Phi-ML meets Engineering: Applying machine learning to automate interpretation of ultrasonic non-destructive evaluation (NDE) data
Thursday 22 Apr 2021
Time: 13:00 - 14:00
Bio
Paul D. Wilcox was born in Nottingham in 1971. He received an MEng degree in Engineering Science from the University of Oxford in 1994 and a PhD from Imperial College London in 1998. He remained in the Non-Destructive Testing (NDT) research group at Imperial College as a Research Associate until 2002, working on the development of guided wave array transducers for large area inspection. Since 2002, he has been with the Department of Mechanical Engineering at the University of Bristol where his current title is Professor of Dynamics. He held an EPSRC Advanced Research Fellowship in Quantitative Structural Health Monitoring from 2007 to 2012 and was Head of the Mechanical Engineering Department from 2015 to 2018. His research interests include array transducers, embedded sensors, ultrasonic particle manipulation, long-range guided wave inspection, structural health monitoring, elastodynamic scattering, data analysis and signal processing. In 2015 he was a co-founder of Inductosense Ltd., a spin-out company which is commercialising inductively-coupled embedded ultrasonic sensors.
Research interests
This pilot project will apply data science to exploit the full potential of quantitative Non-Destructive Evaluation (NDE) measurements of engineering assets, ranging from power stations to aeroplanes. Currently, the majority of NDE data is irreversibly condensed after a measurement is made, sometimes to as little as a binary pass/fail classification. In doing so, a wealth of useful information that could impact on the future safety and economic utility of an asset is lost forever. However, as the fourth industrial revolution unfolds, the enabling infrastructure to store raw NDE data from every inspection is becoming available in the form of scalable, cloud-based, asset-management platforms.
Permanent storage of raw NDE measurement data will become the norm, as it future-proofs historic measurements against new processing techniques as they become available. From a data science perspective this also presents a massive opportunity to extract much more information about the integrity of an engineering asset. NDE data from multiple measurement modalities can be integrated and analysed collectively in the context of historic measurements and data streams from other condition-monitoring modalities.
The specific objectives of the pilot project are to:
- Create the foundations of NDE data science by developing the mathematical taxonomy of ‘big’ NDE data to the point where the challenges can be clearly articulated to the wider data science community.
- Promote, through two workshops, the value of applying data science to NDE to both the industrial end-user community (e.g. Airbus, Rolls-Royce, EDF, Hitachi, BP) and data scientists.
- Develop one or more case studies with an industrial end-user that will demonstrate the value of NDE data science.
