ir. Stefan de Gijsel wins Kooy prize 2022

At the end of 2021, the Kooy Prize jury received seven entries with nominations for the Kooy Prize 2022. At the Kooy Symposium on 15 April in Stroe, it was officially announced that the €2,000 prize had been won by ir Stefan de Gijsel. De Gijsel graduated cum laude from TU Delft at the Faculty of Electrical Engineering, Mathematics and Information and wrote his report 'Data-driven methods for magnetic signatures' here.

The Defence&Security Department of the Royal Institute of Engineers (KIVI), annually awards a prize for the best graduation thesis in a technology relevant to Defence and Security (HBO or TU). This year, it was finally possible to do so again 'live' during the Kooy Symposium in Stroe.

Graduation

The winner - Stefan de Gijsel - carried out his research at TNO and was supervised from the TU Department of Mathematical Physics. Within his graduation research, he applied Data-Science techniques to the research area of magnetic ship signatures. In doing so, he worked on two applications:

1. locating a magnetic dipole and
2. translating magnetic measurements from a marine platform above water into a magnetic underwater signature.

Based on his research, Stefan wrote a paper, "A Compressed Sensing Algorithm for Magnetic Dipole Localisation", which was submitted to the IEEE Sensors Journal in early 2022. At TNO, where he joined after graduation, Stefan's ideas were immediately put into practice during a trial. Defence is very interested in the results of this operationally relevant research.

Pleasantly surprised

Stefan himself is proud and honoured to have won the Kooy prize. "To be honest, it was also a surprise, as my supervisors had secretly entered me! I would therefore like to thank Aad, Reinier and Eugene enormously, without them I would never have been able to achieve these great results. I don't yet know exactly what I will do with the prize. I want to donate part of it to a charity for emergency aid in Ukraine. In addition, my supervisors and I are also going to celebrate winning this prize together."

Jury

The jury consisted of Dr L. Koene (Jury chairman and board member KIVI DV, NLDA/FMW), Ing L.F. Galle (Ministry of Defence, Materiel Directorate), Prof.dr.ir. P.H.A.J.M. van Gelder (TU Delft), Prof. dr. B.R.H.M. Haverkort (University of Twente), Prof.dr.ir. J.M.C. Mol (TU Delft), Prof Dr F. Phillipson (TNO, University of Maastricht), Prof Dr F.E. van Vliet (TNO, University of Twente) and Prof Dr P.J. Oonincx (NLDA/FMW).

RESEARCH

Stefan de Gijsel applied Data-Science techniques to the research area of magnetic ship signatures. Stefan worked on two applications, namely locating a magnetic dipole and translating magnetic measurements from a marine platform above water into a magnetic underwater signature.

The first topic involves locating a magnetic dipole based on magnetic measurements. This is the core of Magnetic Anomaly Detection (MAD). Algorithms that can do this well are necessary to determine the detectability, and hence vulnerability, of our own ships. Stefan dived deep into this problem and developed a robust algorithm that he demonstrated in real-time at the magnetic scale model facility at TNO.

For the localisation problem, an algorithm is proposed to localise a magnetic dipole with a limited number of noise measurements from a sensor array forming a horizontal grid. The algorithm is based on the theory of compressed sensing and exploits the sparsity of magnetic dipole fields in the location domain. Based on the resulting sparse representation, it produces a classification consisting of estimates of the location and both the magnitude and direction of the magnetic moment. The possibility of performing iterations has been explored, with the basis and chosen sensors improving after each location estimate. Results from both simulations and experiments, show that the algorithm is effective for locating magnetic dipoles.

The second topic is translating magnetic measurements from a surface vessel above water to an underwater magnetic signature, especially from ships. This topic has major implications. Currently, a ship still has to go to a special measuring orbit to establish its magnetic signature, but in the future we foresee that such a measurement using a drone can take place anywhere, provided an algorithm can correctly translate the data from abovewater to underwater, where the mine threat manifests itself. For this translation task, Stefan investigated two methods with nice results.

To solve the problem of translating signatures, Stefan investigated two approaches. One algorithm was developed using the so-called Gappy POD technique and one using neural networks. With the Gappy POD technique -which stands for Proper Orthogonal Decomposition, based on incomplete data- a complete signal can be reconstructed with just a few measurements (a gappy measurement). This method is adapted to also translate a field above a ship to a field below a ship. Second, different neural networks are created, trained using data from above and below a ship. It is proven that both the Gappy POD-based algorithm and linear neural networks give good estimates of magnetic signatures. Results improve when more so-called base modes are included and when more sensors are used for measurements.

Based on this research, Stefan wrote a paper, "A Compressed Sensing Algorithm for Magnetic Dipole Localisation", which was submitted to the IEEE Sensors Journal in early 2022. At TNO, where he joined after graduation, Stefan's ideas were immediately put into practice during a trial. Defence has a keen interest in the results of this operationally relevant research.