DEVELOPMENT OF FIVE-HOLE PROBE PROCESSING TECHNIQUES FOR ANGLE OF ATTACK, ANGLE OF SIDESLIP, AND AIRSPEED APPLIED TO EXPERIMENTAL DATA

• Main Author: Birry, Abdurrahman
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/79467
• Institution: Institut Teknologi Bandung

In this thesis, the Artificial Neural Networks (ANN) technique was implemented to predict the flow parameters from five-hole probe measurement. The ANN technique is used to estimate the angle of attack, angle of sideslip, and flow speed. The experimental data of this work were obtained from a low-speed wind tunnel testing. In the testing, the velocity of the tunnel varied from 10 to 50 m/s in increments of 10 m/s. The angle of attack varies from -30° to 90° and the angle of sideslip is varied from -20° to 20°. From the measured data, 80% are randomly chosen to build the ANN model, while the remaining 20% are used to validate the model. The accuracy of the ANN technique is compared to other known methods. The first method uses the formula provided by the manufacturer of the five-hole probe and the second method uses 5th-order polynomial fitting. The novelty of the process in this paper is in the use of single ANN model to predict all three flow parameters. This single model can be used throughout the range of angle of attack, sideslip, and speed. The findings reveal that the ANN model accurately forecasts flow parameters in the low angle of attack range, surpassing the formula and polynomial methods in mean absolute and maximum absolute error. Furthermore, the ANN method significantly outperforms the other two methods in obtaining flow parameters in high angle of attack regions. The proposed model can provide mean absolute and maximum absolute error of 0.164 and 2.605, respectively for angles of attack, of 0.584 and 14.522, respectively for angle of sideslip, and of 0.024 and 0.242, respectively for speed. Further improvement of the ANN model can be obtained by segmenting the applicability region. The initial result shows that 95% improvement is achievable.