Industrial plant piping can be inspected dimensionally on site by registering terrestrial-laser-scan point clouds to the as-designed 3D CAD model (typically with iterative closest point registration), extracting pipe components (typically with RANSAC (random sample consensus) cylinder fitting), and reporting distance-based deviations and per-component geometric parameter errors such as orientation and radius. Nguyen and Choi (2018) demonstrated this pipeline on a real plant with a 91.60% component recognition rate, but the reported deviations mix genuine construction deviations with measurement error contributed by scanner noise (including laser light diffusion on hot thermal piping), occlusion from line-of-sight limits, registration error, and parameter-extraction error; they concluded that practical deployment requires an allowable error that incorporates input noise and algorithmic uncertainties. The open problem is to establish and validate such an allowable-error framework: a decomposition and uncertainty budget that quantifies each error source of the scan-to-CAD pipeline, propagates it into component-level error bounds on every reported geometric quantity, and yields a decision rule that attributes each CAD-versus-as-built mismatch to construction non-conformance or to measurement uncertainty under the noisy, occluded conditions of operating plants. Later literature has advanced automatic pipe recognition, as-built piping reconstruction, registration accuracy, and occlusion-aware scan planning, and uncertainty-budgeting methods exist for adjacent domains (bridges, spatial steel structures, manufacturing metrology), but no validated allowable-error framework for on-site piping scan-versus-CAD inspection has been published; the most recent scan-versus-BIM validation work still sets tolerance thresholds by practice-oriented approximation.

Background

Industrial process plants (refineries, chemical and power plants) contain dense networks of pipes whose positions, orientations, and radii must conform to the as-designed 3D CAD (computer-aided design) model within construction tolerances. Conventional dimensional inspection relies on manual measurement against drawings, which is slow, error-prone, and difficult at plant scale. Terrestrial laser scanning (TLS) — a ground-based instrument that captures millions of surface points by laser ranging, producing a point cloud — enables rapid as-built capture, and scan-versus-BIM comparison (BIM, building information modeling, denoting the semantically rich as-designed model) has become the standard paradigm for automated geometric quality inspection: the as-built point cloud is registered into the coordinate system of the as-designed model, and deviations between the two are computed. For piping, the reference pipeline is that of Nguyen and Choi (2018, Automation in Construction 91:44–52), extending their earlier parametric comparison method (Nguyen and Choi, 2018, Computer-Aided Design and Applications 15:238–246). The scan is registered to the CAD model with the iterative closest point algorithm (ICP), which iteratively rigidly transforms the scan to minimize distances between corresponding points. Pipe components are recognized by segmentation followed by cylinder fitting based on RANSAC (random sample consensus), a robust estimator that fits a cylinder — axis position, axis direction, and radius — to noisy points by repeatedly sampling minimal point sets. As-built condition is then assessed by two complementary metric families: distance-based deviation analysis (mean distance error and point distance error between scan points and the CAD surface) and geometric parameter comparison (orientation error and radius error per recognized pipe segment). On a real industrial plant dataset this pipeline recognized 91.60% of piping components, failing on occluded components (line-of-sight limits) and on thermally loaded piping where laser light diffusion produces noisy returns. Two long-standing strands frame what remains unresolved. First, point-cloud quality and tolerance checking: Rebolj et al. (2017) inverted the question for construction progress monitoring, deriving the point-cloud density and accuracy needed for reliable scan-vs-BIM recognition, and tolerance-based compliance checking has been demonstrated for fabricated pipe spools (Nahangi and Haas, 2014) and prefabricated MEP (mechanical, electrical, and plumbing) modules (Guo et al., 2020), with scan-based measurements benchmarked against a laser tracker (a portable coordinate-measurement instrument tracking a reflective target with sub-millimetric accuracy) in factory settings (Sharif et al., 2021). Second, measurement-uncertainty metrology: standards such as ISO 14253-1 prescribe decision rules for verifying conformance to a specification while accounting for measurement uncertainty, task-specific uncertainty determination for laser scanning has been developed in manufacturing metrology (Vlaeyen et al., 2023, 2022), statistical point-cloud comparison under uncertainty has been demonstrated for TLS-measured bridges (de Asís López et al., 2014), and an explicit two-part error budget separating scanner error from data-processing error was formulated for spatial steel structures (Liu et al., 2018). What none of these provides is what the piping inspection pipeline itself needs. On an operating plant, the scan is noisy (laser light diffuses on hot piping surfaces), partially occluded, registered with error, and processed with error-laden cylinder extraction, so each reported quantity — a distance error, an orientation error, a radius error — is the sum of a possible construction deviation and a stack of measurement and algorithmic errors of unknown magnitude. Without an allowable error that incorporates input noise and algorithmic uncertainties, a deviation report cannot distinguish a pipe that was genuinely installed out of tolerance from a measurement artifact, which blocks automated acceptance decisions on real plants.

Problem Statement

Establish and validate an allowable-error framework for on-site dimensional inspection of industrial plant piping systems by terrestrial-laser-scan point clouds compared against as-designed 3D CAD models. The framework must: (i) decompose the measurement error of the inspection pipeline into its sources — input point-cloud noise (including scanner- and surface-induced effects such as laser light diffusion on thermal piping), occlusion-induced partial coverage from line-of-sight limitations, iterative closest point (ICP) registration error, and random sample consensus (RANSAC) cylinder parameter-extraction error; (ii) propagate these sources, analytically or numerically, into quantitative uncertainty bounds for each reported geometric quantity (deviation distances, and per-component orientation and radius errors); and (iii) define an allowable-error decision rule, stated in terms of this uncertainty budget, that classifies each inspected piping component as genuinely non-conforming to the CAD model versus conforming within measurement uncertainty, so that observed model-versus-reality mismatches can be attributed to construction rather than to the measurement process. Validation must establish an explicit, falsifiable acceptance criterion for the framework — naming the comparison statistic and the coverage or confidence level at which independent reference measurements must fall within the predicted bounds — and must apply it separately under every source named in (i): input point-cloud noise (including laser light diffusion on thermal piping), occlusion-induced partial coverage from line-of-sight limits, ICP registration error, and RANSAC cylinder parameter-extraction error; the demonstration may use real on-site plant data where it is accessible, and otherwise synthetic or controlled scans that reproduce these error sources. The framework must also characterize how the budget responds to scan quality, including the effect of improved acquisition such as multi-position scanning. A resolution requires the decomposition (i), the propagated bounds (ii), the decision rule (iii), this validation applied under every source named in (i), and the scan-quality characterization.

References

  1. Nguyen, C. H. P., & Choi, Y. (2018). Comparison of point cloud data and 3D CAD data for on-site dimensional inspection of industrial plant piping systems. Automation in Construction, 91, 44–52. https://doi.org/10.1016/j.autcon.2018.03.008
  2. Nguyen, C. H. P., & Choi, Y. (2018). Parametric comparing for local inspection of industrial plants by using as-built model acquired from laser scan data. Computer-Aided Design and Applications, 15, 238–246. https://doi.org/10.1080/16864360.2017.1375675
  3. Kim, Y., Nguyen, C. H. P., & Choi, Y. (2020). Automatic pipe and elbow recognition from three-dimensional point cloud model of industrial plant piping system using convolutional neural network-based primitive classification. Automation in Construction, 116, 103236. https://doi.org/10.1016/j.autcon.2020.103236
  4. Xie, Y., Li, S., Liu, T., & Cai, Y. (2023). As-built BIM reconstruction of piping systems using PipeNet. Automation in Construction, 147, 104735. https://doi.org/10.1016/j.autcon.2022.104735
  5. Yu, W., Shu, J., Yang, Z., Ding, H., Zeng, W., & Bai, Y. (2025). Deep learning-based pipe segmentation and geometric reconstruction from poorly scanned point clouds using BIM-driven data alignment. Automation in Construction, 173, 106071. https://doi.org/10.1016/j.autcon.2025.106071
  6. Zhang, Y., Chen, L., Lu, Q., Zou, Y., Xiahou, X., Sølvsten, S., & Hancock, C. (2026). Semi-automated localised updating for as-built BIM of piping systems using point cloud data. Automation in Construction, 181, 106609. https://doi.org/10.1016/j.autcon.2025.106609
  7. Nahangi, M., & Haas, C. T. (2014). Automated 3D compliance checking in pipe spool fabrication. Advanced Engineering Informatics, 28, 360–369. https://doi.org/10.1016/j.aei.2014.04.001
  8. Guo, J., Wang, Q., & Park, J. (2020). Geometric quality inspection of prefabricated MEP modules with 3D laser scanning. Automation in Construction, 111, 103053. https://doi.org/10.1016/j.autcon.2019.103053
  9. Sharif, M., Rausch, C., Ndiongue, S., Haas, C., & Walbridge, S. (2021). Using 3D scanning for accurate estimation of termination points for dimensional quality assurance in pipe spool fabrication. International Journal of Industrialized Construction, 2, 54–69. https://doi.org/10.29173/ijic253
  10. Rebolj, D., Pučko, Z., Čuš Babič, N., Bizjak, M., & Mongus, D. (2017). Point cloud quality requirements for Scan-vs-BIM based automated construction progress monitoring. Automation in Construction, 84, 323–334. https://doi.org/10.1016/j.autcon.2017.09.021
  11. Liu, J., Zhang, Q., Wu, J., & Zhao, Y. (2018). Dimensional accuracy and structural performance assessment of spatial structure components using 3D laser scanning. Automation in Construction, 96, 324–336. https://doi.org/10.1016/j.autcon.2018.09.026
  12. de Asís López, F., Ordóñez, C., Roca-Pardiñas, J., & García-Cortés, S. (2014). Point cloud comparison under uncertainty. Application to beam bridge measurement with terrestrial laser scanning. Measurement, 51, 259–264. https://doi.org/10.1016/j.measurement.2014.02.013
  13. Vlaeyen, M., Haitjema, H., & Dewulf, W. (2023). Virtual task-specific measurement uncertainty determination for laser scanning. Precision Engineering, 80, 208–228. https://doi.org/10.1016/j.precisioneng.2022.12.008
  14. Vlaeyen, M., Haitjema, H., & Dewulf, W. (2022). Uncertainty-based autonomous path planning for laser line scanners. Metrology, 2, 479–494. https://doi.org/10.3390/metrology2040028
  15. Akiyama, J., Zong, Y., Shinada, N., Suzuki, T., & Amano, Y. (2023). High-Resolution Point Cloud Registration Method for Three-Dimensional Piping Measurements. Journal of Robotics and Mechatronics, 35, 1655–1662. https://doi.org/10.20965/jrm.2023.p1655
  16. Becker, F., Macher, H., Landes, T., & Kila, Y. (2026). Evaluation of ICP variants for point cloud/BIM alignment enabling Scan-vs-BIM comparison: Application to maritime construction tolerance verification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIX-B2-2026, 1229–1236. https://doi.org/10.5194/isprs-archives-xlix-b2-2026-1229-2026
  17. Alsuhaibani, A., & Leite, F. (2026). Scan-versus-BIM without Registration for As-Built Verification: Verifying MEP Sleeves ahead of Concrete Pours. Journal of Architectural Engineering, 32(2), 04026011. https://doi.org/10.1061/jaeied.aeeng-2114
  18. Zhu, J., Wang, D., Fu, Z., Zhang, Z., & Han, W. (2025). BIM-based terrestrial laser scanning path planning for large-scale civil infrastructure. Automation in Construction, 178, 106439. https://doi.org/10.1016/j.autcon.2025.106439
  19. Kinnen, T., & Blankenbach, J. (2026). Automated geometry-based analysis for object-level Scan-vs-BIM validation of TBE components. Journal of Information Technology in Construction (ITcon), 31, 75–105. https://doi.org/10.36680/j.itcon.2026.004
  20. Morgan, J., Mostafiz, R. B., & Friedland, C. (2025). Development of a decision support tool for use of 3D terrestrial laser scanning for piping designs. Frontiers in Built Environment, 11, 1663869. https://doi.org/10.3389/fbuil.2025.1663869
  21. British Standards Institution / ISO. (1999). Geometrical product specifications (GPS). Inspection by measurement of workpieces and measuring equipment — Decision rules for proving conformance or non-conformance with specifications (BS EN ISO 14253-1:1999, the British adoption of ISO 14253-1:1998). Standard. https://doi.org/10.3403/01722655