基于气泡法的阀门泄漏检测智能化研究综述  

A Review of Intelligent Valve Leakage Detection Based on Bubble Detection Method

作　　者：赵首琦 卢斌[2] 蓝一凡 曹秀 金志江[3] 迟国冰 ZHAO Shou-qi;LU Bin;LAN Yi-fan;CAO Xiu;JIN Zhi-jiang;CHI Guo-bing(SUFA Technology Industry Co.,Ltd.,CNNC,Suzhou 215129,Jiangsu,China;Wuhan University of Technology,Wuhan 430070,Hubei,China;Institute of Wenzhou,Zhejiang University,Wenzhou 325036,Zhejiang,China;Wenzhou Fengyong Intelligent Technology Co.,Ltd.,Wenzhou 325036,Zhejiang,China)

机构地区：[1]中核苏阀科技实业股份有限公司,江苏苏州215129 [2]武汉理工大学,湖北武汉430070 [3]浙江大学温州研究院,浙江温州325036 [4]温州风涌智能科技有限公司,浙江温州325036

出　　处：《阀门》2025年第4期451-459,共9页Chinese Journal of Valve

基　　金：温州市自主申报科研项目(ZZG2023004);温州市揭榜挂帅科研项目(ZG2024033)。

摘　　要：阀门是调控流体流量、方向及开关状态的关键部件,其早期气密性检测对保障工业流水线安全运行至关重要。本文聚焦阀门泄漏检测的气泡法,系统综述了传统运动目标检测技术的优化策略及深度学习技术的应用进展。传统方法通过融合边缘检测与自适应阈值分割算法,缓解了光照干扰与实时性不足的问题;深度学习技术通过卷积神经网络自动提取气泡特征,实现了重叠气泡分割与形状参数预测,检测精度显著提升。然而,模型泛化能力不足与计算复杂度高仍是瓶颈。未来需通过构建多样化数据集、迁移学习及轻量化模型设计,推动气泡法检测技术的智能化发展。Valves,as critical components for regulating and controlling the flow,direction,and on/off states of fluids,require early-stage airtightness testing to ensure the safe operation of industrial production lines.This study focuses on the widely adopted bubble method for valve leakage detection,providing a systematic review of optimization strategies for traditional moving object detection techniques,alongside recent advances in deep learning applications.Traditional methods partially address challenges such as lighting interference and insufficient real-time performance through the integration of edge detection and adaptive threshold segmentation algorithms.Deep learning techniques,leveraging convolutional neural networks(CNNs),automatically extract bubble features,enabling precise segmentation of overlapping bubbles and prediction of shape parameters,thereby significantly enhancing detection accuracy.However,limitations persist,including inadequate model generalization capability and high computational complexity.Future development directions involve constructing diversified datasets,applying transfer learning,and designing lightweight models to advance the intelligent evolution of bubble-based detection technology.

关 键 词：阀门 密封 气泡法 泄漏检测 深度学习 

分 类 号：TH134[机械工程—机械制造及自动化]