Machine Learning based Laser Failure Mode Detection

03/19/2022

∙

by   Khouloud Abdelli, et al.

∙

0

∙

share

Laser degradation analysis is a crucial process for the enhancement of laser reliability. Here, we propose a data-driven fault detection approach based on Long Short-Term Memory (LSTM) recurrent neural networks to detect the different laser degradation modes based on synthetic historical failure data. In comparison to typical threshold-based systems, attaining 24.41 accuracy, the LSTM-based model achieves 95.52 classical machine learning (ML) models namely Random Forest (RF), K-Nearest Neighbours (KNN) and Logistic Regression (LR).