Prediction of circular jet streams with artificial neural networks

Abstract

In this study, an ANN model was established by using experimental measurement values at low speed sub-audio level in an air tunnel of which length is 75cm and experiment room cross sections are 32cm x 32cm and model results were compared to experimental values and then, the prediction was made for unmeasured jet stream values. The jet stream at value of 30m/s in the wind tunnel is ensured with a compressor connected to the inlet of wind tunnel experiment room. The tunnel speed values of 0, 10 and 20m/s is ensured with a frequency converter axial fan by making suction in same direction with jet stream. In experimental studies, the jet speed in wind tunnel and radial speed diffusion measures are obtained with a hot wire anemometer enable to make two-dimension measure in the wind tunnel experiment room. In the experiment room, measurements are made with measurement stations located in four different distances. To establish the ANN model, the tunnel speed, length rate and radial distances were taken as an input, with these data, by training the ANN model, networks were established and the radial speed diffusions corresponding to these inputs were obtained as an output. With the data obtained from that network, experimental measurement was made and speed profiles in data ranges were predicted and compared to experimental results. To verify the predicted results, these values and experimental results were compared relatively on non-dimensional speed diffusion graphics. Additionally, similar speed diffusion values and non-dimensional speed diffusion graphics were obtained for 5 and 15m/s tunnel speeds without experimental measurements and comparative comments were made. © 2012 IEEE.