Ellie completed a BSc in Environmental Sciences from the University of Leeds in 2020, gaining an understanding and appreciation for the complexity of environmental, social, and technological systems. From this, an overarching goal of climate change mitigation led her to complete an MSc in Energy Systems and Data Analytics at UCL, within which the connectivity of transport, built environment, and social and political sectors were explored using machine learning techniques on big data, to identify plausible solutions for decarbonisation. Projects completed included investigating the driving factors of micromobility demand, and carrying out an uncertainty analysis on global net zero pathways using clustering for her thesis. Outside of her university work Ellie was also involved in additional research roles, including looking into public responses to carbon taxes, and how demographics can impact this, and also the applicability of digital platforms to inform and engage the public in local policy changes.
It is due to the dynamism of the transportation sector, and the challenge that it poses to reaching net zero targets by 2050, that Ellie chose to pursue this area of study. Her specific research interests revolve around sustainability and efficiency; by utilising machine learning techniques she hopes to identify novel approaches that unlock the potential of technological innovation, policy measures, and behavioural change for large scale decarbonisation.
Physical testing of new vehicle components is time consuming and complex, and is thus, highly energy intensive. Additionally, many of these tests develop faults which are only discovered after the test-run, and subsequently need to be repeated. Virtual testing can reduce the need for physical tests and reduce energy consumption, however, these models rely on physical testing for high quality test data to adapt and optimise the virtual models. Some physical testing will also still be required prior to market release to ensure customer safety. Ellie's project aims to improve the quality of data from physical testing, using machine learning and human factors analysis, whilst minimising wasted energy consumption.
Anomaly prevention – a term coined to complement anomaly detection - describes the combined methodology of human factors analysis and systems mapping; this was devised after testing procedure and human factors were identified as significant contributors to data quality. Anomaly prevention will identify the processes outside of the test bed that negatively impact data quality and suggest mitigations. Anomaly detection – a method of finding unexpected patterns or points in data via statistical or machine learning techniques - presents a solution to detect anomalies in real time during testing, thus, minimising physical testbed time whilst increasing the reliability of data to feed into virtual models. Together, these methods have the potential to reduce the energy intensity of testing and development, whilst simultaneously increasing the speed at which low-carbon technologies can be released into the public domain to aid large scale decarbonisation of the transport sector.
© Copyright 2024 AAPS CDT, Centre for Doctoral Training in Advanced Automotive Propulsion Systems at the University of Bath
