In this article, we profile Jonathan Sadeghi, who is studying a PhD in Engineering at the Institute for Risk and Uncertainty, University of Liverpool, focused on conducting research in Next Generation Nuclear.
What is your area of study?
I’m an Engineering PhD Student, at the University of Liverpool, based in the Institute for Risk and Uncertainty. My PhD is funded through the EPSRC Centre for Doctoral Training in Nuclear Fission- Next Generation Nuclear (NGN CDT), which focuses on developing future research to support the UK's strategic nuclear programmes including nuclear legacy clean-up, new build power stations and, defence and security. My studies focus on conducting research intersecting the disciplines of Machine Learning, Nuclear Engineering, Software Engineering and Uncertainty Quantification.
Why did you choose this subject for your PhD?
I saw an advert for a PhD with the Nuclear Centre for Doctoral Training, which includes time working on an industry project. I was studying physics at the time, and the thought of working in the nuclear sector really appealed. The project being advertised by the Institute of Risk and Uncertainty sounded very interesting, involving probability and statistics in computing, collaborating with Wood Group (formerly Amec Foster-Wheeler), an engineering consultancy for the nuclear sector. It’s a very different career path compared to physics but I already had a strong interest in probability and statistics.
What’s your background?
I studied for my undergraduate degree, an MPhys in Physics with Theoretical Physics, at the University of Manchester which took four years, and then a PGDip in Nuclear Science and Engineering. Going from theoretical, to collaborative and industry focused studies has been a big change but I really enjoy working with industry, and at the Institute for Risk and Uncertainty.
Risk and uncertainty is a growing area within industry, what is your interest?
It’s really important to focus on the decision making, because I think the essence of decision making is uncertainty. If we know what’s going to happen, decision making is easy. If we don’t know what’s going to happen, the only way to make a sound decision is to understand your uncertainty. For example, designing an aircraft and having distinct teams focus separately on the physics and the uncertainty doesn’t make a lot of sense scientifically. For me, going from a theoretical physics background to working more with uncertainty makes sense, and is one of the great things about being multidisciplinary, which is how we work at Liverpool.
What will be the impact of risk and uncertainty on the engineering sector?
I think there’s a really interesting intersection between uncertainty quantification and machine learning (another area of interest for me) or maybe more generally Artificial Intelligence (AI). Applying AI techniques to engineering challenges is a big growth area which hasn't been exploited as yet, not to mention adding in uncertainty and understanding the impact of this. With machine learning research taking place in industry, I think we’ll see a lot more uncertainty quantification integrated into the design process. My PhD focuses more on probabilistic safety analysis and I think machine learning techniques have a role to play here, helping to improve safety in industry but this is not a new idea, people have been using more simplistic models for probably 20 or 30 years!
Why did you choose to study Physics? Have you always had a scientific interest?
I’ve always been interested in maths and found it fun to solve problems in that way, and I’ve always quite enjoyed programming, making games when I was younger. One of the great things about physics (and this applies to uncertainty quantification as well) is you get a good balance of applying mathematics to problems, applying computational techniques, applying physical concepts to problems. The interplay of those different skills is what makes the area a lot of fun to work in. It’s constantly changing, which obviously makes life a lot more interesting, and you also have the whole problem solving aspect of things, which is enjoyable when you solve a problem and everything goes to plan. I think if you ask any of my colleagues, they would say something similar, as I think that's what draws people into the area.
What are your most notable projects, awards, and areas of study?
One of the projects we’ve been working on at Liverpool is Open Cossan, a generalised software tool for uncertainty quantification, and where I've developed some of the code. My main contribution was what we call the interval predictor model toolbox, which is a really robust machine learning toolbox in our code that engineers can use. The great thing is you don't have to be an expert in uncertainty quantification. You can be an engineer, and have a model set up to do CFD for example, and use the software and run your analysis in less than a day. We’ve got a few interesting published papers based on this, including one in Nuclear Engineering and Design where we analysed the structural reliability of prestressed concrete containments.
I also participated in the Enhanced Fidelity Transonic (EFT) Wing Project, led by Airbus. Here, I worked on potentially using machine learning techniques to enable data science analysis on new data coming from some of the computer clusters. The overall objective of the project was to significantly enhance the performance assessment fidelity of transonic wings, thereby reducing risk and uncertainty in the aircraft design process, and enabling aircraft to be driven to higher performance standards.
Published research and conference papers
Cossan Software: A Multidisciplinary and Collaborative Software for Uncertainty Quantification, Edoardo Patelli, Matteo Broggi, Silvia Tolo and Jonathan Sadeghi, Proceedings of the 2nd ECCOMAS Thematic Conference on Uncertainty Quantification in Computational Sciences and Engineering.
History Matching with Robust Predictive Metamodels, Jonathan Sadeghi and Edoardo Patelli, 12th International Conference on Structural Safety & Reliability.
Structural reliability of pre-stressed concrete containments, Nawal K. Prinja, Azeezat Ogunbadejo, Jonathan Sadeghi, and Edoardo Patelli, Nuclear Engineering and Design.
Robust propagation of probability boxes by Interval Predictor Models, Jonathan Sadeghi, Marco de Angelis and Edoardo Patelli. Proceedings of the joint ICVRAM ISUMA UNCERTAINTIES conference, Florianopolis, SC, Brazil, April 2018.
What are you looking forward to in your area of expertise?
One of the things I am really interested in is the scenario technique to chance constrained optimisation. Essentially it allows you to solve optimisation problems that are subject to uncertainty. It could be that you have a particular design and you don't know what type of situations it's going to run up against, or you might have a model and you don't really know what the data is going to look like and what you are going to see in the future, but want to make sure your model is accurate. At Liverpool, like many other institutions, we are doing work in this area, and hopefully we will soon see this technique moving into industry, which would be exciting.
With regards to uncertainty quantification, a really cool project in Liverpool is the digital reactor design project, involving the creation of a digital twin of a nuclear reactor. If you couple that with design optimisation and uncertainty quantification, essentially you can design a reactor and know how it's going to work in advance. When this comes to fruition, it will be a game changer.
Who do you most aspire to and why?
It’s hard to say really, there’s a lot of people who I have a lot of respect for in this area – there are people who have changed their field. A good example in the UK are the Bayesian statisticians, who have completely changed the way we do things in our area.
There are recognised individuals like Tony O’Hagan, David McKay, who have carried out really interesting research that’s changed things, and Scott Ferson who invented probability boxes. Every scientist would love to say “I am going to change things as much as these people have”, and that’s got to be the dream for every PhD student.
What are your career aspirations?
After my PhD, I would really like to get stuck into some industry challenges, applying some of the techniques I have studied to engineering problems. I am really passionate about taking the research we do, and applying this to industrial codes for use in software engineering. Possibly, this could mean applying machine learning to engineering problems. However, I’m still interested in research, and seeing that many big contributors in this field have regularly moved between academic research and industry has inspired me to do the same – it seems like a great way to exchange ideas.