206 ニューラルネットによる学習型流速場測定法

抄録

This paper provides a new Particle Image Velocimetry (PIV) with learning function by Neural Network. This method is based on the Fourier transform method and detects a special shift by using a multiple-layer Neural Network, on purpose to improve the precision of measurement. The network learns the input-output patterns in one-dimensional problem by Error Back Propagation method.
Performance test is carried out in two-dimensional problem and the test results show that the present method induces less errors for the image patterns with velocity fluctuation than the original Fourier transform method.