Like many areas of technology, artificial intelligence and deep learning are making their way into the field of laser materials processing, bringing with it benefits such as advanced, automated process monitoring that enables laser parameters to be adjusted ‘on the fly’. This unlocks capabilities such as ‘closed-loop’ processing, which by ensuring an optimal, stable laser process, can dramatically reduce the number of defects that occur in finished parts.
This arrival of artificial intelligence and deep learning in laser processing can be attributed to the exceptional work of scientists such as Dr Ben Mills, a Principal Research Fellow at the University of Southampton’s Optoelectronics Research Centre.
“My research is primarily focussed on the application of deep learning for the optimisation, modelling and real-time control of laser machining using ultrafast femtosecond lasers,” Mills explained. “In the past year, my research direction has also included the application of deep learning for control and optimisation of the coherent beam combination of fibre lasers, as well as the application of deep learning for a range of sensing and medical research fields.”
With deep learning having now been demonstrated across a range of laser materials processing applications in a research lab environment, the primary challenge now, according to Mills, is to translate these recent breakthroughs into tangible efficiency and productivity enhancements in the industrial context. ‘This will require collaboration between academia and industry, and the creation of large-scale data sets for training and validating the neural networks,’ he said.
While no doubt being an incredibly rewarding and fulfilling career, working with such cutting-edge technologies doesn’t come without their own, fairly unique challenges that need to be overcome.
“Back in 2017, when my team first started applying deep learning to laser materials processing, the initial results were so surprising and unconventional that we were unable to progress our papers through peer review. In fact, it took nearly two years to get our first set of results published,” said Mills. “Despite these setbacks, we kept developing our techniques and resubmitting to journals, and we finally had our breakthrough moment a few years later when we started to receive worldwide recognition for our work.”
Many of the techniques presented by Mills and his colleagues in those first papers are now standard techniques across deep learning in photonics. “Today there are many journals that encourage research manuscripts demonstrating deep learning in photonics, due to the recent increase in the number of academics working in applied deep learning,” he remarked.
Mills assured that overcoming the barriers faced in his career would not have been nearly as possible without the guidance of several mentors and role models, with a particular tip of the hat to Professor Robert Eason, an ORC colleague with whom he has worked for more than a decade. “It is rare, but vitally important, to have someone who you know has the time and the desire to push you forward in your career,” Mills remarked. “In addition, I must mention the many contacts I have made in the UK and EU laser community and industry, who have expanded my horizons beyond the research laboratory and into real-world applications.”
Also key to Mills’ success as a researcher has of course been the UK funding provided by the UKRI and EPSRC, “...without whom I would never have been able to become an independent research lead,” he commented.
And it is gaining this independence that Mills believes is the most important advice that can be given to those starting their own research careers. “Try and become scientifically independent as soon as possible,” he affirmed. “To do this, you need to develop your own research ideas, and obtain your own funding to pursue them. Whilst there are many opportunities, competition is extremely tough, but then again, success rate is always zero if you don’t apply! Striking out on your own is vital, but it can be challenging for sure.”
Dr Mills can be found online at www.linkedin.com/in/ben-mills-31953b45
He plans to attend the SPIE Photonics West and CLEO conferences next year, where he will present on the application of deep learning for fibre lasers and laser machining.
Job title: Principal Research Fellow
Organisation: Optoelectronics Research Centre, University of Southampton
Based in: Southampton, UK
Education: PhD in Photonics (2005-2009) at the Optoelectronics Research Centre. BSc Physics (Hons) (2002-2005) University of Southampton.