Neural network model of optical fiber direction coupler design

Abstract

The coupling length and coupling ratio is defined as input and output respectively, which are used to train the neural network model. Radial base function neural network model is built for optical fiber direction coupler, which is in turn simulated and designed through the model. This method has the advantages of speediness, accuracy and reliability, which are identified by the design example. It can be used to design the other kind of optical passive device. This way is a novel design approach, which can save the cost of device design, decrease period of design and has a good prospect of application.

Artificial Neural Networks for the Chromatic Dispersion Prediction of Photonic Crystal Fibers

“…On the other hand, ANNs have been used in photonics for the modeling of planar waveguides based couplers and optical fiber based couplers and also for the analysis of PCFs and patch antennas . The main advantages of an ANN based models are their simplicity, the reduced time, and computational effort and also its application for synthesis problems.…”

Aplicação de Redes Neurais Artificiais em Conjunto com o Método Vetorial da Propagação de Feixes na Análise de um Acoplador Direcional Baseado em Fibra Ótica

“…Essa habilidade permite, sem a necessidade de conhecer o objeto investigado, modelar um sistema conhecendo apenas os valores de entrada e saída. Tal funcionalidade é usada na modelagem de acopladores de fibras óticas [5] e também para análise e síntese de antenas de microfita na área de microondas [6] e [7]. A maior vantagem do uso de modelos baseados em redes neurais é a sua simplicidade, o pouco tempo e esforço computacional requeridos após a fase de treinamento, além da sua aplicação em problemas de síntese.…”

Efficient Neural Network Modeling of Photonic Crystal Fiber Chromatic Dispersion

“…On the other hand, neural networks have been used in photonics for the modeling of planar couplers [6] and optical fiber couplers [6,7] and photonic crystal fibers [8]. The main advantages of a neural network based models are their simplicity, the reduced time and computational effort and also their application for synthesis problems.…”