Dynamic Stress Prediction in Centrifugal Compressor Blades Using Fluid Structure Interaction

Abstract

A computational model is developed that predicts stresses in the blades of a centrifugal compressor. The blade vibrations are caused by the wakes coming off stationary inlet guide vanes upstream of the impeller, which create a periodic excitation on the impeller blades. When this excitation frequency matches the resonant frequency of the impeller blades, resonant vibration is experienced. This vibration leads to high cycle fatigue, which is a leading cause of blade failure in turbomachinery. Although much research has been performed on axial flow turbomachinery, little has been published for radial machines such as centrifugal compressors and radial inflow turbines. A time domain coupled fluid-structure computational model is developed. The model couples the codes unidirectionally, where pressures are transferred to the structural code during the transient solution, and the fluid mesh remains unaffected by the structural displacements. A Fourier analysis is performed of the resulting strains to predict both amplitude and frequency content. This modeling method was first applied to a compressor in a single stage centrifugal compressor test rig. The analysis results were then validated by experimental blade strain measurements from a rotating test. The model correlated very well with the experimental results. In this work, a model is developed for a liquefied natural gas (LNG) centrifugal compressor that experienced repeated blade failures. The model determined stress levels in the blades, which helped to predict the likely cause of failure. The method was also used to investigate design changes to improve the robustness of the impeller design.