Huang, Po-WeiPo-WeiHuangZane Wang, Yu ChiYu ChiZane WangFU-CHENG WANG2024-03-012024-03-012024-03-1502632241https://scholars.lib.ntu.edu.tw/handle/123456789/640172Valve leakage is a serious safety issue in petroleum and petrochemical production processes. Traditional valve manufacturers examine valve leakage by manual visual inspection, but this method is inefficient and prone to errors. Hence, this paper applies machine vision techniques to develop a measurement method that can detect valve leakage according to the global standards of the petroleum industry, such as those recommended by the American Petroleum Institute. A collection device and machine vision techniques were applied to count leaked bubbles. The results were then compared with traditional human visual counting. The proposed machine vision method can achieve a mean absolute error of less than 1% while also reducing two-thirds of the workload of traditional human visual counting. Experimental results confirmed the effectiveness of this method. This study is the first to adapt machine-vision technology to industrial valve leakage measurement and opens up new possibilities for further exploration of valve leakage.Bubble | Intensity | Leakage | Machine vision | Standard | Valve[SDGs]SDG6journal article10.1016/j.measurement.2024.1142212-s2.0-85183924401https://api.elsevier.com/content/abstract/scopus_id/85183924401