Application of neural networks to fluid measurements with a five-hole probe

For determining the velocity vector with a five-hole pressure probe without rotation around the axis a previous calibration is necessary. A polynom is usually used for the analysis of the calibration. For getting a sufficient accuracy a minimum number of calibration points is needed for approximatin g the polynom coefficients. The paper describes a method based on artificia l neural network which reduces the number of calibration points while simul taneous increase of the accuracy.