A Study on a Bubbling 2-D Column by Using Image Processing with Neural Network

• Main Authors: Chai-Chung Chen, 陳嘉忠
• Other Authors: Jyh-Tong Teng
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/48385136743788281192

碩士 === 中原大學 === 機械工程學系 === 87 === The study of the bubble behavior in a fluidized bed can help the design of nozzles and gaseous distribution plates used in the fluidized bed. Traditionally the measurements of the bubble related phenomena in a fluidized bed were performed by the pressure perturbation technique, but this method is intrusive. Several optical instruments such as X-ray or γ-ray have also been used in the study of bubbles in a 3-D fluidized bed, but the apparatuses are much more expensive and bubbles in a 3-D fluidized bed may overlap each other. Such an overlapping effect will make study of the behavior of individual bubbles more difficult. The purpose of this study is to develop a measurement tool and an evaluation methodology for a 2-D column based on the image processing technology. The apparatuses including CCD(charge couple device),image grabber and PC are non-intrusive. SOM(self-organization feature map) and Fuzzy ART(fuzzy adaptive resonance theory) of artificial neural network are used to find out the threshold criterion for image segmentation. To reduce the noise of the image, median filtering is applied to each frame. Besides, a bubble auto judgment rule is developed in this study. Furthmore, the effects of bubble splitting and coalescing is considered. Several bubble related parameters such as area, velocity, shape ratio and aspect ratio will be determined by the image processing technology and compared with the correlations reported in the literature. For examining the suitability and accuracy in image segmentation, other objects’s images were analyzed. The method developed in this study is more accurate than conventional method, and it can reduce the time needed for analyzing the experimental results.