Hybrid Experimental-CFD Mapping Approach for Twin-Entry Turbo Turbine

The steady flow hot-gas stand test is a widely used method for experimentally characterising turbocharger turbines to produce maps for use in 1D engine simulations. However, for twin entry turbine stages with two volutes, measuring multiple maps at different ratios of mass flow in each volute is time-consuming. This study investigated how computational fluid dynamics (CFD) simulation could reduce the experimental effort for mapping twin-entry turbines, especially for unequal admission conditions. The study is based on a case study of a medium-duty twin-entry turbine, characterising its performance both experimentally and using 3D simulations with ANSYS CFX®. In total, nine maps were produced: one at equal admission, two single admissions, and six unequal admissions conditions. The unequal admission maps were recorded at constant pressure ratios between the two scrolls; the scroll pressure ratio varied from 0.58 to 1.75. Each map contains 24 data points, comprising four constant speedlines each containing six points at different expansion ratios. The comparison of CFD and experiments showed a good agreement with an average deviation below 5.3% and 1.5% for the total-to-static efficiency and the mass flow parameter respectively. Turbine performance maps are typically used in gas dynamic codes where the measured data is used to fit a mathematical interpolation and extrapolation model to predict turbine performance across the full range of speeds and expansion ratios. Here a state-of-the-art fitting algorithm was used to produce turbine models with different combinations of experimental and CFD data points to determine the trade-off between the number of points and the quality of the fitted model. In all cases, the results were compared to a reference model that used all nine experimental measured maps. The results showed that using five maps can yield relatively similar results to the full nine maps fit for both efficiency and mass flow, but reducing this further to three maps results in mean differences of 5% and 13% for the efficiency and mass flow parameter respectively. Using the five-map configuration, three different combinations of experimental and CFD-generated data were used to create a hybrid approach. The resultant fit was systematically compared to the benchmark nine-map experimental model, showing a trade-off (an average of 5%) between experimental and pure CFD-generated maps. The hybrid turbine model results showed an accuracy of less than 1% error compared with the pure CFD-generated model data. The study suggests that CFD simulation can be used to reduce the experimental effort for mapping twin-entry turbines, especially for unequal admission conditions, but the trade-off in accuracy must be considered.