Session: 36-09 Uncertainty Quantification & Sensitivity Analysis (1)

Paper Number: 121341

121341 - Impact of Input Data Quality on the Uncertainties of a Compressor Blade Parametrization Process 

Accounting for a significant part of an airline's operational expenses, the cost of maintenance and fuel continues to rise. One method to counteract this trend is to consider the impact of individual parts on aero-engine efficiency and thus base maintenance decisions on more information. In the course of this desired shift towards a knowledge-based approach to aero-engine maintenance, a process for automatic parametrization and evaluation of compressor blades is developed at the Institute of Jet Propulsion and Turbomachinery. The real and individual blades are 3D-scanned and their profiles are then extracted along the blade height. These profiles are then parametrized in terms of airfoil parameters, such as chord length, thickness, metal angles, or camber. Based on this parametrized data, statistical analysis of the blade degradation is possible. Furthermore the process includes the reconstruction of 3D-blade geometries from parameters and the automatic generation of CFD simulations. Therefore, the impact of degradation, based on specific parameters, can be quantified.  
 
As the individual blades are analysed from 3D-scanned pointclouds, the quality of these pointclouds has a direct impact on the results of the parametrization and thereby the entire process. To quantify the impact and evaluate the uncertainties of the parametrization process, a test structure is developed. The test structure consists of individual tests for two main steps of the process which, where applicable, are linked for a resulting total uncertainty of the resulting blade parameters. The first tested process step is the alignment of input data, which automatically orients the blade pointcloud in accordance to the blades position in the aero-engine.  This eliminates the need to precisely locate the physical blade during 3D scanning. The second process step tested is the parametrization of the blades profiles. This includes the automatic detection and measurement of the aforementioned parameters, as well the calculation of chordwise thickness and camber distributions. Furthermore the relative positioning of the profiles is calculated, to collect information about the lean of the blade.  

The test procedures are based on a monte-carlo approach, in which the process steps are applied to slightly varied input data. This simulates changes of input data quality based on two quality parameters. The first data quality parameter is the density of the three-dimensional pointclouds or total number of points in case of the two-dimensional profiles. The second quality parameter is the noise magnitude, which describes the point to point inaccuracy of the scan. To test the impact of the quality parameters on the process uncertainties, artificial input data is automatically created with a given density and noise magnitude. Both parameters are subject to the operator's choice in the scanning process or the accuracy of the equipment used, determining the scan process duration and cost.

Based on the artificial input data, calculation of the process uncertainty due to the input data quality is possible. Therefore, the resulting statistical distribution of output data is analysed.

Results of these tests show a sufficiently small uncertainty of the process, as long as the input data is of high enough quality. The pointcloud density has a small impact as long as a certain threshold of about 1000 points per analyzed profile section or 800.000 per Blade is reached. Especially the parametrization step however shows a significant dependence from the noise magnitude. With the results of this study, minimum quality criteria for future input data can be set depending on the required results and thereby the cost and time required for 3D scanning and processing optimized. 

Presenting Author: Benedikt Schulten Institute of Jet Propulsion and Turbomachinery

Presenting Author Biography: I studied aeronautical engineering at TU Braunschweig with a focus on jet engines. Now I am a research assistant at the Institute for Jet Propulsion and Turbomachinery and work on a project aiming to develop a parameter based evaluation process for compressor blades.

Authors:

Benedikt Schulten Institute of Jet Propulsion and Turbomachinery
Jan Goeing Institute of Jet Propulsion and Turbomachinery
Sebastian Lueck Institute of Jet Propulsion and Turbomachinery
Jens Friedrichs Institute of Jet Propulsion and Turbomachinery