LUFO IV: compressor design under consideration of cost and robustness

The target of the project “compressor design under consideration of cost and robustness” is to consider the manufacturing tolerances in the performance calculation of the compressor. Therefore the difference of manufactured blades and the CAD geometry has to be quantified in form of parameters to allow their consideration in the numerical simulation.

Blades of rotor 3, 6 and 9 of the high pressure compressor of BR 710 jet engine are measured using the scanning system “Kolibri Flex” of the company IVB (Jena, Germany). The system uses structured light (fringe projection) to scan the complete surface optically in three dimensions. The analysis of the measured blades is done by a method developed at the institute of fluid mechanics at Dresden University of Technology. Section outlines on rotation symmetrical faces are extracted from the point cloud for performing the calculation of camber line and chord line, which define the thickness and camber distribution. These distributions give access to parameters engineers prefer: e.g. thickness of leading and trailing edge, chord length, stagger angle, maximum thickness etc. The connection of the individual spanwise sections is received through stacking laws for rebuilding the measured geometry in three dimensions. An interesting aspect of this approach is that reverse engineering in the sense of updating the mostly NURBS-based CAD-model is not necessary because the mentioned profile parameters are used to describe the measured geometry (and e.g. manufacturing tolerances).

As the main result of the measuring process the profile parameters including their statistical properties (scatter, mean value, correlation with other parameters) are gained. This information is used within a Monte Carlo Simulation (MCS) that creates realizations of the blades. Their geometry is created by adding the difference between the measured blade and the CAD-geometry to the geometry used in CFD-calculation. This is done by a morphing algorithm developed at the institute as well. The aerodynamical characteristics (profile loss, flow turning, etc.) of the morphed blades are calculated by 3D-CFD simulations using NUMECA FINE TURBO. Herein, each blade represents a realization in the sense of MCS and is not a copy of a measured blade, but the statistical properties of all realizations created in the MCS are equivalent to the statistical properties of all measured blades. The advantage of this approach is that any desired number of realizations can be created independent of the number of measured blades, as long as a sufficient number of blades are measured to provide representative statistic results. This is an important fact for this project, because the CFD-simulations are performed on a quarter (90° model) containing approx. 15 blades. This means that the number of required realizations is much higher than the number of blades that could be measured. Mentioning the quarter another advantage of the morphing approach becomes clear. With the use of individual blades within one 90° model, the influence of other effects like the changing of the throughflow section between two blades can be quantified. Furthermore the morphing method allows creating the desired geometry without the time consuming remesh process.

Finally throughflow-calculations are performed to consider the manufacturing related scatter in the performance of the whole compressor. Herein, the results of the CFD-calculations (e.g. correlations between the geometrical profile parameters and the aerodynamic parameters of one single blade) are used.

© TFA

As a result of the project a sensitivity analysis shall clarify the influence of the manufacturing tolerances on the performance of the entire compressor. This might be interesting for the quality assurance department. They only have to ensure the parameters of great influence to be within strict tolerances. On the other hand, manufacturing of the blades could be cheaper if the tolerances of parameters with little influence are increased. Furthermore the model can be used to evaluate the robustness of the design depending on the parameters with scatter.