Optical Module Performance Testing Method
The failure of the optical module is mainly caused by the deterioration of the performance of the transmitting laser. Aiming at the implementation method of optical module performance monitoring, this paper proposes a data-driven implementation method for optical module performance prediction and maintenance. Through data annotation/feature engineering/K-Means clustering/GBDT model, the performance status of optical modules can be effectively predicted. Compared with the threshold-based implementation method, the module performance can be predicted in advance, and early warning information can be generated in time to remind timely maintenance.
Solutions for Optical Module Performance Maintenance
The implementation method of data-driven performance prediction and maintenance of optical modules is to effectively predict the performance status of optical modules through data annotation/feature engineering/K-Means clustering/GBDT model. Predict the performance of optical modules, and generate early warning information in time to prompt system maintenance.
For large data centers/5G networks, hundreds of thousands of optical modules, such as SFP+, QSFP+, and QSFP28 optical modules, and the minute-level data collection frequency will generate massive DDM information data sets. In the two cases, an automatic degradation point detection algorithm is proposed, which automatically labels “T” samples and “F” samples according to the abnormality of the bias current Tx_Bias and the bias voltage Vcc, avoiding the need for manual data labeling during the data labeling process.
In the network topology, the device ports are interconnected, and the optical modules will appear in pairs. The transmitting end of the optical module at the local end and the receiving end of the optical module at the opposite end are connected by optical fibers. The DDM information set selects the local end (Temp, Vcc, Tx_Bias, Tx_Power)+ Analysis is performed on the end (Rx_Power) data. Aiming at the problems of large data volume and low data dimension of DDM information data, feature engineering technology is applied to expand the data dimension. There are far more “T” samples than “F” samples in the automatically labeled data set in a certain period of time, and the “T” sample data set is clustered by using the k-means method, so that the “T” samples and the “F” sample set tend to be closer to each other. It is in balance, which is convenient for model training and improves the training accuracy.
By comparing the commonly used logistic regression, gradient boosting decision tree and artificial neural network models, from the dimensions of algorithm complexity and model accuracy, the gradient boosting decision tree model is selected to predict whether the optical module fails. Compared with the existing threshold-based judgment rules, the artificial intelligence method can predict the failure of the optical module in advance and prompt the system for maintenance.
Effects on Communication Networks
Effective prediction of optical module performance in advance
Establishing a data analysis model between DDM information and performance degradation of optical modules can effectively predict the performance status of optical modules. Compared with the rules based on thresholds, it can predict the performance of optical modules in advance and generate early warning information in time.
Realize the correlation analysis of equipment transceiver modules
The correlation analysis of the optical modules of the sending port/receiving port between the devices in the network is established, and the performance status of the optical module at the local/peer end is monitored at the same time.
Efficiently improve communication system operation and maintenance
It provides a non-intrusive data-driven online optical module performance trend analysis method for large-scale optical module application scenarios such as 5G and data centers, which can effectively improve the intelligent operation and maintenance level of the communication system and ensure the normal operation of the business.