Work

Time to failure prediction for MLCCs: A machine learning approach based on leakage current data

Year: 2025

Type: article

Abstract: The development of modern electronics poses challenges to the long-term reliability of multi-layer ceramic capacitors (MLCCs), especially under extreme conditions. While existing research mainly focuses on predicting the mean time to failure for MLCC populations, studies on predicting the time to failure (TTF) for individual capacitors are scarce. This study addresses this gap by developing a machine learning model to predict the TTF of individual MLCCs. Based on features extracted from leakage current curves and failure data generated via highly accelerated life testing, feature engineering and model optimization were applied to select a random forest model, achieving an R2 of 0.8971. This approach enables real-time monitoring and prediction of individual capacitor failures, offering valuable insights for improving reliability, particularly in the predictive maintenance of critical systems.

Source: APL Machine Learning

DOI: 10.1063/5.0262717