Session: 30-05 Modeling, algorithms and computational techniques II

Paper Number: 83402

83402 - Gradient Enhanced Kriging Using Modal Sensitivity Approximations in a Reduced Basis Space for As-Manufactured Airfoil Analysis 

This work develops a new approach to efficiently predict the effect of as-manufactured geometry variations on airfoil modal response. A Gradient Enhanced Kriging modeling approach is formulated that uses the frequency, mode shape, and eigensensitivity results of a training data set in order to reduce the total number of training models required. Computational efficiency of this approach is further enabled through incorporation of a modal superposition approach for eigenvalue and eigenvector sensitivity calculations. This approach avoids the need to recompute full eigensolutions in a conventional finite difference sensitivity calculations approach.  Efficiency in mode shape approximation is enabled through transformation of displacement variation into a principal component analysis based reduced-basis space. This reduces the required number of surrogate models by three orders of magnitude. A novel approach is developed to perform gradient based emulation in this space that projects the principal component modes onto first Taylor series expansions of the physical mode shape variation at each training data site. This provides the gradients of the mode shape variation principal component participation. Results show that gradient-based approximations of these participations lead to an improvement in accuracy and efficiency compared to non-gradient based methods.  The process is demonstrated on the Parametric Blade Study Rotor Four.

Presenting Author: Jeffrey Brown US Air Force Research Laboratory

Presenting Author Biography: Dr. Jeff Brown is a Principal Engineer in the Engine Integrity Branch of the AFRL Turbine Engine Division. His research efforts are focused on implementing Physics-based Digital Engineering solutions for a variety of customers.

Authors:

Jeffrey Brown US Air Force Research Laboratory
Emily Carper AFRL
Dan Gillaugh AFRL
Alex Kaszynski AFRL
Joseph Beck AFRL