基于点云数据的航空发动机管道分割技术

孙龙辉; 沈建新; 徐建国

doi:10.13224/j.cnki.jasp.20250060

基于点云数据的航空发动机管道分割技术

doi: 10.13224/j.cnki.jasp.20250060

南京航空航天大学机电学院，南京 210016

详细信息

作者简介:
孙龙辉（1999-），男，硕士，研究方向为点云处理。E-mail：slh18137640977@163.com

通讯作者:
沈建新（1969-），男，教授，博士，研究方向为数字化设计与制造。E-mail：cadatc@nuaa.edu.cn

中图分类号: V263
计量
- 文章访问数: 28
- HTML浏览量: 20
- PDF量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2025-02-08
- 网络出版日期: 2026-08-14

Pipe segmentation technology for aircraft engines based on point cloud data

College of Mechanical and Electrical Engineering，Nanjing University of Aeronautics and Astronautics，Nanjing 210016，China

摘要

摘要:
航空发动机通常布有大量自由弯曲、交错分布的管道，为避免运行中管道间的摩擦或共振引发故障，装配过程中需严格控制管道间距。然而，现有检测方法依赖人工操作，且存在劳动强度大、效率低、测量精度不稳定的问题。利用激光扫描仪采集航空发动机管道点云数据并进行间距计算，可以显著提高管道间距测量的自动化程度和效率。然而，航空发动机管道点云数据量庞大且分割困难，为后续的管道间距测量带来了挑战。因此，本研究以曲率、法向量夹角、法向量叉积三种关键特征作为分割判断条件，提出了一种基于层次化多特征的管道分割算法。计算结果表明，该算法通过逐层细化分割的方式精确分割出管道点云，改善了现有算法的过分割和欠分割问题，准确率达到了94.65%，为后续计算管道间距奠定了基础。
- 航空发动机管道 /
- 管道间距测量 /
- 点云数据 /
- 管道分割 /
- 层次化多特征
Abstract:
Aero-engine usually has a large number of freely-bent and interlaced pipes. To avoid faults caused by friction or resonance between pipes during operation, the pipe spacing must be strictly controlled during assembly. However, the existing detection methods rely on manual operation and have problems such as high labor intensity, low efficiency, and unstable measurement accuracy. Using a laser scanner to collect point cloud data of aero-engine pipes and calculate the pipe spacing can significantly improve the degree of automation and efficiency of pipe spacing measurement. However, the point cloud data of aero-engine pipes is huge and difficult to segment, which brings challenges to the subsequent pipe spacing measurement. Therefore, a hierarchical multi-feature-based pipe segmentation algorithm is proposed by taking curvature, normal vector angle, and normal vector cross product as the segmentation judgment conditions. The calculation results show that this algorithm can accurately segment the pipe point cloud through a layer-by-layer refinement method, improving the over-segmentation and under-segmentation problems of existing algorithms, with an accuracy rate of 94.65%, which lays a foundation for subsequent pipe spacing calculation.
- aircraft engine pipes /
- pipe spacing measurement /
- point cloud data /
- pipe segmentation /
- hierarchical multi-feature

HTML

图 1 航空发动机点云数据

Figure 1. Aero engine point cloud data

下载: 全尺寸图片幻灯片

图 2 发动机管道测量模型

Figure 2. Engine pip measurement model

下载: 全尺寸图片幻灯片

图 3 体素网格

Figure 3. Voxel grid

下载: 全尺寸图片幻灯片

图 4 滤波对比效果

Figure 4. Filter contrast effect

下载: 全尺寸图片幻灯片

图 5 管道几何结构

Figure 5. Pipe geometric structure

下载: 全尺寸图片幻灯片

图 6 技术路线

Figure 6. Technical route

下载: 全尺寸图片幻灯片

图 7 点云法向量夹角

Figure 7. Angle between point cloud normal vectors

下载: 全尺寸图片幻灯片

图 8 法向量叉积一致性

Figure 8. Consistency of normal vector cross product

下载: 全尺寸图片幻灯片

图 9 原始点云均匀采样

Figure 9. Uniform sampling of the original point cloud

下载: 全尺寸图片幻灯片

图 10 发动机管道测试区域

Figure 10. Engine pipe test area

下载: 全尺寸图片幻灯片

图 11 消融实验结果对比

Figure 11. Comparison of ablation results

下载: 全尺寸图片幻灯片

图 12 分割结果对比

Figure 12. Comparison of segmentation results

下载: 全尺寸图片幻灯片

表 1 点云数据信息

Table 1. Point cloud data information

数据编号	数据类型	点云数目
Engine data1	扫描数据	1310895
Engine data2	扫描数据	1170201
Model data	模型数据	6514650

下载: 导出CSV

表 2 不同点的曲率差异

Table 2. Curvature differences at different points

点类型	点云坐标	曲率
管道点	（−176.5, 42.8, 1673.8）	0.0048
	（−175.1, 41.5, 1674.4）	0.0034
	（−177.1, 42.3, 1674.4）	0.0045
	（−174.1, 41.2, 1674.9）	0.0033
	（−175.6, 41.6, 1674.5）	0.0027
	（−175.1, 40.8, 1675.4）	0.0018
非管道点	（−167.5, 48.3, 1660.8）	0.0257
	（−166.5, 48.3, 1660.9）	0.0405
	（−169.1, 49.8, 1660.6）	0.0368
	（−167.6, 48.8, 1660.7）	0.0328
	（−167.5, 51.8, 1660.7）	0.0352
	（−168.0, 53.7, 1660.8）	0.0281

下载: 导出CSV

表 3 消融实验的算法运行时间对比

Table 3. Comparison of ablation experiment running time

分割算法	运行时间/s
基于曲率	4.87
基于法向量夹角	15.122
基于法向量叉积	25.88
基于曲率+法向量夹角	14.35
基于曲率+法向量叉积	23.34
基于曲率+法向量夹角+法向量叉积	21.43
基于法向量夹角+曲率+法向量叉积	21.55

下载: 导出CSV

表 4 消融实验各算法性能对比

Table 4. Comparison of ablation experimental algorithms

分割算法	准确率/%	召回率/%	F1分数/%
基于曲率	16.00	74.33	26.33
基于法向量夹角	27.54	88.80	42.04
基于法向量叉积	21.87	83.54	34.66
基于曲率+法向量夹角	37.43	74.33	48.36
基于曲率+法向量叉积	26.56	69.17	38.39
基于曲率+法向量夹角+ 法向量叉积	45.91	69.17	55.19
基于法向量夹角+曲率+ 法向量叉积	92.20	71.93	80.81

下载: 导出CSV

表 5 分割算法性能比较

Table 5. Performance comparison of segmentation algorithms

Methods	准确率（%）	召回率（%）	F1分数（%）
层次化多特征分割	94.65	96.07	95.02
区域增长分割	76.87	71.06	72.33
欧式聚类分割	65.77	68.19	60.76

下载: 导出CSV

参考文献(25)

[1]	韦正渊. 航空发动机外形点云数据中管路轴线生成技术研究[D]. 南京: 南京航空航天大学, 2021. Wei Zhengyuan. Research on the Centerline Extraction Technology from Pipeline in Point Cloud of Aero-engine Aspect [D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2021. (in Chinese Wei Zhengyuan. Research on the Centerline Extraction Technology from Pipeline in Point Cloud of Aero-engine Aspect [D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2021. (in Chinese)
[2]	王轩. 航空发动机典型管路布局及振动特性分析[D]. 天津: 中国民航大学, 2018. Wang Xuan. Analysis of typical pipeline layout and vibration characteristics of aero-engine[D]. Tianjin: Civil Aviation University of China, 2018. (in Chinese Wang Xuan. Analysis of typical pipeline layout and vibration characteristics of aero-engine[D]. Tianjin: Civil Aviation University of China, 2018. (in Chinese)
[3]	柴银生. 航空发动机典型管路结构件振动特性分析[D]. 天津: 中国民航大学, 2022. Chai Yinsheng. Analysis of vibration characteristics of typical pipeline components of aero-engine[D]. Tianjin: Civil Aviation University of China, 2022. (in Chinese Chai Yinsheng. Analysis of vibration characteristics of typical pipeline components of aero-engine[D]. Tianjin: Civil Aviation University of China, 2022. (in Chinese)
[4]	张香华, 王立志, 党伟. 航空发动机管路设计研究[J]. 中国新技术新产品, 2018(12): 80-81. Zhang Xianghua, Wang Lizhi, Dang Wei. Research on pipeline design of aero-engine[J]. New Technology & New Products of China, 2018(12): 80-81. (in Chinese doi: 10.13612/j.cnki.cntp.2018.12.043 Zhang Xianghua, Wang Lizhi, Dang Wei. Research on pipeline design of aero-engine[J]. New Technology & New Products of China, 2018(12): 80-81. (in Chinese) doi: 10.13612/j.cnki.cntp.2018.12.043
[5]	潘乃康. 航空发动机外部管路振动特性分析及卡箍布局优化的研究[D]. 沈阳: 东北大学, 2021. Pan Naikang. Analysis of vibration characteristics of aero-engine external pipeline and study on optimization of clamp layout[D]. Shenyang: Northeastern University, 2021. (in Chinese Pan Naikang. Analysis of vibration characteristics of aero-engine external pipeline and study on optimization of clamp layout[D]. Shenyang: Northeastern University, 2021. (in Chinese)
[6]	Ji Wenhao, Sun Wei, Wang Donghai, et al. Optimization of aero-engine pipeline for avoiding vibration based on length adjustment of straight-line segment[J]. Frontiers of Mechanical Engineering, 2022, 17(1): 11. doi: 10.1007/s11465-021-0667-x
[7]	Yuan Hexiang, Yu Jiapeng, Jia Duo, et al. Group-based multiple pipe routing method for aero-engine focusing on parallel layout[J]. Frontiers of Mechanical Engineering, 2021, 16(4): 798-813. doi: 10.1007/s11465-021-0645-3
[8]	樊晶晶, 马骊群, 孙安斌, 等. 基于点云数据的发动机管路最小间距计算方案[J]. 航空动力学报, 2019, 34(11): 2347-2353. Fan Jingjing, Ma Liqun, Sun Anbin, et al. Strategy of calculating minimal-distance of engine pipelines based on point cloud data[J]. Journal of Aerospace Power, 2019, 34(11): 2347-2353. (in Chinese doi: 10.13224/j.cnki.jasp.2019.11.006 Fan Jingjing, Ma Liqun, Sun Anbin, et al. Strategy of calculating minimal-distance of engine pipelines based on point cloud data[J]. Journal of Aerospace Power, 2019, 34(11): 2347-2353. (in Chinese) doi: 10.13224/j.cnki.jasp.2019.11.006
[9]	金鹏, 刘检华, 刘少丽, 等. 基于中心线的管路端点位置精确测量方法[J]. 计算机集成制造系统, 2016, 22(10): 2284-2293. Jin Peng, Liu Jianhua, Liu Shaoli, et al. Centerline-based measuring method for endpoints of tube[J]. Computer Integrated Manufacturing Systems, 2016, 22(10): 2284-2293. (in Chinese doi: 10.13196/j.cims.2016.10.002 Jin Peng, Liu Jianhua, Liu Shaoli, et al. Centerline-based measuring method for endpoints of tube[J]. Computer Integrated Manufacturing Systems, 2016, 22(10): 2284-2293. (in Chinese) doi: 10.13196/j.cims.2016.10.002
[10]	Dimitrov A, Golparvar-fard M. Segmentation of building point cloud models including detailed architectural/structural features and MEP systems[J]. Automation in Construction, 2015, 51: 32-45. doi: 10.1016/j.autcon.2014.12.015
[11]	Kawashima K, Kanai S, Date H. Automatic recognition of piping system from large-scale terrestrial laser scan data[J]. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2012, 38: 283-288.
[12]	Kim Y, Nguyen C H P, Choi Y. Automatic pipe and elbow recognition from three-dimensional point cloud model of industrial plant piping system using convolutional neural network-based primitive classification[J]. Automation in Construction, 2020, 116: 103236. doi: 10.1016/j.autcon.2020.103236
[13]	Nurunnabi A, Belton D, West G. Robust segmentation in laser scanning 3D point cloud data[C]//2012 International Conference on Digital Image Computing Techniques and Applications. Piscataway, US: IEEE, 2012: 1-8.
[14]	Kawashima K, Kanai S, Date H. Automatic recognition of a piping system from laser scanned points by eigenvalue analysis[J]. Key Engineering Materials, 2012, 523/524: 932-938.
[15]	Chaperon T, Goulette F. Extracting cylinders in full 3D data using a random sampling method and the Gaussian image[C]//Proceedings of the Vision Modeling and Visualization Conference 2001. New York: ACM, 2001: 35-42.
[16]	Lu Zhuheng, Mao Weiwei, Dai Yuewei, et al. Slicing-tracking-detection: simultaneous multi-cylinder detection from large-scale and complex point clouds[J]. IEEE Transactions on Visualization and Computer Graphics, 2022, 28(12): 4172-4185. doi: 10.1109/TVCG.2021.3082572
[17]	Liu Yongjin, Zhang Junbin, Hou Jichun, et al. Cylinder detection in large-scale point cloud of pipeline plant[J]. IEEE Transactions on Visualization and Computer Graphics, 2013, 19(10): 1700-1707. doi: 10.1109/TVCG.2013.74
[18]	Ahmed M F, Haas C T, Haas R. Automatic detection of cylindrical objects in built facilities[J]. Journal of Computing in Civil Engineering, 2014, 28(3): 04014009. doi: 10.1061/(ASCE)CP.1943-5487.0000329
[19]	Zhang Jie, Sun Junhua. 4-D spatiotemporal detection and modeling of free-bending pipelines in cluttered 3-D point cloud[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 5017612. doi: 10.1109/tim.2021.3115209
[20]	Yan Jieqiong, Zhou Laishui, Wang Jun, et al. Structural feature-preserving point cloud denoising method for aero-engine profile[J]. International Journal of Aerospace Engineering, 2022, 2022: 9565062. doi: 10.1155/2022/9565062
[21]	邢影, 宋涛, 赵延, 等. 3D-SIFT特征提取与体素滤波结合的点云精简算法[J]. 激光杂志, 2023, 44(3): 163-169. Xing Ying, Song Tao, Zhao Yan, et al. Point cloud reduction algorithm of 3D-SIFT feature point extraction and voxel filtering[J]. Laser Journal, 2023, 44(3): 163-169. (in Chinese doi: 10.14016/j.cnki.jgzz.2023.03.163 Xing Ying, Song Tao, Zhao Yan, et al. Point cloud reduction algorithm of 3D-SIFT feature point extraction and voxel filtering[J]. Laser Journal, 2023, 44(3): 163-169. (in Chinese) doi: 10.14016/j.cnki.jgzz.2023.03.163
[22]	Zhang Jie, Sun Zhengmao, Sun Junhua. 3-DFineRec: fine-grained recognition for small-scale objects in 3-D point cloud scenes[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 2500711. doi: 10.1109/tim.2021.3139685
[23]	张景蓉. 三维管道点云分割与识别技术研究[D]. 南京: 南京理工大学, 2019. Zhang Jingrong. Research on 3D Pipeline Point Cloud Segmentation and Recognition Technology [D]. Nanjing: Nanjing University of Science and Technology, 2019. (in Chinese Zhang Jingrong. Research on 3D Pipeline Point Cloud Segmentation and Recognition Technology [D]. Nanjing: Nanjing University of Science and Technology, 2019. (in Chinese)
[24]	Cao Tao, Zhou Laishui, Qu Liyu, et al. Aircraft pipe gap inspection on raw point cloud from a single view[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 5014814. doi: 10.1109/tim.2022.3186058
[25]	梅向明, 黄敬之. 微分几何[M]. 5版. 北京: 高等教育出版社, 2019. Mei Xiangming, Huang Jingzhi. Differential geometry[M]. 5th ed. Beijing: Higher Education Press, 2019. (in Chinese Mei Xiangming, Huang Jingzhi. Differential geometry[M]. 5th ed. Beijing: Higher Education Press, 2019. (in Chinese)