Modal amplitude and phase estimation of multimode near field patterns based on artificial neural network with the help of grey-wolf-optimizer

• Published in: Optical fiber technology 2021-12, Vol.67, p.102720, Article 102720
• Main Authors: Sugawara, Naoto, Fujisawa, Takeshi, Nakamura, Kodai, Sawada, Yusuke, Mori, Takayoshi, Sakamoto, Taiji, Imada, Ryota, Matsui, Takashi, Nakajima, Kazuhide, Saitoh, Kunimasa
• Format: Article
• Language: English
• Subjects: Artificial neural network; Modal decomposition technique; Mode division multiplexing; Mode scrambler
• ISSN: 1068-5200; 1095-9912
• DOI: 10.1016/j.yofte.2021.102720

•A novel method for modal decomposition using neural network and gray-wolf optimizer is proposed.•With the help of meta-heuristic optimizer, the accuracy of the decomposition is increased.•The method is used to decompose the NFPs of mode scramblers and the correlation value is improved. A simple and efficient method for estimating modal amplitude and phase of multimode near field patterns (NFPs) based on artificial-neural-network (ANN) with the help of the optimization method is proposed. The inferred amplitude and phase of measured NFPs based on ANN are refined by using a grey-wolf optimizer (GWO). By using the proposed method, the image correlation between reproduced and measured NFPs is improved without re-training of ANN, which is the most time-consuming part of ANN-based numerical modal decomposition technique. Numerical examples of three and six mode cases are presented for the estimation using simple ANN. For six-mode case, the correlation is greatly improved by using the optimizer. Finally, the estimation of the measured NFPs from three-mode exchanger and six-mode mode conversion grating is implemented, and 5% improvement in the correlation value is observed for six-mode case. The proposed method offers alternative way to improve the correlation without using elaborated ANN.