Implementation of optical module performance prediction and maintenance on data-driven

Dongmei Liu,Yongjun Yang,Zhifei Tang,Zheng He
DOI: https://doi.org/10.1117/12.2627880
2022-03-27
Abstract:The SFF.8472 protocol defines five key DDM attributes: operating temperature (Temp), operating voltage (Vcc), bias current (Tx_Bias), receiving power (Rx_Power), and transmitting power (Tx_Power). All the five DDM attributes are related to the degradation degree and expected remaining life of the optical module. Therefore, DDM can be used for diagnosis and early warning of optical module failure. The optical power feedback control unit inside the optical module can control the Tx_Power at a stable level, but with the aging of the laser, the quantum efficiency of the laser will decrease. Power control is achieved by improving the Tx_Bias of the laser, so we can predict its life by monitoring the Tx_Bias of the laser. Meanwhile, the Tx_Bias of the laser is closely related to Temp and Vcc, and the influence of Temp and Vcc needs to be considered when setting the Tx_Bias limit. Optical module failure is mainly caused by theperformance degradation of the transmitting laser. The data-driven optical module performance prediction and maintenance is implemented based on data labeling/feature engineering/K Means clustering/GBDT model, effectively predicting the optical module performance status. Compared with the threshold-based implementation method, the data-driven method can predict module performance in advance and generate timely warning messages to remind timely maintenance.
What problem does this paper attempt to address?