Optical fiber dispersion compensation: supervised machine learning with regression approach

Abstract The performance of Optical Communication Systems is severely hampered by pulse broadening, which limits both coverage speed and distance. To compensate for pulse broadening effects both Fiber Bragg Gratings (FBGs) and Dispersion Compensating Fibers are widely used, evaluated, and documented. This work investigates a new method to mitigate the dispersion effect using a combination of DCF and a multi-stage FBGs. The proposed system includes multiple designs that use more than one stage of cascaded FBG with a DCF to increase the effectiveness of the FBG network. In the proposed system, the Quality factor (Q-factor) and Bit Error Rate (BER) results for each design are evaluated using a variety of FBG apodization functions. Different profiles show different Q-factors over a range of Continuous Wave laser power ranging from −20 to 20 dBm, for FBG grating lengths up to 70 mm. The proposed model demonstrates a significant cost improvement of 97.5% for DCF length and 98.9% highest Pulse Width Reduction Percentage, using the OptiSystem software. Regression modeling and Machine Learning methods are also utilized in FBG and DCF systems. Through these approaches, we are able to predict the actions and characteristics of the system, which aids in determining the estimate of the number of phases necessary to achieve the optimal aspect ratio for the Q-factor with minimal BER, in addition to reduce cost and time consumed. This leads to better system performance as compared to previously published attempts.