基于曲率阈值与空间聚类的胶条轮廓提取

Rubber Strip Profile Extraction Based on Curvature Threshold and Spatial Clustering

摘要:

换热器是核电站的重要热力设备，板片胶条作为实现密封与形成流道的重要部件，在长期运行过程中易发生老化、磨损与腐蚀，因此对胶条轮廓的精确提取具有重要意义；针对板片波纹结构三维数据复杂、易受噪声干扰以及胶条轮廓难提取等问题，提出一种融合曲率阈值与空间聚类的双重筛选方法。方法通过对原始点云进行下采样与滤波预处理，并结合法向量Z轴分量作为曲率特征，通过分位数阈值提取高曲率区域；采用基于密度的（Density-Based Spatial Clustering of Applications with Noise，DBSCAN）空间聚类算法分割胶条点云，并通过随机抽样一致性（Random Sample Consensus，RANSAC）局部平面投影得到二维轮廓；结合凸包分析实现胶条位置提取与测量；实验结果表明，该方法在胶条轮廓提取中的平均精度达到±5 mm，重复性测量误差小于1 mm，并已在阳江核电站现场得到验证。

Abstract:

Heat exchanger is an important thermal equipment of nuclear power plant. As an important part of sealing and forming flow channel, plate strip is prone to aging, wear and corrosion during long-term operation. Therefore, it is of great significance to accurately extract the profile of strip. Aiming at the problems of complex three-dimensional data of plate corrugated structure, easy to be interfered by noise and difficult to extract the contour of rubber strip, a double screening method combining curvature threshold and spatial clustering is proposed. Firstly, the original point cloud is preprocessed by down-sampling and filtering, and the normal vector Z-axis component is used as the curvature feature, and the high curvature region is extracted by the quantile threshold. The density-based spatial clustering of applications with noise (DBSCAN) spatial clustering algorithm is used to segment the strip point cloud, and the two-dimensional contour is obtained by random sample consensus (RANSAC) local plane projection. Combined with convex hull analysis, the strip position extraction and measurement are realized. The experimental results show that the average accuracy of this method in the strip contour extraction is ±5 mm, and the repeatability measurement error is less than 1 mm, which has been verified in the field of Yangjiang nuclear power plant.