Enhancing gas turbine thrust estimation performance under varied uncertainties using a simplified comprehensive dataset

As the in-flight thrust of gas turbine engines (GTE) cannot be measured, research has focused on developing estimators that function as data-driven thrust sensors. Current studies do not adequately consider the impacts of uncertainty factors. Addressing this issue, this study proposes to develop thrust estimators using a simplified comprehensive dataset. The dataset is generated based on the categorization and quantification of uncertainty factors. The innovations include: 1) The mechanism of dataset construction is analyzed, and the physically comprehensive thrust dataset (PCTD) with fewer samples is generated; 2) The sample capture learning strategy (SCLS) is designed, which generates a streamlined training set for thrust estimation. Both approaches avoid excessive sample numbers, which lead to increased time and computational costs. Results show that thrust estimators developed by the PCTD outperforms those built using general datasets. The average and maximum relative error under simulated real conditions are only 0.0519% and 0.6494%. Moreover, using the SCLS improves accuracy across the full envelope, especially the maximum error, which is reduced by 47.6% on average.