Rapid Visualization of Compressor Blade Finite Element Models Using Surrogate Modeling

Abstract

The design process for compressor blades is a highly iterative and often slow process. This research applied and measured the impact of using surrogates to quickly model the stresses on a compressor blade. By modeling distinct points on a finite element (FE) model with unique surrogates, the stress field of the entire FE model was quickly predicted. This required that the distinct points remain in the same relative location on each blade used in training the surrogate. This research studied the ability of mesh morphing, and using the surface nodes as those distinct points, to satisfy this requirement. The results show that mesh morphing performed well on the tested compressor blades. The research also found that the surrogate accuracy depended not only on the number of training samples, but also the number and types of parameters being emulated. The surrogate models achieved less than 5% error on all the tested blades. Finally, the method provided a 96% decrease in time required for a structural iteration of a compressor blade. Such speeds eliminate bottlenecks that may occur in the structural design process. The combination of mesh morphing and surrogate modeling in compressor blade analysis enables exploration of various geometric parameters and their effect on structural responses. Application of this process would produce a more thoroughly refined and understood compressor blade design.