk of interconnected assets to future extreme events, but also require real-time knowledge of the spatial risk posed as extreme events, such as flooding, occur. Such knowledge is critical so that improved real-time decisions and interventions with regards to asset performance, defence and failure consequences can be understood.
With the emergence of sensor networks and IoT capability, it is now becoming increasingly viable to develop decision support platforms that allow the current and near future exposure, vulnerability and adaptive capacity of critical infrastructure to be assessed in near real-time to understand and predict into the near future spatial risk of critical infrastructure.
This PhD will focus on the development of real-time decision support capability for (interdependent) infrastructure water systems in relation to emerging extreme events such as flooding and dependency failures. The PhD will focus on the development of the analysis framework required and combination of machine learning, artificial intelligence and simulation modelling capability that will allow near real-time prediction and assessment of infrastructure from asset to system scale during emergent extreme events.
The PhD will build upon work undertaken during the 10-years of research developed during the EPSRC funded Infrastructure Transitions Research Consortium (ITRC; 2010-2020) that led to the development of the UKs first dedicated National Infrastructure Modelling Database (NISMOD-DB++). The PhD will also utilise the capability developed during the NERC Flood-PREPARED project for the ingestion, integration and streaming of spatially distributed real-time sensor network data streams. Utilising and building upon these data manegment capabilities, the PhD will build upon the Bayesian network model developed at Newcastle as part of the Connected Places Catapult CReDO infrastructure systems digital twin. In particular, the PhD will look to investigate:
- How do we best integrate real-time data feeds with existing large scale asset databases in a timely and efficient manner for improved analytical real time situation awareness?
- How best can we capture and represent across large spatially interdependent infrastructure systems the real-time dynamics of asset vulnerability, exposure and adaptive capacity? and,
- What combination of machine learning, artificial intelligence and simulation modelling is appropriate to provide real-time infrastructure system scale prediction of emergent risk and response options to developing events?
During the MRes year the student will gain the key knowledge, understanding and skills required to initiate the design of the decision support capability and consider how best to address the research questions. The successful candidate will gain skills in big data management, including handling of real-time data streams. Via taught modules knowledge and understanding of AI and machine learning, including statistical machine learning approaches, will be acquired, along with an introduction to the spatial network simulation models that have been used to study and understand infrastructure systems. The MRes project will act as a foundation piece of work for the subsequent PhD. MRes project options include (i) preliminary work of ingesting and melding real-time data-feeds into the National Infrastructure Modelling Database capability, (ii) Investigating how the condition, vulnerability and exposure of different infrastructure assets can be captured, represented in a robust manner, or (iii) evaluating the ability to adapt and modify infrastructure network failure models for near real-time prediction.
Initial work on the PhD will focus on the development of the methods for the efficient and robust integration and melding of spatially distributed real-time data feeds with existing location-based asset data in order to facilitate subsequent near real-time prediction. This will be achieved by utilising and adapting elements of the National Infrastructure Modelling Database (NISMOD-DB++) and existing real-time data handling and management software tools. Thereafter, via a number of agreed case studies with Northumbrian Water Group, the near real-time analytical and modelling requirements will be investigated. It is envisaged that this will initially start and build upon the Bayesian network approach utilised within the CReDO digital twin project to model asset risk to emerging events in real-time. Via a series of cases studies, the analytical framework will be further extended to consider increasingly complex infrastructure system representations and extreme event scenarios (i.e., simultaneous spatially distributed failure or stressing of the system).
During the course of the PhD several research visits/secondments to Northumbrian Water Group will be arranged to ensure bi-directional translation and knowledge exchange. Opportunity exists for significant engagement with the Alan Turing Institute’s Digital Twin hub regarding the use of AI, machine learning and simulation modelling for the development of infrastructure digital twins.
Candidates with a minimum of a 2:1 degree in computing, engineering or statistics. Experience and knowledge of data science, including the use and application of AI and machine learning libraries/packages in a language such as python or R is desirable. Previous experience in the management and analysis of geospatial data via GIS would be desirable.
For further information, please contact Professor Stuart Barr
