Session: 05-12 Instrumentation II: Pressure Probes

Paper Number: 128955

128955 - Application of Machine Learning Techniques in Calibration and Data Reduction of Multi-Hole Probes 

Aerodynamic probes such as multi-hole probes have been established as robust, simple and cost-effective tools for determining three-component velocity components and fluid properties. It is shown that these probes provide accurate measurements for flow angles up to 75 degrees. A directional probe can be operated in yaw-null or differential modes. In the non-nulling or differential mode which is the focus of this work, the probe stays stationary in the flow, Mach number and flow angles are obtained from a calibration map based on the differential pressure readings across the opposite taps of the probe. The calibration of a multi-hole probe depends on its geometrical construction and as long as the structural integrity is maintained, the probe holds its calibration. The probe calibration is usually carried out in a free jet or a closed wind tunnel. It is presented using non-dimensional pressure differences in the form of calibration maps. Generally, these calibration maps are functions of flow direction (pitch and yaw angles) and Mach number. Thus, in order to obtain accurate measurements by utilizing a multi-hole probe, a large amount of calibration data is needed in the ranges of interest for flow angles and Mach number.

This work presents procedures to implement machine-learning methods in the existing algorithms for multi-hole probe calibrations and data reduction. It is shown here, that utilizing artificial neural networks (ANNs) can reduce the amount of calibration data that needs to be acquired in order to obtain a specific calibration uncertainty, by more than 50% while simultaneously reducing data reduction times significantly. In this approach, ANNs were used instead of surface fitting methods to relate directional calibration coefficients to flow angles. Then, the flow angles were used as an input to another set of ANNs to define Mach number and static pressure iteratively. In a second approach explained in this work, novel calibration coefficients were used to directly relate the pressure measurements from multi-hole probe to the quantities of interest thus, eliminating the need for iterative algorithms. This method has an average increase of less than 1% in the calibration uncertainty for the large flow angles while reducing the data reduction times to a few seconds.

The probe calibration dataset used in this work were acquired in NASA CE-12 free jet calibration facility. The calibration data set ranged from Mach 0.01 to 0.9 and flow angles varied from -25 to 25 degrees. The methodology to avoid over-fitting and under-fitting are elucidated.

Presenting Author: Arman Mirhashemi NASA Glenn Research Center

Presenting Author Biography: Arman Mirhashemi is a research engineer in the turbomachinery and turboelectric systems branch at NASA Glenn research center. His area of research is focused on turbofan power extraction for hybrid-electric architectures, thermal management challenges in turbomachinery applications, and application of artificial intelligence techniques in the propulsion field. He received his PhD from the University of Notre Dame and followed that as a postdoctoral research fellow at the Notre Dame Turbomachinery Laboratory (NDTL) before joining NASA GRC.

Authors:

Arman Mirhashemi NASA Glenn Research Center
Paht Juangphanich NASA Glenn Research Center
Kenji Miki NASA Glenn Research Center