Scholarlink

<p>Estimating how long 1550&#xa0;nm DFB lasers will last is crucial for maintaining the reliability of optical communication networks. Traditionally, this is done using accelerated aging tests and physical models like the Arrhenius equation. In recent times, machine learning has become popular for predicting device failure, but most approaches either depend entirely on physics or entirely on data. In this work, we combine both. First, we calculate the Mean Time To Failure (MTTF) using the classical Arrhenius method, based on parameters like threshold current, junction temperature, slope efficiency, and optical power. Then, we apply machine learning models like Linear Regression, Random Forest, and Gradient Boosting to learn the gap between the estimated and actual MTTF values. Random Forest outperformed both Linear Regression and Gradient Boosting, demonstrating the lowest prediction error and the highest explanatory power across all evaluation metrics. This hybrid method blends theory with data, offering a more reliable and explainable way to predict laser lifetime in photonic systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid lifetime prediction of DFB lasers using Arrhenius theory and machine learning

  • 期刊论文

  • 发表日期：

    2025年12月29日

查看全文
  • Lakshmi G,
  • Brindha S

摘要

Estimating how long 1550 nm DFB lasers will last is crucial for maintaining the reliability of optical communication networks. Traditionally, this is done using accelerated aging tests and physical models like the Arrhenius equation. In recent times, machine learning has become popular for predicting device failure, but most approaches either depend entirely on physics or entirely on data. In this work, we combine both. First, we calculate the Mean Time To Failure (MTTF) using the classical Arrhenius method, based on parameters like threshold current, junction temperature, slope efficiency, and optical power. Then, we apply machine learning models like Linear Regression, Random Forest, and Gradient Boosting to learn the gap between the estimated and actual MTTF values. Random Forest outperformed both Linear Regression and Gradient Boosting, demonstrating the lowest prediction error and the highest explanatory power across all evaluation metrics. This hybrid method blends theory with data, offering a more reliable and explainable way to predict laser lifetime in photonic systems.