Miniature Multihole Pressure Probes and Their Neural-Network-Based Calibration

Abstract

We present the development of miniature multihole pressure probes and a novel neural-network-based calibration algorithm for them. Seven-hole probes of tip diameters as low as 0.035 in. (0.9 mm) were successfully fabricated with high tip surface quality. Any of the typical probe tip geometries, i.e., hemispherical, conical, or faceted, could be fabricated. The miniature probes were calibrated and tested in a wind tunnel. A backpropagation-based neural-network calibration algorithm was developed for these probes, with flexibility in network architecture design and network self-optimization capabilities. In the feedforward mode the algorithm yields computational speeds an order of magnitude higher than those typically achieved by similar accuracy interpolation algorithms. The new algorithm has prediction accuracies of 0.28 deg in the flow angles and 0.35% in the velocity magnitude