Machine Learning based Raman amplifier design

Student thesis: Master

Abstract

In this Master thesis, a machine learning framework used to design Raman amplifiers to create gain profiles in theC-band and the C+L-band is presented. The goal of the machine learning framework is to find the pump powers and wavelengths that result in a targeted gain profile. Simulations in the C-band and C+L-band resulted in maximum errors with a mean of 0.711and 0.868 dB and standard deviations of 0.449 and 0.414 dB.Fixing the wavelengths when employing 5 pumps in the C and the C+L-band resulted in a maximum error with a mean of 0.093dB and 0.109 dB and a standard deviation of 0.060 dB and 0.077dB. The framework can produce flat gain profiles for 2,3 and5 pumps in the C-band and for 4 and 6 pumps in the C+L-band. The framework is flexible and can be extended beyond theC+L-band to other bands.
Date of Award12 Aug 2020
Original languageEnglish
SupervisorChigo M. Okonkwo (Supervisor 1) & Menno van den Hout (Supervisor 2)

Keywords

  • Machine Learning
  • Optimisation
  • Raman Amplification
  • Ultra wide band Optical
  • Optical Communication
  • Space division multiplexing
  • Non-linear optical transmission

Cite this

'