Dynamic Stress Prediction, 18-R9816

Printer Friendly Version

Principal Investigator
J. Jeffrey Moore

Inclusive Dates:  04/01/08 – 10/30/09

Background - Advances in turbomachinery technology have increased the performance of gas turbine engines and compressors by increasing the operating speeds, pressures, and temperatures of the machinery. Although greater performance is obtained, the harsher operating environments can lead to premature failure of components. To avoid the high cost of such a failure, turbomachinery companies often request detailed engineering design reviews from SwRI, particularly for unproven designs. In cases where a failure has already occurred, SwRI is frequently asked to perform root cause failure analyses (RCFA) so that similar failures can be avoided in the future.

Approach - Most analytical methods for dynamic stress prediction with aerodynamic loading rely on some combination of finite element analysis (FEA) and computational fluid dynamics (CFD). Previously, these models were not coupled, and assumptions of the loading in the FEA model were required. A fluid-structure interaction (FSI) method is available in the commercial FEA/CFD software, where FEA and CFD solvers simulate transient response data alternately at each time step. However, the method has never been validated for turbomachinery applications. The proposed project will perform an FSI transient solution that will solve for the flow field and stress field at every time step thereby calculating the dynamic stresses in the blading. As an alternative to CFD and FEA, an empirical method has been developed at SwRI that predicts dynamic stresses by manipulating transfer functions obtained from experimental data. The method introduces the aerodynamic loading on a point-by-point basis and does not capture the spatial and temporal variation in loading. Although deemed conservative, it may not accurately capture the dynamic interaction between blades. It also assumes the loading profile for aerodynamic loading. A recent method based on the proper orthogonal decomposition (a statistical signal processing tool) and linear system theory has been developed for empirical modeling that shows promise for stress prediction. The method has firm mathematical foundations and is able to predict a structure's response to arbitrary loading conditions. The method is currently limited to predicting displacement responses and has only been validated for structures with 1-D geometry. Therefore, the proposed research will extend the method to accommodate strain predictions for structures with complex geometry.

To validate the FSI calculations, the project used an existing Centrifugal Gas Turbine (CGT) test rig pictured in Figure 1. This test rig was modified into a centrifugal compressor rig by removing the combustor, turbine buckets, and nozzles, keeping only the centrifugal compressor. It has been previously operated to 30,000 rpm. The rotor was further modified to accept the strain gages, compact amplifier, and slip rig to measure the dynamic strains on the blading, as shown in Figure 2. The measurement involves instrumenting a centrifugal compressor impeller with strain gages on the rotor blades and measuring the blade vibration caused by interaction with the IGV wakes while spinning at 27,000 rpm. A compact rotating amplifier was developed to avoid significant noise associated with running unamplified signals through a slip ring resulting in accurate strain measurement.

Figure 1. CGT Test Rig.

Figure 2. Slip Ring Assembly.

Accomplishments - Significant progress has been made this past year and a half on the project, including developing of the proper orthogonal decomposition method that uses strain gages and was validated on a gas turbine blade. On the computation side, a successful model was built and solved to perform a coupled computational fluid dynamics and finite element analysis to solve for the dynamic stresses caused by wake interaction of the inlet guide vane and centrifugal compressor impeller. This model was solved on an SwRI high-performance computing cluster. Preliminary results have been obtained from the test rig demonstrating success in the rotating blade strain measurement. Additional CFD studies are under way, and comparisons will be made to the measurements over a range of operating flows.

2009 Program Home