Study of 3D-flow-field velocimetry for turbomachine

本研究开发了用于PIV测量三维空间非定常速度分布的单摄像机图像分析系统，并将其应用于静、旋转圆盘间强剪切流的非定常流动测量中。首先，构建了带反射棱镜的消旋观测系统，测量了转速高达300RPM的叶轮或叶轮机械风扇相对于旋转物体的速度分布。该观测系统采用3CCD摄像机和带两个滤色器的迈克尔逊干涉仪测量示踪粒子在三维空间中的分布深度。每个示踪剂粒子的粒子深度是使用一对以不同焦点设置同时捕获的绿色和红色帧来测量的。通过离焦深度检测技术进行深度检测。数值模拟和实验结果表明，在原始图像空间分辨率较高的情况下，基于离焦深度的深度检测算法在测量深度范围内的误差率为3%。其次，提出了一种基于自组织映射神经网络的三维粒子图像测速算法。通过对立方体空腔内均匀流和旋转流的数值模拟，验证了该系统的跟踪性能，并将该系统成功地应用于静、旋转盘间隙强剪切流的速度分布测量。实验结果表明，利用该方法可以从轴向获取流动可视化图像，从而测量出切向速度分量在轴向的分布。

In this research, the image analyzing system with a single camera was developed for the PIV measurement of the unsteady velocity distribution in 3D space and it was applied to the unsteady flow measurement of the strong shear flow between a static and rotating disks to demonstrate its utility.First, a derotating observation system with a reflective prism was constructed to measure the velocity distribution relative to the rotating object, such as an impeller or a fan of turbomachine at the speed up to 300RPM. A 3CCD camera and the Michelson interferometer with two color filters were employed in the observation system to measure the depth of tracer particles distributed in 3D space. The particle depth for each tracer particle was measured using a pair of green and red frames simultaneously captured with different focus settings. The depth detection was performed by means of the technique of a depth-from-defocus. It was confirmed in numerical simulations and experiments that the depth detection based on a depth-from-defocus could be carried out with the error ratio of 3% to the measurable depth range if the spatial resolution of original images was higher.Secondly, a 3D particle tracking algorithm based on self-organizing maps neural network was developed for 3D Particle Image Velocimetry (3D-PTV). The tracking performance was checked in numerical simulations for the uniform flow and the rotating flow in a cubic cavity.Furthermore, the present observation system with a single camera was successfully applied to the velocity distribution measurement for the strong shear flow in the gap between a static and rotating disks. In this experiment, it was shown that the distribution of tangential velocity component along the axial coordinate could be measured with the flow visualization images captured from the axial direction using the present method.