No abstract

This paper addresses the consideration of variability due to manufacturing and abrasion of high pressure turbine (HPT) blades through probabilistic CFD investigations. Currently, the influences of those effects onto relevant aerodynamic quantities are covered by safety factors and conservative assumptions. The probabilistic investigation on the other hand, based on realistic input parameters, will enable designers to possibly reduce safety factors and provide the foundation for more robust blade designs. In order to establish a sound statistical database, 500 new and used HPT blades were digitised with an optical 3D scanning system in order to record the outer geometric variability. As described in the work of Heinze et al. [1], the measured point clouds are parameterized using classical profile parameters and a process to rebuild various blade geometries has been developed. In terms of probabilistic investigations, in our case by applying sampling-based methods, a large number of 3D CFD simulations for a 1.5 HPT stage have been conducted using probabilistic blade geometries, followed by a sensitivity analysis and a detailed investigation of the probabilistic system behaviour. For visualization of possible correlations, the technique of Statistics on Passage (SoP) was chosen, as described in [2].

During the design process of compressor blades predominantly deterministic models are used for High Cycle Fatigue (HCF) strength investigations. The scatter of HCF that results e.g. through abrasion of the production machines [1] or inhomogeneities of the blade material, is accounted by safety factors and conservative assumptions. A more realistic approach to consider these uncertainties is the application of probabilistic methods. Therefore, further information about HCF and eigenfrequency scatter of the really produced blades can be used for a robust design during the design process. Within a measurement campaign 400 blades of a Rolls-Royce High Pressure Compressor were randomly selected and scanned using an automated process that applies the optical measurement technique of strip projection. The measurement data of the airfoil were subdivided into constant spanwise profile slices. Geometric airfoil parameters were determined on each of the profile slices [2]. Due to the large number of scanned blades each geometric airfoil parameter can be described as a distribution function with corresponding parameters. These distribution functions are the input parameters for the probabilistic investigation — the Monte-Carlo-Simulation (MCS). Within the MCS an automatically transfer process varies at first the profile slices of a CAD-airfoil and in a second step morphs an existent 3D finite element mesh applying the meshmorphing tool of the FE preprocessor Hypermesh. The HCF and eigenfrequency scatter of all blades were calculated with the interpretation of the MCS results and parameters were detected with the largest influence on HCF-strength and eigenfrequencies. A detailed interpretation of the HCF-strength at one example shows the power of the probabilistic investigation. The interpretation helps the engineer to understand the entire system and to design a robust blade.

The measurement of compressor or turbine blades is very interesting for quality control and inspection checks. Especially for mechanical wear valuation it is important to inspect the whole geometry of a blade with main focus on the airfoil. This 3D measurement task, which involves many identical components, and high accuracy, as well as high point density in the airfoil edge regions, demands detailed planning of the appropriate 3D-measuring machine, measurement planning, reproducibility as well as analysis and provision of measuring values for the user. Preliminary investigations were carried out to select an appropriate measuring method providing both a complete geometric and as automatic as possible acquisition of blade geometries. Another constraint was that the technique had to make efficient use of time and be accurate. A data processing routine using software developed at TU Dresden is used for automated analysis and provision of measuring values for the user. The approach to linking automated complete 3D data acquisition by means of strip projection with automated analysis of measuring data opens up the option of using it as an efficient 3D-measuring machine for high accuracy requirements and higher quantities. The method shown can be transferred to other objects that need to be scanned in larger quantities.

No abstract