Trust-Aware Predictive Emissions Monitoring for Gas Turbine Fleets with Limited Labelled Data
Researchers have developed a trust-aware probabilistic framework to improve emissions prediction for gas turbine fleets, particularly when labeled data is scarce. The system combines multiple machine learning models with confidence estimation and uncertainty quantification to generate reliability scores for predictions on unlabeled turbines. This approach significantly reduces prediction errors, with the highest-confidence predictions showing a substantial drop in Mean Absolute Error, indicating its potential for more trustworthy industrial deployments. AI
IMPACT Enhances the reliability of AI-driven predictive maintenance and monitoring in industrial settings.