← All industry projectsOLED degradation modeling
LG DISPLAY · 2024–2025

OLED degradation modeling

Predictive models and interpretable compensation parameters

I developed machine-learning models of OLED panel aging and generated parameter maps for residual-image compensation. Explainability analysis helped identify the factors driving the predictions.

IndustryLG Display · 2024–2025Applied ML
Explore the approach ↓
01 / THE PROBLEM

How can predicted degradation become actionable compensation data?

Residual-image compensation needs information about how an OLED panel ages. A prediction model is useful only if its output can inform that compensation process. This project connected learned degradation estimates with parameter maps while examining which factors were driving the predictions.

02 / THE APPROACH

Connect prediction to compensation parameters

01

Learn panel degradation behavior

Develop machine-learning models to predict OLED panel degradation. The target is the panel’s aging behavior, which supplies information for the compensation system.

02

Generate spatial parameter maps

Translate predicted degradation into parameter maps for residual-image compensation. This gives the downstream compensation process an actionable representation of the model output.

03

Examine the factors behind predictions

Use explainable-AI techniques to analyze the factors contributing to degradation predictions and inform compensation modeling.

03 / MY CONTRIBUTION

Prediction models and interpretability analysis

Built degradation-prediction models and parameter maps and used explainable-AI techniques to examine the factors behind the predictions.

04 / OUTCOME & SCOPE

Models and maps for residual-image compensation

I developed the degradation models and compensation parameter maps, then used explainability analysis to examine the factors behind their predictions. The work linked a learned model to an engineering output that could feed the panel-compensation workflow.