Sensing with multimode optical fibres: A machine learning perspective

Abstract: 

In multimode optical fibres, the light confined in each mode has slightly different sensitivities to surrounding parameters, be it temperature, strain, or refractive index of the ambient environment (cladding). When used for sensing a specific parameter, multimode fibres can be viewed as the co-location of many slightly different sensors in both time and space. However, the cross-sensitivity from other parameters will lead to a different standard deviation for each sensing mode. Those parameter-specific estimations can be combined to produce a measurement with a much smaller variance by applying machine learning techniques on the entire multimode data.

I will discuss the principle behind the application of machine learning to multimode fibre sensing, with emphasis on sensing under very strong cross-perturbations and distributed sensing. The probabilistic nature of machine learning and its consequences on laboratory-based sensing experiments will be discussed. Perspectives on the future of using machine learning in conjunction with multimode fibre sensing in real-world applications will be presented.

Speaker bio:

Linh Nguyen is a research fellow with the Future Industries Institutes at UniSa. Before joining UniSA he has spent times at Edith Cowan University and the University of Adelaide. He has published more than 60 peer reviewed papers in international journals and conference proceedings attracting more than 2200 citations, h-index = 27 (Google citation). Research projects in which he has been involved with in major capacity (either a named fellow, a named chief investigator or a lead researcher) have totalled nearly 5M dollars to date. His research interest is in photonics and optical science in general, with a particular focus on optical sensing and fibre laser, and most recently light transport through multimode waveguide system and machine learning assisted photonics.