Plane Cascade Aerodynamic Performance Prediction Based on Metric Learning for Multi-Output Gaussian Process Regression

Multi-output Gaussian process regression measures the similarity between samples based on Euclidean distance and assigns same weight to each feature. However, there are significant differences in aerodynamic performance of plane cascades composed symmetric asymmetric blade shapes, also geometry formed by different shapes experimental working conditions. There large geometric condition parameters features, which makes it difficult accurately measure when fewer samples. For this problem, a metric learning for multi-output method (ML_MOGPR) prediction cascade is proposed. It shares multiple output distributions during training input new embedding space reduce bias improve overall accuracy. analysis ML_MOGPR results, accuracy significantly improved compared with (MOGPR), backpropagation neural network (BPNN), multi-task (MTLNN). The results show that effective predicting cascade, can quickly make preliminary estimate meet parameter estimation requirements early stage.