Rapid Temperature Prediction of Turbine Guide Vanes Based on Multi-Source Heterogeneous Data Fusion

Turbine cooling design is challenged by the limited high-fidelity experimental data and the high computational cost of Computational Fluid Dynamics (CFD). A multi-source heterogeneous data fusion framework is proposed in the presented work to overcome the challenge based on a weighted Extreme Gradient Boosting (XGBoost) algorithm. The framework establishes a nonlinear mapping from geometric and operational parameters to the three-dimensional temperature field. By introducing adjustable sample weights, the model adaptively fuses sparse reference data with full-field simulation data. It is shown that by using a single data source, approximately 70 samples suffice to capture the dominant patterns within the design space, achieving predictions with a Mean Absolute Percentage Error (MAPE) below 0.5% in less than 0.5 seconds per field. Furthermore, it is revealed that sparse reference data can effectively correct systematic CFD biases through sample weight optimization, given a sufficient number of full-field data samples. It is resulted that the proposed framework provides an efficient and accurate solution for evaluating turbine cooling performance rapidly.