Analysis of Compressor Cascade Deviation and Surrogate Model Construction Based on Retained Lift

The through-flow method still plays an important role in the design of modern aero-engine, and its accuracy depends on the loss and deviation model. The presence of tip clearance will impact the deviation distribution, while the retained lift is somewhat related to this effect. To achieve a more precise deviation model, this paper utilises the machine learning approach. The database comprises cascades with tip clearances in training, from which obtains a span-wise deviation model and executes its validation by comparing with experiment result. The database is obtained by calculating 16 different geometries of the cascades with tip clearance in different working conditions, introducing the geometrical parameters of the cascades and retained lift as feature engineering. The deviation and the retained lift follow the same trend with tip clearance size and operating conditions variation. Predict the span-wise distribution of the retained lift using the k-nearest neighbour regression, then combine with the traditional model to get the distribution of the deviation. The results show that the coefficient of determination of the retained lift coefficient prediction in the test set reaches 81.02%, and the mean absolute error is around 1.32%. Moreover, the trend predictions of cascade deviation distribution for different tip clearance size are all in good agreement with the experimental results. The coefficient of determination of the prediction with the simulation is 75.23%, and the mean absolute error is 1.74%.