Design and optimization of 1 × 8 PLC splitter with backpropagation neural networks and genetic algorithm

Abstract

Abstract The planar lightwave circuit (PLC) splitter serves as a crucial component in fiber-to-the-home. Diverse applications in complex scenarios necessitate a variety of PLC splitters, thereby escalating the complexity of their design. A novel optimization method based on orthogonal experimental design integrating a backpropagation neural network with a genetic algorithm is proposed. The optimization results demonstrate that for the 1 × 8 PLC splitter example, the insertion loss (IL), wavelength-dependent loss (WDL), and uniformity loss (UL) of the designed device are 9.2202 dB, 0.0915 dB, and 0.0914 dB, respectively. Compared to the previous results obtained using a single-parameter optimization method, which yielded an IL, WDL, and UL of 9.5 dB, 0.35 dB, and 0.30 dB respectively, there is a significant performance improvement.