大流量浆体长输管道机械撞击损伤检测方法  

Detection Method for Mechanical Impact Damage of Large-flow Slurry Long-distance Pipeline

作　　者：冯斌 FENG Bin(CCTEG Wuhan Engineering Company,Wuhan 430064,China)

机构地区：[1]中煤科工集团武汉设计研究院有限公司,湖北武汉430064

出　　处：《机械制造与自动化》2025年第2期215-220,共6页Machine Building & Automation

摘　　要：提出大流量浆体长输管道机械撞击损伤检测方法,改善管道损伤图像处理能力,提高机械撞击损伤检测效果。利用管道视觉检测机器人采集大流量浆体长输管道全景图像,通过灰度梯度正则化方法消除管道全景图像所含噪声,经管道图像的分割及边界跟踪预处理后,分别提取预处理后管道图像尺寸、形状、纹理特征,将特征提取结果作为改进卷积神经网络缺陷检测模型的输入,实现大流量浆体长输管道机械撞击损伤类型的检测。实验结果表明:该方法可提高采集管道全景图像清晰度,增强纹理细节信息;可实现损伤区域的分割,确定缺陷边缘;可准确检测撞击损伤类型,检测误差低于5%。A method for detecting mechanical impact damage of large-flow slurry long-distance pipeline is proposed to improve the image processing ability of pipeline damage and enhance the detection effect of mechanical impact damage.The pipeline vision inspection robot is applied to collect the panoramic image of the large-flow slurry Rlong-distance pipeline,and the noise contained in the panoramic image of the pipeline is eliminated by the gray gradient regularization method.After the pipeline image segmentation and boundary tracking preprocessing,the size,shape and texture features of the pipeline image after preprocessing are extracted respectively,and the feature extraction results are used as the input of the improved convolution neural network defect detection model to realize the detection of mechanical impact damage types of large-flow slurry long-distance pipeline.The experimental results show that the proposed method can improve the clarity of the panoramic image of the collected pipeline and enhance the texture details.And the damaged can be segmented,the defect edge can be determined,and the impact damage type can be accurately detected with detection error less than 5%.

关 键 词：灰度梯度正则化 传输管道 机械撞击损伤 最大类间方差 边界跟踪 缺陷检测模型 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]