A synthetic laser reliability dataset generated using generative adversarial networks (GANs) is provided. The data includes normalized current measurements estimated at the following times: 2, 20, 40, 60, 80, 100, 150, 500, 1000, and 1500 hours. The data can be used to train machine learning models to solve different predictive maintenance tasks such as prediction of performance degradation, remainng useful prediction, and so on.
A monitoring data, which includes several OTDR traces incorporating various types of fiber events (e.g. reflective, non-reflective, merged events) induced along an optical fiber link, is provided. Different fiber faults such as fiber cut, and fiber bend are modeled using optical components such as connectors and variable optical attenuators (VOAs). The data can be used to train machine learning models for solving fiber fault diagnosis problems.
The monitored data is obtained using the optical time domain reflectometry (OTDR) principle, which is commonly used for troubleshooting fiber optic cables or links. The data set contains raw OTDR traces that include one or two reflective events caused by the placement of one or two reflectors and/or an open physical contact (PC) at the end of the monitored optical fiber link.
A synthetic data for low power (P ≤10 mW) InGaAsP MQW-DFB lasers operating at a wavelength (λ) ranging from 1.53 to 1.57 µm at a case temperature laying between -40 ℃ to 85 ℃ with side mode suppression ratio of more than 35 dB is generated and can be used for laser lifetime prediction using machine learning based approaches.
The dataset includes processed sequences of optical time domain reflectometry (OTDR) traces incorporating different types of fiber faults namely fiber cut, fiber eavesdropping (fiber tapping), dirty connector and bad splice. The dataset can be used for developping ML-based approaches for optical fiber fault detection, localization, idenification, and characterization.