Development of a Fuzzy Inferencing Computer Board Which Detects Ordered Motion Patterns From Flow-Visualized Pictures.

流动显示是流体力学中一种非常有用的方法，但它也存在一些缺陷。指出了在从流视图像中识别模式的最后过程中存在研究者偏见的风险。本计画的目的是开发一种电脑板，它可以高速执行模糊推理，并且应该能够在控制偏见的情况下从流动视觉图片中检测出有组织的图案。经过几次重新设计，计算机板的生产获得了成功。在项目的过程中，很明显，软件或赋予推理机的规则比硬件本身重要得多。为了检查该系统的能力，选择了通过烟雾可视化的三角翼周围的流动的图片、喷水喷射中的飞行磨料树脂的图片和从卫星上拍摄的日本上空的云的图片，以及该系统是否可以发现涡流、树脂和条纹云。测试是成功的，但它被发现，给系统的推理规则的影响很大。虽然这项研究是非常新的，并在初步阶段，未来的规则是有希望的进一步改进。

Although flow-visualization is a very useful method in fluid dynamics, it also has defects. It has been pointed out that there is a risk of researcher's prejudice coming-in in the final process of recognizing pattern from the flow-visual pictures. The aim of this project is to develop a computer board that can perform fuzzy inferencing at high speed and should be capable of detecting organized pattern from the flow-visual pictures with controlled prejudice. The production of the computer board was successful after several redesignings. In the process of the project, it became clear that the software or the rules given to the inferencing machine is much more important than the hardware itself. In order to check the capability of the system, a picture of a flow around a delta-wing visualized by smoke, pictures of flying abrasive resins in waterjet and a picture of clouds over Japan taken from a satelite were chosen, and whether the system can find vortices, resins and stripe clouds. The tests were successful, but it was found out that the inferencing rules given to the system affect the results strongly.Though this research is very new and is in a preliminary stage, future is promising by further improvement in the rules.