A Study on Learning Parameters in Application of Radial Basis Function Neural Network Model to Rotor Blade Design Approximation

Meta-model sre generally applied to approximate multi-objective optimization, reliability analysis, based design etc., not only in order improve the efficiencies of numerical calculation and convergence, but also facilitate analysis sensitivity. The radial basis function neural network (RBFNN) is meta-model employing hidden layer units output linear units, characterized by relatively fast training, generalization compact type networks. It important minimize some scattered noisy data space prevent local minima gradient optimization or using RBFNN. Since must be smoothed out for RBFNN as any actual structural problem, smoothing parameter properly determined. This study aims identify effect various learning parameters including spline on performance regarding approximation. An rotor blade problem was considered investigate characteristics approximation with respect range parameter, number training data, layers. In design, sensitivity such main effects were evaluated according variation parameters. From evaluation results it found that had larger influence accuracy than while layers little performances meta-model.